mentat/query-parser/src/parse.rs

1236 lines
48 KiB
Rust
Raw Normal View History

// Copyright 2016 Mozilla
//
// Licensed under the Apache License, Version 2.0 (the "License"); you may not use
// this file except in compliance with the License. You may obtain a copy of the
// License at http://www.apache.org/licenses/LICENSE-2.0
// Unless required by applicable law or agreed to in writing, software distributed
// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
// CONDITIONS OF ANY KIND, either express or implied. See the License for the
// specific language governing permissions and limitations under the License.
extern crate combine;
extern crate edn;
extern crate mentat_parser_utils;
extern crate mentat_query;
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
use std; // To refer to std::result::Result.
use std::collections::BTreeSet;
use self::combine::{
eof,
look_ahead,
many,
many1,
optional,
parser,
satisfy,
satisfy_map,
Parser,
ParseResult,
Stream,
};
use self::combine::combinator::{any, choice, or, try};
2018-05-31 19:13:28 +00:00
use errors::{
Error,
ErrorKind,
Result,
ResultExt,
};
use self::mentat_parser_utils::{
KeywordMapParser,
ResultParser,
ValueParseError,
};
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
use self::mentat_parser_utils::value_and_span::Stream as ValueStream;
use self::mentat_parser_utils::value_and_span::{
Item,
OfExactlyParsing,
forward_any_keyword,
forward_namespaced_keyword,
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
keyword_map,
list,
map,
seq,
vector,
};
use self::mentat_query::{
Aggregate,
Binding,
Direction,
Element,
FindQuery,
FindSpec,
FnArg,
FromValue,
Limit,
Order,
OrJoin,
OrWhereClause,
NamedPullAttribute,
Parse and Algebrize `not` & `not-join`. (#302) (Closes #303, #389, #422 ) r=rnewman * Part 1 - Parse `not` and `not-join` * Part 2 - Validate `not` and `not-join` pre-algebrization * Address review comments rnewman. * Remove `WhereNotClause` and populate `NotJoin` with `WhereClause`. * Fix validation for `not` and `not-join`, removing tests that were invalid. * Address rustification comments. * Rebase against `rust` branch. * Part 3 - Add required types for NotJoin. * Implement `PartialEq` for `ConjoiningClauses` so `ComputedTable` can be included inside `ColumnConstraint::NotExists` * Part 4 - Implement `apply_not_join` * Part 5 - Call `apply_not_join` from inside `apply_clause` * Part 6 - Translate `not-join` into `NOT EXISTS` SQL * Address review comments. * Rename `projected` to `unified` to better describe the fact that we are not projecting any variables. * Check for presence of each unified var in either `column_bindings` or `input_bindings` and bail if not there. * Copy over `input_bindings` for each var in `unified`. * Only copy over the first `column_binding` for each variable in `unified` rather than the whole list. * Update tests. * Address review comments. * Make output from Debug for NotExists more useful * Clear up misunderstanding. Any single failing clause in the not will cause the entire not to be considered empty * Address review comments. * Remove Limit requirement from cc_to_exists. * Use Entry.or_insert instead of matching on the entry to add to column_bindings. * Move addition of value_bindings to before apply_clauses on template. * Tidy up tests with some variable reuse. * Addressed nits, * Address review comments. * Move addition of column_bindings to above apply_clause. * Update tests. * Add test to ensure that unbound vars fail * Improve test for unbound variable to check for correct variable and error * address nits
2017-04-28 09:44:11 +00:00
NotJoin,
Pattern,
PatternNonValuePlace,
PatternValuePlace,
Predicate,
Pull,
PullAttributeSpec,
PullConcreteAttribute,
QueryFunction,
SrcVar,
TypeAnnotation,
UnifyVars,
Variable,
VariableOrPlaceholder,
WhereClause,
WhereFn,
};
pub struct Query<'a>(std::marker::PhantomData<&'a ()>);
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Query, variable, Variable, {
satisfy_map(Variable::from_value)
});
def_parser!(Query, keyword, edn::Keyword, {
satisfy_map(|v: &edn::ValueAndSpan| v.inner.as_keyword().cloned())
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Query, source_var, SrcVar, {
satisfy_map(SrcVar::from_value)
});
// TODO: interning.
def_parser!(Query, query_function, QueryFunction, {
satisfy_map(QueryFunction::from_value)
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Query, fn_arg, FnArg, {
satisfy_map(FnArg::from_value).or(vector().of_exactly(many::<Vec<FnArg>, _>(Query::fn_arg())).map(FnArg::Vector))
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Query, arguments, Vec<FnArg>, {
(many::<Vec<FnArg>, _>(Query::fn_arg()))
});
def_parser!(Query, direction, Direction, {
satisfy_map(|v: &edn::ValueAndSpan| {
match v.inner {
edn::SpannedValue::PlainSymbol(ref s) => {
let name = s.0.as_str();
match name {
"asc" => Some(Direction::Ascending),
"desc" => Some(Direction::Descending),
_ => None,
}
},
_ => None,
}
})
});
def_parser!(Query, order, Order, {
seq().of_exactly((Query::direction(), Query::variable()))
.map(|(d, v)| Order(d, v))
.or(Query::variable().map(|v| Order(Direction::Ascending, v)))
});
def_matches_plain_symbol!(Query, the, "the");
def_matches_plain_symbol!(Query, pull, "pull");
def_matches_plain_symbol!(Query, wildcard, "*");
def_matches_keyword!(Query, alias_as, "as");
pub struct Where<'a>(std::marker::PhantomData<&'a ()>);
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Where, pattern_value_place, PatternValuePlace, {
satisfy_map(PatternValuePlace::from_value)
});
def_parser!(Query, natural_number, u64, {
any().and_then(|v: &edn::ValueAndSpan| {
match v.inner {
edn::SpannedValue::Integer(x) if (x > 0) => {
Ok(x as u64)
},
ref spanned => {
let e = Box::new(Error::from_kind(ErrorKind::InvalidLimit(spanned.clone().into())));
Err(combine::primitives::Error::Other(e))
},
}
})
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Where, pattern_non_value_place, PatternNonValuePlace, {
satisfy_map(PatternNonValuePlace::from_value)
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_matches_plain_symbol!(Where, and, "and");
def_matches_plain_symbol!(Where, or, "or");
def_matches_plain_symbol!(Where, or_join, "or-join");
