mentat/parser-utils/src/value_and_span.rs

773 lines
30 KiB
Rust
Raw Normal View History

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
// 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.
#![allow(dead_code)]
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;
use std::cmp::Ordering;
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::fmt::{
Debug,
Display,
Formatter,
};
use combine::{
ConsumedResult,
ParseError,
Parser,
ParseResult,
StreamOnce,
parser,
satisfy_map,
};
use combine::primitives; // To not shadow Error.
use combine::primitives::{
FastResult,
};
use combine::combinator::{
Expected,
FnParser,
};
use edn;
use macros::{
KeywordMapParser,
};
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
/// A wrapper to let us order `edn::Span` in whatever way is appropriate for parsing with `combine`.
#[derive(Clone, Copy, Debug)]
pub struct SpanPosition(pub edn::Span);
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
impl Display for SpanPosition {
fn fmt(&self, f: &mut Formatter) -> ::std::fmt::Result {
self.0.fmt(f)
}
}
impl PartialEq for SpanPosition {
fn eq(&self, other: &Self) -> bool {
self.cmp(other) == Ordering::Equal
}
}
impl Eq for SpanPosition { }
impl PartialOrd for SpanPosition {
fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
Some(self.cmp(other))
}
}
impl Ord for SpanPosition {
fn cmp(&self, other: &Self) -> Ordering {
(self.0).0.cmp(&(other.0).0)
}
}
/// An iterator specifically for iterating `edn::ValueAndSpan` instances in various ways.
///
/// Enumerating each iteration type allows us to have a single `combine::Stream` implementation
/// yielding `ValueAndSpan` items, which allows us to yield uniform `combine::ParseError` types from
/// disparate parsers.
#[derive(Clone)]
pub enum Iter<'a> {
Empty,
Atom(std::iter::Once<&'a edn::ValueAndSpan>),
Vector(std::slice::Iter<'a, edn::ValueAndSpan>),
List(std::collections::linked_list::Iter<'a, edn::ValueAndSpan>),
/// Iterates a map {:k1 v1, :k2 v2, ...} as a single `flat_map` slice [k1, v1, k2, v2, ...].
Map(std::iter::FlatMap<std::collections::btree_map::Iter<'a, edn::ValueAndSpan, edn::ValueAndSpan>,
std::iter::Chain<std::iter::Once<&'a edn::ValueAndSpan>, std::iter::Once<&'a edn::ValueAndSpan>>,
fn((&