Parse and Algebrize `not` & `not-join`. (#302) (Closes #303, #389, #422 ) r=rnewman * Part 1 - Parse `not` and `not-join` * Part 2 - Validate `not` and `not-join` pre-algebrization * Address review comments rnewman. * Remove `WhereNotClause` and populate `NotJoin` with `WhereClause`. * Fix validation for `not` and `not-join`, removing tests that were invalid. * Address rustification comments. * Rebase against `rust` branch. * Part 3 - Add required types for NotJoin. * Implement `PartialEq` for `ConjoiningClauses` so `ComputedTable` can be included inside `ColumnConstraint::NotExists` * Part 4 - Implement `apply_not_join` * Part 5 - Call `apply_not_join` from inside `apply_clause` * Part 6 - Translate `not-join` into `NOT EXISTS` SQL * Address review comments. * Rename `projected` to `unified` to better describe the fact that we are not projecting any variables. * Check for presence of each unified var in either `column_bindings` or `input_bindings` and bail if not there. * Copy over `input_bindings` for each var in `unified`. * Only copy over the first `column_binding` for each variable in `unified` rather than the whole list. * Update tests. * Address review comments. * Make output from Debug for NotExists more useful * Clear up misunderstanding. Any single failing clause in the not will cause the entire not to be considered empty * Address review comments. * Remove Limit requirement from cc_to_exists. * Use Entry.or_insert instead of matching on the entry to add to column_bindings. * Move addition of value_bindings to before apply_clauses on template. * Tidy up tests with some variable reuse. * Addressed nits, * Address review comments. * Move addition of column_bindings to above apply_clause. * Update tests. * Add test to ensure that unbound vars fail * Improve test for unbound variable to check for correct variable and error * address nits
2017-04-28 09:44:11 +00:00
def_matches_plain_symbol!(Where, not, "not");
def_matches_plain_symbol!(Where, not_join, "not-join");
def_matches_plain_symbol!(Where, type_symbol, "type");
def_parser!(Where, rule_vars, BTreeSet<Variable>, {
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
seq()
.of_exactly(many1(Query::variable()).and_then(unique_vars))
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Where, or_pattern_clause, OrWhereClause, {
Where::clause().map(|clause| OrWhereClause::Clause(clause))
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Where, or_and_clause, OrWhereClause, {
seq()
.of_exactly(Where::and()
.with(many1(Where::clause()))
.map(OrWhereClause::And))
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Where, or_where_clause, OrWhereClause, {
choice([Where::or_pattern_clause(), Where::or_and_clause()])
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Where, or_clause, WhereClause, {
seq()
.of_exactly(Where::or()
.with(many1(Where::or_where_clause()))
.map(|clauses| {
2017-04-04 21:54:08 +00:00
WhereClause::OrJoin(OrJoin::new(UnifyVars::Implicit, clauses))
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
}))
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Where, or_join_clause, WhereClause, {
seq()
.of_exactly(Where::or_join()
.with(Where::rule_vars())
.and(many1(Where::or_where_clause()))
.map(|(vars, clauses)| {
2017-04-04 21:54:08 +00:00
WhereClause::OrJoin(OrJoin::new(UnifyVars::Explicit(vars), clauses))
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
}))
});
Parse and Algebrize `not` & `not-join`. (#302) (Closes #303, #389, #422 ) r=rnewman * Part 1 - Parse `not` and `not-join` * Part 2 - Validate `not` and `not-join` pre-algebrization * Address review comments rnewman. * Remove `WhereNotClause` and populate `NotJoin` with `WhereClause`. * Fix validation for `not` and `not-join`, removing tests that were invalid. * Address rustification comments. * Rebase against `rust` branch. * Part 3 - Add required types for NotJoin. * Implement `PartialEq` for `ConjoiningClauses` so `ComputedTable` can be included inside `ColumnConstraint::NotExists` * Part 4 - Implement `apply_not_join` * Part 5 - Call `apply_not_join` from inside `apply_clause` * Part 6 - Translate `not-join` into `NOT EXISTS` SQL * Address review comments. * Rename `projected` to `unified` to better describe the fact that we are not projecting any variables. * Check for presence of each unified var in either `column_bindings` or `input_bindings` and bail if not there. * Copy over `input_bindings` for each var in `unified`. * Only copy over the first `column_binding` for each variable in `unified` rather than the whole list. * Update tests. * Address review comments. * Make output from Debug for NotExists more useful * Clear up misunderstanding. Any single failing clause in the not will cause the entire not to be considered empty * Address review comments. * Remove Limit requirement from cc_to_exists. * Use Entry.or_insert instead of matching on the entry to add to column_bindings. * Move addition of value_bindings to before apply_clauses on template. * Tidy up tests with some variable reuse. * Addressed nits, * Address review comments. * Move addition of column_bindings to above apply_clause. * Update tests. * Add test to ensure that unbound vars fail * Improve test for unbound variable to check for correct variable and error * address nits
2017-04-28 09:44:11 +00:00
def_parser!(Where, not_clause, WhereClause, {
seq()
.of_exactly(Where::not()
.with(many1(Where::clause()))
.map(|clauses| {
WhereClause::NotJoin(
NotJoin {
unify_vars: UnifyVars::Implicit,
clauses: clauses,
})
}))
});
def_parser!(Where, not_join_clause, WhereClause, {
seq()
.of_exactly(Where::not_join()
.with(Where::rule_vars())
.and(many1(Where::clause()))
.map(|(vars, clauses)| {
WhereClause::NotJoin(
NotJoin {
unify_vars: UnifyVars::Explicit(vars),
clauses: clauses,
})
}))
});
def_parser!(Query, func, (QueryFunction, Vec<FnArg>), {
(Query::query_function(), Query::arguments())
});
def_parser!(Query, aggregate, Aggregate, {
Query::func()
.map(|(func, args)| Aggregate {
func, args,
})
});
def_parser!(Query, pull_concrete_attribute_ident, PullConcreteAttribute, {
forward_namespaced_keyword()
.map(|k| PullConcreteAttribute::Ident(::std::rc::Rc::new(k.clone())))
});
def_parser!(Query, pull_concrete_attribute, PullAttributeSpec, {
(Query::pull_concrete_attribute_ident(),
optional(try(Query::alias_as()
.with(forward_any_keyword()
.map(|alias| ::std::rc::Rc::new(alias.clone()))))))
.map(|(attribute, alias)|
PullAttributeSpec::Attribute(
NamedPullAttribute {
attribute,
alias: alias,
}))
});
def_parser!(Query, pull_wildcard_attribute, PullAttributeSpec, {
Query::wildcard().map(|_| PullAttributeSpec::Wildcard)
});
def_parser!(Query, pull_attribute, PullAttributeSpec, {
choice([
try(Query::pull_concrete_attribute()),
try(Query::pull_wildcard_attribute()),
// TODO: reversed keywords, entids (with aliases, presumably…).
])
});
// A wildcard can appear only once.
// If a wildcard appears, only map expressions can be present.
fn validate_attributes<'a, I>(attrs: I) -> std::result::Result<(), &'static str>
where I: IntoIterator<Item=&'a PullAttributeSpec> {
let mut wildcard_seen = false;
let mut non_map_or_wildcard_seen = false;
for attr in attrs {
match attr {
&PullAttributeSpec::Wildcard => {
if wildcard_seen {
return Err("duplicate wildcard pull attribute");
}
wildcard_seen = true;
if non_map_or_wildcard_seen {
return Err("wildcard with specified attributes");
}
},
// &PullAttributeSpec::LimitedAttribute(_, _) => {
&PullAttributeSpec::Attribute(_) => {
non_map_or_wildcard_seen = true;
if wildcard_seen {
return Err("wildcard with specified attributes");
}
},
// TODO: map form.
}
}
Ok(())
}
def_parser!(Query, pull_attributes, Vec<PullAttributeSpec>, {
vector().of_exactly(many1(Query::pull_attribute()))
.and_then(|attrs: Vec<PullAttributeSpec>|
validate_attributes(&attrs)
.and(Ok(attrs))
.map_err(|e| combine::primitives::Error::Unexpected(e.into())))
});
/// A vector containing just a parenthesized filter expression.
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Where, pred, WhereClause, {
// Accept either a nested list or a nested vector here:
// `[(foo ?x ?y)]` or `[[foo ?x ?y]]`
vector()
.of_exactly(seq()
.of_exactly(Query::func()
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
.map(|(f, args)| {
WhereClause::Pred(
Predicate {
operator: f.0,
args: args,
})
})))
});
/// A type annotation.
def_parser!(Where, type_annotation, WhereClause, {
// Accept either a nested list or a nested vector here:
// `[(type ?x :db.type/string)]` or `[[type ?x :db.type/long]]`
vector()
.of_exactly(seq()
.of_exactly((Where::type_symbol(), Query::variable(), Query::keyword())
.map(|(_, var, ty)| {
WhereClause::TypeAnnotation(
TypeAnnotation {
value_type: ty,
variable: var,
})
})))
});
/// A vector containing a parenthesized function expression and a binding.
def_parser!(Where, where_fn, WhereClause, {
// Accept either a nested list or a nested vector here:
// `[(foo ?x ?y) binding]` or `[[foo ?x ?y] binding]`
vector()
.of_exactly(
(seq().of_exactly(
Query::func()),
Bind::binding())
.map(|((f, args), binding)| {
WhereClause::WhereFn(
WhereFn {
operator: f.0,
args: args,
binding: binding,
})
}))
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Where, pattern, WhereClause, {
vector()
.of_exactly(
// While *technically* Datomic allows you to have a query like:
// [:find … :where [[?x]]]
// We don't -- we require at least e, a.
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
(optional(Query::source_var()), // src
Where::pattern_non_value_place(), // e
Where::pattern_non_value_place(), // a
optional(Where::pattern_value_place()), // v
optional(Where::pattern_non_value_place())) // tx
.and_then(|(src, e, a, v, tx)| {
let v = v.unwrap_or(PatternValuePlace::Placeholder);
let tx = tx.unwrap_or(PatternNonValuePlace::Placeholder);
// Pattern::new takes care of reversal of reversed
// attributes: [?x :foo/_bar ?y] turns into
// [?y :foo/bar ?x].
//
// This is a bit messy: the inner conversion to a Pattern can
// fail if the input is something like
//
// ```edn
// [?x :foo/_reversed 23.4]
// ```
//
// because
//
// ```edn
// [23.4 :foo/reversed ?x]
// ```
//
// is nonsense. That leaves us with a nested optional, which we unwrap here.
Pattern::new(src, e, a, v, tx)
.map(WhereClause::Pattern)
.ok_or(combine::primitives::Error::Expected("pattern".into()))
}))
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Where, clause, WhereClause, {
choice([try(Where::pattern()),
// It's either
// (or-join [vars] clauses…)
// or
// (or clauses…)
// We don't yet handle source vars.
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
try(Where::or_join_clause()),
try(Where::or_clause()),
Parse and Algebrize `not` & `not-join`. (#302) (Closes #303, #389, #422 ) r=rnewman * Part 1 - Parse `not` and `not-join` * Part 2 - Validate `not` and `not-join` pre-algebrization * Address review comments rnewman. * Remove `WhereNotClause` and populate `NotJoin` with `WhereClause`. * Fix validation for `not` and `not-join`, removing tests that were invalid. * Address rustification comments. * Rebase against `rust` branch. * Part 3 - Add required types for NotJoin. * Implement `PartialEq` for `ConjoiningClauses` so `ComputedTable` can be included inside `ColumnConstraint::NotExists` * Part 4 - Implement `apply_not_join` * Part 5 - Call `apply_not_join` from inside `apply_clause` * Part 6 - Translate `not-join` into `NOT EXISTS` SQL * Address review comments. * Rename `projected` to `unified` to better describe the fact that we are not projecting any variables. * Check for presence of each unified var in either `column_bindings` or `input_bindings` and bail if not there. * Copy over `input_bindings` for each var in `unified`. * Only copy over the first `column_binding` for each variable in `unified` rather than the whole list. * Update tests. * Address review comments. * Make output from Debug for NotExists more useful * Clear up misunderstanding. Any single failing clause in the not will cause the entire not to be considered empty * Address review comments. * Remove Limit requirement from cc_to_exists. * Use Entry.or_insert instead of matching on the entry to add to column_bindings. * Move addition of value_bindings to before apply_clauses on template. * Tidy up tests with some variable reuse. * Addressed nits, * Address review comments. * Move addition of column_bindings to above apply_clause. * Update tests. * Add test to ensure that unbound vars fail * Improve test for unbound variable to check for correct variable and error * address nits
2017-04-28 09:44:11 +00:00
try(Where::not_join_clause()),
try(Where::not_clause()),
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
try(Where::type_annotation()),
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
try(Where::pred()),
try(Where::where_fn()),
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
])
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Where, clauses, Vec<WhereClause>, {
// Right now we only support patterns and predicates. See #239 for more.
(many1::<Vec<WhereClause>, _>(Where::clause()))
});
pub struct Find<'a>(std::marker::PhantomData<&'a ()>);
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_matches_plain_symbol!(Find, period, ".");
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_matches_plain_symbol!(Find, ellipsis, "...");
def_matches_plain_symbol!(Find, placeholder, "_");
def_parser!(Find, variable_element, Element, {
Query::variable().map(Element::Variable)
});
def_parser!(Find, corresponding_element, Element, {
Query::the().with(Query::variable())
.map(Element::Corresponding)
});
def_parser!(Find, aggregate_element, Element, {
Query::aggregate().map(Element::Aggregate)
});
def_parser!(Find, pull_element, Element, {
Query::pull().with(Query::variable().and(Query::pull_attributes()))
.map(|(var, attrs)| Element::Pull(Pull { var: var, patterns: attrs }))
});
enum ElementType {
Corresponding,
Pull,
Aggregate,
}
def_parser!(Find, seq_elem, Element, {
let element_parser_for_type = |ty: ElementType| {
match ty {
ElementType::Corresponding => Find::corresponding_element(),
ElementType::Pull => Find::pull_element(),
ElementType::Aggregate => Find::aggregate_element(),
}
};
// This slightly tortured phrasing ensures that we don't consume
// when the first item in the list -- the function name -- doesn't
// match, but once we decide what the list is, we go ahead and
// commit to that branch.
seq().of_exactly(
// This comes first because otherwise (the ?x) will match as an aggregate.
look_ahead(Query::the()).map(|_| ElementType::Corresponding)
// Similarly, we have to parse pull before general functions.
.or(look_ahead(Query::pull()).map(|_| ElementType::Pull))
.or(look_ahead(Query::func()).map(|_| ElementType::Aggregate))
.then(element_parser_for_type))
});
def_parser!(Find, elem, Element, {
try(Find::variable_element())
.or(Find::seq_elem())
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Find, find_scalar, FindSpec, {
Find::elem().skip(Find::period())
.map(FindSpec::FindScalar)
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Find, find_coll, FindSpec, {
vector().of_exactly(Find::elem().skip(Find::ellipsis()))
.map(FindSpec::FindColl)
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Find, elements, Vec<Element>, {
many1::<Vec<Element>, _>(Find::elem())
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Find, find_rel, FindSpec, {
Find::elements().map(FindSpec::FindRel)
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
def_parser!(Find, find_tuple, FindSpec, {
vector().of_exactly(Find::elements())
.map(FindSpec::FindTuple)
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
/// Parse a stream of values into one of four find specs.
///
/// `:find` must be an array of plain var symbols (?foo), pull expressions, and aggregates. For now
/// we only support variables and the annotations necessary to declare which flavor of :find we
/// want:
///
///
/// `?x ?y ?z ` = FindRel
/// `[?x ...] ` = FindColl
/// `?x . ` = FindScalar
/// `[?x ?y ?z]` = FindTuple
def_parser!(Find, spec, FindSpec, {
// Any one of the four specs might apply, so we combine them with `choice`. Our parsers consume
// input, so we need to wrap them in `try` so that they operate independently.
choice::<[&mut Parser<Input = _, Output = FindSpec>; 4], _>
([&mut try(Find::find_scalar()),
&mut try(Find::find_coll()),
&mut try(Find::find_tuple()),
&mut try(Find::find_rel())])
});
fn unique_vars<T, E>(vars: Vec<Variable>) -> std::result::Result<BTreeSet<Variable>, combine::primitives::Error<T, E>> {
let given = vars.len();
let set: BTreeSet<Variable> = vars.into_iter().collect();
if given != set.len() {
// TODO: find out what the variable is!
let e = Box::new(Error::from_kind(ErrorKind::DuplicateVariableError));
Err(combine::primitives::Error::Other(e))
} else {
Ok(set)
}
}
def_parser!(Find, vars, BTreeSet<Variable>, {
many(Query::variable()).and_then(unique_vars)
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
/// This is awkward, but will do for now. We use `keyword_map()` to optionally accept vector find
/// queries, then we use `FindQueryPart` to collect parts that have heterogeneous types; and then we
/// construct a `FindQuery` from them.
def_parser!(Find, query, FindQuery, {
let find_map = keyword_map_of!(
("find", Find::spec()),
("in", Find::vars()),
("limit", Query::variable().map(Limit::Variable).or(Query::natural_number().map(Limit::Fixed))),
("order", many1(Query::order())),
("where", Where::clauses()),
("with", Find::vars()) // Note: no trailing comma allowed!
);
(or(keyword_map(), vector()))
.of_exactly(find_map)
.and_then(|(find_spec, in_vars, limit, order_clauses, where_clauses, with_vars) | -> std::result::Result<FindQuery, combine::primitives::Error<&edn::ValueAndSpan, &edn::ValueAndSpan>> {
let limit = limit.unwrap_or(Limit::None);
// Make sure that if we have `:limit ?x`, `?x` appears in `:in`.
let in_vars = in_vars.unwrap_or(BTreeSet::default());
if let Limit::Variable(ref v) = limit {
if !in_vars.contains(v) {
let e = Box::new(Error::from_kind(ErrorKind::UnknownLimitVar(v.name())));
return Err(combine::primitives::Error::Other(e));
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
}
}
Ok(FindQuery {
default_source: SrcVar::DefaultSrc,
find_spec: find_spec.ok_or(combine::primitives::Error::Unexpected("expected :find".into()))?,
in_sources: BTreeSet::default(), // TODO
in_vars: in_vars,
limit: limit,
order: order_clauses,
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
where_clauses: where_clauses.ok_or(combine::primitives::Error::Unexpected("expected :where".into()))?,
with: with_vars.unwrap_or(BTreeSet::default()),
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
})
})
});
pub struct Bind<'a>(std::marker::PhantomData<&'a ()>);
def_parser!(Bind, bind_scalar, Binding, {
Query::variable()
.skip(eof())
.map(|var| Binding::BindScalar(var))
});
def_parser!(Bind, variable_or_placeholder, VariableOrPlaceholder, {
Query::variable().map(VariableOrPlaceholder::Variable)
.or(Find::placeholder().map(|_| VariableOrPlaceholder::Placeholder))
});
def_parser!(Bind, bind_coll, Binding, {
vector()
.of_exactly(Query::variable()
.skip(Find::ellipsis()))
.map(Binding::BindColl)
});
def_parser!(Bind, bind_rel, Binding, {
vector().of_exactly(
vector().of_exactly(
many1::<Vec<VariableOrPlaceholder>, _>(Bind::variable_or_placeholder())
.map(Binding::BindRel)))
});
def_parser!(Bind, bind_tuple, Binding, {
vector().of_exactly(
many1::<Vec<VariableOrPlaceholder>, _>(Bind::variable_or_placeholder())
.map(Binding::BindTuple))
});
def_parser!(Bind, binding, Binding, {
// Any one of the four binding types might apply, so we combine them with `choice`. Our parsers
// consume input, so we need to wrap them in `try` so that they operate independently.
choice([try(Bind::bind_scalar()),
try(Bind::bind_coll()),
try(Bind::bind_tuple()),
try(Bind::bind_rel())])
});
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
pub fn parse_find_string(string: &str) -> Result<FindQuery> {
let expr = edn::parse::value(string)?;
Find::query()
.parse(expr.atom_stream())
.map(|x| x.0)
.map_err(|e| Error::from_kind(ErrorKind::FindParseError(e.into())))
}
#[cfg(test)]
mod test {
extern crate combine;
extern crate edn;
extern crate mentat_query;
use std::rc::Rc;
use self::combine::Parser;
use self::edn::OrderedFloat;
use self::mentat_query::{
Binding,
Element,
FindSpec,
NonIntegerConstant,
Pattern,
PatternNonValuePlace,
PatternValuePlace,
SrcVar,
Variable,
VariableOrPlaceholder,
};
use super::*;
fn variable(x: edn::PlainSymbol) -> Variable {
Variable(Rc::new(x))
}
fn ident_kw(kw: edn::Keyword) -> PatternNonValuePlace {
PatternNonValuePlace::Ident(kw.into())
}
fn ident(ns: &str, name: &str) -> PatternNonValuePlace {
ident_kw(edn::Keyword::namespaced(ns, name))
}
#[test]
fn test_pattern_mixed() {
let e = edn::PlainSymbol::plain("_");
let a = edn::Keyword::namespaced("foo", "bar");
let v = OrderedFloat(99.9);
let tx = edn::PlainSymbol::plain("?tx");
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
let input = edn::Value::Vector(vec!(edn::Value::PlainSymbol(e.clone()),
edn::Value::Keyword(a.clone()),
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
edn::Value::Float(v.clone()),
edn::Value::PlainSymbol(tx.clone())));
assert_parses_to!(Where::pattern, input, WhereClause::Pattern(Pattern {
source: None,
entity: PatternNonValuePlace::Placeholder,
attribute: ident_kw(a),
value: PatternValuePlace::Constant(NonIntegerConstant::Float(v)),
tx: PatternNonValuePlace::Variable(variable(tx)),
}));
}
#[test]
fn test_pattern_vars() {
let s = edn::PlainSymbol::plain("$x");
let e = edn::PlainSymbol::plain("?e");
let a = edn::PlainSymbol::plain("?a");
let v = edn::PlainSymbol::plain("?v");
let tx = edn::PlainSymbol::plain("?tx");
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
let input = edn::Value::Vector(vec!(edn::Value::PlainSymbol(s.clone()),
edn::Value::PlainSymbol(e.clone()),
edn::Value::PlainSymbol(a.clone()),
edn::Value::PlainSymbol(v.clone()),
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
edn::Value::PlainSymbol(tx.clone())));
assert_parses_to!(Where::pattern, input, WhereClause::Pattern(Pattern {
source: Some(SrcVar::NamedSrc("x".to_string())),
entity: PatternNonValuePlace::Variable(variable(e)),
attribute: PatternNonValuePlace::Variable(variable(a)),
value: PatternValuePlace::Variable(variable(v)),
tx: PatternNonValuePlace::Variable(variable(tx)),
}));
}
#[test]
fn test_pattern_reversed_invalid() {
let e = edn::PlainSymbol::plain("_");
let a = edn::Keyword::namespaced("foo", "_bar");
let v = OrderedFloat(99.9);
let tx = edn::PlainSymbol::plain("?tx");
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
let input = edn::Value::Vector(vec!(edn::Value::PlainSymbol(e.clone()),
edn::Value::Keyword(a.clone()),
edn::Value::Float(v.clone()),
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
edn::Value::PlainSymbol(tx.clone())));
let input = input.with_spans();
let mut par = Where::pattern();
let result = par.parse(input.atom_stream());
match result {
Err(_) => (),
_ => assert!(false, "Expected a parse error"),
}
}
#[test]
fn test_pattern_reversed() {
let e = edn::PlainSymbol::plain("_");
let a = edn::Keyword::namespaced("foo", "_bar");
let v = edn::PlainSymbol::plain("?v");
let tx = edn::PlainSymbol::plain("?tx");
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
let input = edn::Value::Vector(vec!(edn::Value::PlainSymbol(e.clone()),
edn::Value::Keyword(a.clone()),
edn::Value::PlainSymbol(v.clone()),
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
edn::Value::PlainSymbol(tx.clone())));
// Note that the attribute is no longer reversed, and the entity and value have
// switched places.
assert_parses_to!(Where::pattern, input, WhereClause::Pattern(Pattern {
source: None,
entity: PatternNonValuePlace::Variable(variable(v)),
attribute: ident("foo", "bar"),
value: PatternValuePlace::Placeholder,
tx: PatternNonValuePlace::Variable(variable(tx)),
}));
}
#[test]
fn test_rule_vars() {
let e = edn::PlainSymbol::plain("?e");
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
let input = edn::Value::Vector(vec![edn::Value::PlainSymbol(e.clone())]);
assert_parses_to!(Where::rule_vars, input,
btreeset!{variable(e.clone())});
}
#[test]
fn test_repeated_vars() {
let e = edn::PlainSymbol::plain("?e");
let f = edn::PlainSymbol::plain("?f");
let input = edn::Value::Vector(vec![edn::Value::PlainSymbol(e.clone()),
edn::Value::PlainSymbol(f.clone()),]);
assert_parses_to!(|| vector().of_exactly(Find::vars()), input,
btreeset!{variable(e.clone()), variable(f.clone())});
let g = edn::PlainSymbol::plain("?g");
let input = edn::Value::Vector(vec![edn::Value::PlainSymbol(g.clone()),
edn::Value::PlainSymbol(g.clone()),]);
let input = input.with_spans();
let mut par = vector().of_exactly(Find::vars());
let result = par.parse(input.atom_stream())
.map(|x| x.0)
.map_err(|e| if let Some(combine::primitives::Error::Other(x)) = e.errors.into_iter().next() {
// Pattern matching on boxes is rocket science until Rust Nightly features hit
// stable. ErrorKind isn't Clone, so convert to strings. We could pattern match
// for exact comparison here.
x.downcast::<Error>().ok().map(|e| e.to_string())
} else {
None
});
assert_eq!(result, Err(Some("duplicates in variable list".to_string())));
}
#[test]
fn test_or() {
let oj = edn::PlainSymbol::plain("or");
let e = edn::PlainSymbol::plain("?e");
let a = edn::PlainSymbol::plain("?a");
let v = edn::PlainSymbol::plain("?v");
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
let input = edn::Value::List(
vec![edn::Value::PlainSymbol(oj),
edn::Value::Vector(vec![edn::Value::PlainSymbol(e.clone()),
edn::Value::PlainSymbol(a.clone()),
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
edn::Value::PlainSymbol(v.clone())])].into_iter().collect());
assert_parses_to!(Where::or_clause, input,
WhereClause::OrJoin(
2017-04-04 21:54:08 +00:00
OrJoin::new(UnifyVars::Implicit,
vec![OrWhereClause::Clause(
WhereClause::Pattern(Pattern {
source: None,
entity: PatternNonValuePlace::Variable(variable(e)),
attribute: PatternNonValuePlace::Variable(variable(a)),
value: PatternValuePlace::Variable(variable(v)),
tx: PatternNonValuePlace::Placeholder,
}))])));
}
#[test]
fn test_or_join() {
let oj = edn::PlainSymbol::plain("or-join");
let e = edn::PlainSymbol::plain("?e");
let a = edn::PlainSymbol::plain("?a");
let v = edn::PlainSymbol::plain("?v");
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
let input = edn::Value::List(
vec![edn::Value::PlainSymbol(oj),
edn::Value::Vector(vec![edn::Value::PlainSymbol(e.clone())]),
edn::Value::Vector(vec![edn::Value::PlainSymbol(e.clone()),
edn::Value::PlainSymbol(a.clone()),
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
edn::Value::PlainSymbol(v.clone())])].into_iter().collect());
assert_parses_to!(Where::or_join_clause, input,
WhereClause::OrJoin(
OrJoin::new(UnifyVars::Explicit(btreeset!{variable(e.clone())}),
2017-04-04 21:54:08 +00:00
vec![OrWhereClause::Clause(
WhereClause::Pattern(Pattern {
source: None,
entity: PatternNonValuePlace::Variable(variable(e)),
attribute: PatternNonValuePlace::Variable(variable(a)),
value: PatternValuePlace::Variable(variable(v)),
tx: PatternNonValuePlace::Placeholder,
}))])));
}
Parse and Algebrize `not` & `not-join`. (#302) (Closes #303, #389, #422 ) r=rnewman * Part 1 - Parse `not` and `not-join` * Part 2 - Validate `not` and `not-join` pre-algebrization * Address review comments rnewman. * Remove `WhereNotClause` and populate `NotJoin` with `WhereClause`. * Fix validation for `not` and `not-join`, removing tests that were invalid. * Address rustification comments. * Rebase against `rust` branch. * Part 3 - Add required types for NotJoin. * Implement `PartialEq` for `ConjoiningClauses` so `ComputedTable` can be included inside `ColumnConstraint::NotExists` * Part 4 - Implement `apply_not_join` * Part 5 - Call `apply_not_join` from inside `apply_clause` * Part 6 - Translate `not-join` into `NOT EXISTS` SQL * Address review comments. * Rename `projected` to `unified` to better describe the fact that we are not projecting any variables. * Check for presence of each unified var in either `column_bindings` or `input_bindings` and bail if not there. * Copy over `input_bindings` for each var in `unified`. * Only copy over the first `column_binding` for each variable in `unified` rather than the whole list. * Update tests. * Address review comments. * Make output from Debug for NotExists more useful * Clear up misunderstanding. Any single failing clause in the not will cause the entire not to be considered empty * Address review comments. * Remove Limit requirement from cc_to_exists. * Use Entry.or_insert instead of matching on the entry to add to column_bindings. * Move addition of value_bindings to before apply_clauses on template. * Tidy up tests with some variable reuse. * Addressed nits, * Address review comments. * Move addition of column_bindings to above apply_clause. * Update tests. * Add test to ensure that unbound vars fail * Improve test for unbound variable to check for correct variable and error * address nits
2017-04-28 09:44:11 +00:00
#[test]
fn test_not() {
let e = edn::PlainSymbol::plain("?e");
let a = edn::PlainSymbol::plain("?a");
let v = edn::PlainSymbol::plain("?v");
Parse and Algebrize `not` & `not-join`. (#302) (Closes #303, #389, #422 ) r=rnewman * Part 1 - Parse `not` and `not-join` * Part 2 - Validate `not` and `not-join` pre-algebrization * Address review comments rnewman. * Remove `WhereNotClause` and populate `NotJoin` with `WhereClause`. * Fix validation for `not` and `not-join`, removing tests that were invalid. * Address rustification comments. * Rebase against `rust` branch. * Part 3 - Add required types for NotJoin. * Implement `PartialEq` for `ConjoiningClauses` so `ComputedTable` can be included inside `ColumnConstraint::NotExists` * Part 4 - Implement `apply_not_join` * Part 5 - Call `apply_not_join` from inside `apply_clause` * Part 6 - Translate `not-join` into `NOT EXISTS` SQL * Address review comments. * Rename `projected` to `unified` to better describe the fact that we are not projecting any variables. * Check for presence of each unified var in either `column_bindings` or `input_bindings` and bail if not there. * Copy over `input_bindings` for each var in `unified`. * Only copy over the first `column_binding` for each variable in `unified` rather than the whole list. * Update tests. * Address review comments. * Make output from Debug for NotExists more useful * Clear up misunderstanding. Any single failing clause in the not will cause the entire not to be considered empty * Address review comments. * Remove Limit requirement from cc_to_exists. * Use Entry.or_insert instead of matching on the entry to add to column_bindings. * Move addition of value_bindings to before apply_clauses on template. * Tidy up tests with some variable reuse. * Addressed nits, * Address review comments. * Move addition of column_bindings to above apply_clause. * Update tests. * Add test to ensure that unbound vars fail * Improve test for unbound variable to check for correct variable and error * address nits
2017-04-28 09:44:11 +00:00
assert_edn_parses_to!(Where::not_clause,
"(not [?e ?a ?v])",
WhereClause::NotJoin(
NotJoin {
unify_vars: UnifyVars::Implicit,
clauses: vec![
WhereClause::Pattern(Pattern {
source: None,
entity: PatternNonValuePlace::Variable(variable(e)),
attribute: PatternNonValuePlace::Variable(variable(a)),
value: PatternValuePlace::Variable(variable(v)),
tx: PatternNonValuePlace::Placeholder,
})],
}));
}
#[test]
fn test_not_join() {
let e = edn::PlainSymbol::plain("?e");
let a = edn::PlainSymbol::plain("?a");
let v = edn::PlainSymbol::plain("?v");
Parse and Algebrize `not` & `not-join`. (#302) (Closes #303, #389, #422 ) r=rnewman * Part 1 - Parse `not` and `not-join` * Part 2 - Validate `not` and `not-join` pre-algebrization * Address review comments rnewman. * Remove `WhereNotClause` and populate `NotJoin` with `WhereClause`. * Fix validation for `not` and `not-join`, removing tests that were invalid. * Address rustification comments. * Rebase against `rust` branch. * Part 3 - Add required types for NotJoin. * Implement `PartialEq` for `ConjoiningClauses` so `ComputedTable` can be included inside `ColumnConstraint::NotExists` * Part 4 - Implement `apply_not_join` * Part 5 - Call `apply_not_join` from inside `apply_clause` * Part 6 - Translate `not-join` into `NOT EXISTS` SQL * Address review comments. * Rename `projected` to `unified` to better describe the fact that we are not projecting any variables. * Check for presence of each unified var in either `column_bindings` or `input_bindings` and bail if not there. * Copy over `input_bindings` for each var in `unified`. * Only copy over the first `column_binding` for each variable in `unified` rather than the whole list. * Update tests. * Address review comments. * Make output from Debug for NotExists more useful * Clear up misunderstanding. Any single failing clause in the not will cause the entire not to be considered empty * Address review comments. * Remove Limit requirement from cc_to_exists. * Use Entry.or_insert instead of matching on the entry to add to column_bindings. * Move addition of value_bindings to before apply_clauses on template. * Tidy up tests with some variable reuse. * Addressed nits, * Address review comments. * Move addition of column_bindings to above apply_clause. * Update tests. * Add test to ensure that unbound vars fail * Improve test for unbound variable to check for correct variable and error * address nits
2017-04-28 09:44:11 +00:00
assert_edn_parses_to!(Where::not_join_clause,
"(not-join [?e] [?e ?a ?v])",
WhereClause::NotJoin(
NotJoin {
unify_vars: UnifyVars::Explicit(btreeset!{variable(e.clone())}),
Parse and Algebrize `not` & `not-join`. (#302) (Closes #303, #389, #422 ) r=rnewman * Part 1 - Parse `not` and `not-join` * Part 2 - Validate `not` and `not-join` pre-algebrization * Address review comments rnewman. * Remove `WhereNotClause` and populate `NotJoin` with `WhereClause`. * Fix validation for `not` and `not-join`, removing tests that were invalid. * Address rustification comments. * Rebase against `rust` branch. * Part 3 - Add required types for NotJoin. * Implement `PartialEq` for `ConjoiningClauses` so `ComputedTable` can be included inside `ColumnConstraint::NotExists` * Part 4 - Implement `apply_not_join` * Part 5 - Call `apply_not_join` from inside `apply_clause` * Part 6 - Translate `not-join` into `NOT EXISTS` SQL * Address review comments. * Rename `projected` to `unified` to better describe the fact that we are not projecting any variables. * Check for presence of each unified var in either `column_bindings` or `input_bindings` and bail if not there. * Copy over `input_bindings` for each var in `unified`. * Only copy over the first `column_binding` for each variable in `unified` rather than the whole list. * Update tests. * Address review comments. * Make output from Debug for NotExists more useful * Clear up misunderstanding. Any single failing clause in the not will cause the entire not to be considered empty * Address review comments. * Remove Limit requirement from cc_to_exists. * Use Entry.or_insert instead of matching on the entry to add to column_bindings. * Move addition of value_bindings to before apply_clauses on template. * Tidy up tests with some variable reuse. * Addressed nits, * Address review comments. * Move addition of column_bindings to above apply_clause. * Update tests. * Add test to ensure that unbound vars fail * Improve test for unbound variable to check for correct variable and error * address nits
2017-04-28 09:44:11 +00:00
clauses: vec![WhereClause::Pattern(Pattern {
source: None,
entity: PatternNonValuePlace::Variable(variable(e)),
attribute: PatternNonValuePlace::Variable(variable(a)),
value: PatternValuePlace::Variable(variable(v)),
tx: PatternNonValuePlace::Placeholder,
})],
}));
}
#[test]
fn test_find_sp_variable() {
let sym = edn::PlainSymbol::plain("?x");
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
let input = edn::Value::Vector(vec![edn::Value::PlainSymbol(sym.clone())]);
assert_parses_to!(|| vector().of_exactly(Query::variable()), input, variable(sym));
}
#[test]
fn test_find_scalar() {
let sym = edn::PlainSymbol::plain("?x");
let period = edn::PlainSymbol::plain(".");
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
let input = edn::Value::Vector(vec![edn::Value::PlainSymbol(sym.clone()), edn::Value::PlainSymbol(period.clone())]);
assert_parses_to!(|| vector().of_exactly(Find::find_scalar()),
input,
FindSpec::FindScalar(Element::Variable(variable(sym))));
}
#[test]
fn test_find_coll() {
let sym = edn::PlainSymbol::plain("?x");
let period = edn::PlainSymbol::plain("...");
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
let input = edn::Value::Vector(vec![edn::Value::PlainSymbol(sym.clone()),
edn::Value::PlainSymbol(period.clone())]);
assert_parses_to!(Find::find_coll,
input,
FindSpec::FindColl(Element::Variable(variable(sym))));
}
#[test]
fn test_find_rel() {
let vx = edn::PlainSymbol::plain("?x");
let vy = edn::PlainSymbol::plain("?y");
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
let input = edn::Value::Vector(vec![edn::Value::PlainSymbol(vx.clone()), edn::Value::PlainSymbol(vy.clone())]);
assert_parses_to!(|| vector().of_exactly(Find::find_rel()),
input,
FindSpec::FindRel(vec![Element::Variable(variable(vx)),
Element::Variable(variable(vy))]));
}
#[test]
fn test_find_tuple() {
let vx = edn::PlainSymbol::plain("?x");
let vy = edn::PlainSymbol::plain("?y");
Improve parsing of nested `edn::ValueAndSpan` streams. r=rnewman (#393) * Pre: Expose more in edn. * Pre: Make it easier to work with ValueAndSpan. with_spans() is a temporary hack, needed only because I don't care to parse the bootstrap assertions from text right now. * Part 1a: Add `value_and_span` for parsing nested `edn::ValueAndSpan` instances. I wasn't able to abstract over `edn::Value` and `edn::ValueAndSpan`; there are multiple obstacles. I chose to roll with `edn::ValueAndSpan` since it exposes the additional span information that we will want to form good error messages in the future. * Part 1b: Add keyword_map() parsing an `edn::Value::Vector` into an `edn::Value::map`. * Part 1c: Add `Log`/`.log(...)` for logging parser progress. This is a terrible hack, but it sure helps to debug complicated nested parsers. I don't even know what a principled approach would look like; since our parser combinators are so frequently expressed in code, it's hard to imagine a data-driven interpreter that can help debug things. * Part 2: Use `value_and_span` apparatus in tx-parser/. I break an abstraction boundary by returning a value column `edn::ValueAndSpan` rather than just an `edn::Value`. That is, the transaction processor shouldn't care where the `edn::Value` it is processing arose -- even we care to track that information we should bake it into the `Entity` type. We do this because we need to dynamically parse the value column to support nested maps, and parsing requires a full `edn::ValueAndSpan`. Alternately, we could cheat and fake the spans when parsing nested maps, but that's potentially expensive. * Part 3: Use `value_and_span` apparatus in query-parser/. * Part 4: Use `value_and_span` apparatus in root crate. * Review comment: Make Span and SpanPosition Copy. * Review comment: nits. * Review comment: Make `or` be `or_exactly`. I baked the eof checking directly into the parser, rather than using the skip and eof parsers. I also took the time to restore some tests that were mistakenly commented out. * Review comment: Extract and use def_matches_* macros. * Review comment: .map() as late as possible.
2017-04-06 17:06:28 +00:00
let input = edn::Value::Vector(vec![edn::Value::PlainSymbol(vx.clone()),
edn::Value::PlainSymbol(vy.clone())]);
assert_parses_to!(Find::find_tuple,
input,
FindSpec::FindTuple(vec![Element::Variable(variable(vx)),
Element::Variable(variable(vy))]));
}
#[test]
fn test_natural_numbers() {
let text = edn::Value::Text("foo".to_string());
let neg = edn::Value::Integer(-10);
let zero = edn::Value::Integer(0);
let pos = edn::Value::Integer(5);
// This is terrible, but destructuring errors is frustrating.
let input = text.with_spans();
let mut par = Query::natural_number();
let x = par.parse(input.atom_stream()).err().expect("an error").errors;
let result = format!("{:?}", x);
assert_eq!(result, "[Other(Error(InvalidLimit(Text(\"foo\")), State { next_error: None, backtrace: None }))]");
let input = neg.with_spans();
let mut par = Query::natural_number();
let x = par.parse(input.atom_stream()).err().expect("an error").errors;
let result = format!("{:?}", x);
assert_eq!(result, "[Other(Error(InvalidLimit(Integer(-10)), State { next_error: None, backtrace: None }))]");
let input = zero.with_spans();
let mut par = Query::natural_number();
let x = par.parse(input.atom_stream()).err().expect("an error").errors;
let result = format!("{:?}", x);
assert_eq!(result, "[Other(Error(InvalidLimit(Integer(0)), State { next_error: None, backtrace: None }))]");
let input = pos.with_spans();
let mut par = Query::natural_number();
assert_eq!(None, par.parse(input.atom_stream()).err());
}
#[test]
fn test_fn_arg_collections() {
let vx = edn::PlainSymbol::plain("?x");
let vy = edn::PlainSymbol::plain("?y");
let input = edn::Value::Vector(vec![edn::Value::Vector(vec![edn::Value::PlainSymbol(vx.clone()),
edn::Value::PlainSymbol(vy.clone())])]);
assert_parses_to!(|| vector().of_exactly(Query::fn_arg()),
input,
FnArg::Vector(vec![FnArg::Variable(variable(vx)),
FnArg::Variable(variable(vy)),
]));
}
#[test]
fn test_bind_scalar() {
let vx = edn::PlainSymbol::plain("?x");
assert_edn_parses_to!(|| list().of_exactly(Bind::binding()),
"(?x)",
Binding::BindScalar(variable(vx)));
}
#[test]
fn test_bind_coll() {
let vx = edn::PlainSymbol::plain("?x");
assert_edn_parses_to!(|| list().of_exactly(Bind::binding()),
"([?x ...])",
Binding::BindColl(variable(vx)));
}
#[test]
fn test_bind_rel() {
let vx = edn::PlainSymbol::plain("?x");
let vy = edn::PlainSymbol::plain("?y");
let vw = edn::PlainSymbol::plain("?w");
assert_edn_parses_to!(|| list().of_exactly(Bind::binding()),
"([[?x ?y _ ?w]])",
Binding::BindRel(vec![VariableOrPlaceholder::Variable(variable(vx)),
VariableOrPlaceholder::Variable(variable(vy)),
VariableOrPlaceholder::Placeholder,
VariableOrPlaceholder::Variable(variable(vw)),
]));
}
#[test]
fn test_bind_tuple() {
let vx = edn::PlainSymbol::plain("?x");
let vy = edn::PlainSymbol::plain("?y");
let vw = edn::PlainSymbol::plain("?w");
assert_edn_parses_to!(|| list().of_exactly(Bind::binding()),
"([?x ?y _ ?w])",
Binding::BindTuple(vec![VariableOrPlaceholder::Variable(variable(vx)),
VariableOrPlaceholder::Variable(variable(vy)),
VariableOrPlaceholder::Placeholder,
VariableOrPlaceholder::Variable(variable(vw)),
]));
}
#[test]
fn test_the() {
assert_edn_parses_to!(Find::seq_elem,
"(the ?y)",
Element::Corresponding(Variable::from_valid_name("?y")));
assert_edn_parses_to!(Find::find_tuple,
"[(the ?x) ?y]",
FindSpec::FindTuple(vec![Element::Corresponding(Variable::from_valid_name("?x")),
Element::Variable(Variable::from_valid_name("?y"))]));
assert_edn_parses_to!(Find::spec,
"[(the ?x) ?y]",
FindSpec::FindTuple(vec![Element::Corresponding(Variable::from_valid_name("?x")),
Element::Variable(Variable::from_valid_name("?y"))]));
let expected_query =
FindQuery {
find_spec: FindSpec::FindTuple(vec![Element::Corresponding(Variable::from_valid_name("?x")),
Element::Variable(Variable::from_valid_name("?y"))]),
where_clauses: vec![
WhereClause::Pattern(Pattern {
source: None,
entity: PatternNonValuePlace::Variable(Variable::from_valid_name("?x")),
attribute: PatternNonValuePlace::Placeholder,
value: PatternValuePlace::Variable(Variable::from_valid_name("?y")),
tx: PatternNonValuePlace::Placeholder,
})],
default_source: SrcVar::DefaultSrc,
with: Default::default(),
in_vars: Default::default(),
in_sources: Default::default(),
limit: Limit::None,
order: None,
};
assert_edn_parses_to!(Find::query,
"[:find [(the ?x) ?y]
:where [?x _ ?y]]",
expected_query);
// If we give a malformed pull expression, we don't fall through to aggregates.
assert_parse_failure_contains!(Find::elem,
"(pull x [])",
r#"errors: [Unexpected(Token(ValueAndSpan { inner: PlainSymbol(PlainSymbol("x")), span: Span(6, 7) })), Expected(Borrowed("variable"))]"#);
}
#[test]
fn test_where_fn() {
assert_edn_parses_to!(Where::where_fn,
"[(f ?x 1) ?y]",
WhereClause::WhereFn(WhereFn {
operator: edn::PlainSymbol::plain("f"),
args: vec![FnArg::Variable(Variable::from_valid_name("?x")),
FnArg::EntidOrInteger(1)],
binding: Binding::BindScalar(Variable::from_valid_name("?y")),
}));
assert_edn_parses_to!(Where::where_fn,
"[(f ?x) [?y ...]]",
WhereClause::WhereFn(WhereFn {
operator: edn::PlainSymbol::plain("f"),
args: vec![FnArg::Variable(Variable::from_valid_name("?x"))],
binding: Binding::BindColl(Variable::from_valid_name("?y")),
}));
assert_edn_parses_to!(Where::where_fn,
"[(f) [?y _]]",
WhereClause::WhereFn(WhereFn {
operator: edn::PlainSymbol::plain("f"),
args: vec![],
binding: Binding::BindTuple(vec![VariableOrPlaceholder::Variable(Variable::from_valid_name("?y")),
VariableOrPlaceholder::Placeholder]),
}));
assert_edn_parses_to!(Where::where_fn,
"[(f) [[_ ?y]]]",
WhereClause::WhereFn(WhereFn {
operator: edn::PlainSymbol::plain("f"),
args: vec![],
binding: Binding::BindRel(vec![VariableOrPlaceholder::Placeholder,
VariableOrPlaceholder::Variable(Variable::from_valid_name("?y"))]),
}));
}
#[test]
fn test_type_anno() {
assert_edn_parses_to!(Where::type_annotation,
"[(type ?x :db.type/string)]",
WhereClause::TypeAnnotation(TypeAnnotation {
value_type: edn::Keyword::namespaced("db.type", "string"),
variable: Variable::from_valid_name("?x"),
}));
// We don't check for valid types, or even that the type is namespaced.
assert_edn_parses_to!(Where::clause,
"[[type ?foo :db_type_long]]",
WhereClause::TypeAnnotation(TypeAnnotation {
value_type: edn::Keyword::plain("db_type_long"),
variable: Variable::from_valid_name("?foo"),
}));
}
#[test]
fn test_pull() {
assert_edn_parses_to!(Query::pull_attribute,
"*",
PullAttributeSpec::Wildcard);
assert_edn_parses_to!(Query::pull_attributes,
"[*]",
vec![PullAttributeSpec::Wildcard]);
assert_edn_parses_to!(Find::elem,
"(pull ?v [*])",
Element::Pull(Pull {
var: Variable::from_valid_name("?v"),
patterns: vec![PullAttributeSpec::Wildcard],
}));
let foo_bar = ::std::rc::Rc::new(edn::Keyword::namespaced("foo", "bar"));
let foo_baz = ::std::rc::Rc::new(edn::Keyword::namespaced("foo", "baz"));
let foo_horse = ::std::rc::Rc::new(edn::Keyword::namespaced("foo", "horse"));
let horse = ::std::rc::Rc::new(edn::Keyword::plain("horse"));
assert_edn_parses_to!(Query::pull_concrete_attribute,
":foo/bar",
PullAttributeSpec::Attribute(
PullConcreteAttribute::Ident(foo_bar.clone()).into()));
assert_edn_parses_to!(Query::pull_attribute,
":foo/bar",
PullAttributeSpec::Attribute(
PullConcreteAttribute::Ident(foo_bar.clone()).into()));
assert_edn_parses_to!(Find::elem,
"(pull ?v [:foo/bar :as :foo/horse, :foo/baz])",
Element::Pull(Pull {
var: Variable::from_valid_name("?v"),
patterns: vec![
PullAttributeSpec::Attribute(
NamedPullAttribute {
attribute: PullConcreteAttribute::Ident(foo_bar.clone()),
alias: Some(foo_horse),
}),
PullAttributeSpec::Attribute(
PullConcreteAttribute::Ident(foo_baz.clone()).into()),
],
}));
assert_edn_parses_to!(Find::elem,
"(pull ?v [:foo/bar :as :horse, :foo/baz])",
Element::Pull(Pull {
var: Variable::from_valid_name("?v"),
patterns: vec![
PullAttributeSpec::Attribute(
NamedPullAttribute {
attribute: PullConcreteAttribute::Ident(foo_bar.clone()),
alias: Some(horse),
}),
PullAttributeSpec::Attribute(
PullConcreteAttribute::Ident(foo_baz.clone()).into()),
],
}));
assert_parse_failure_contains!(Find::elem,
"(pull ?x [* :foo/bar])",
r#"errors: [Unexpected(Borrowed("wildcard with specified attributes"))]"#);
}
#[test]
fn test_query_with_pull() {
let q = "[:find ?x (pull ?x [:foo/bar]) :where [?x _ _]]";
let expected_query =
FindQuery {
find_spec: FindSpec::FindRel(vec![
Element::Variable(Variable::from_valid_name("?x")),
Element::Pull(Pull {
var: Variable::from_valid_name("?x"),
patterns: vec![
PullAttributeSpec::Attribute(
PullConcreteAttribute::Ident(
::std::rc::Rc::new(edn::Keyword::namespaced("foo", "bar"))
).into()
),
] })]),
where_clauses: vec![
WhereClause::Pattern(Pattern {
source: None,
entity: PatternNonValuePlace::Variable(Variable::from_valid_name("?x")),
attribute: PatternNonValuePlace::Placeholder,
value: PatternValuePlace::Placeholder,
tx: PatternNonValuePlace::Placeholder,
})],
default_source: SrcVar::DefaultSrc,
with: Default::default(),
in_vars: Default::default(),
in_sources: Default::default(),
limit: Limit::None,
order: None,
};
assert_edn_parses_to!(Find::query,
q,
expected_query);
}
}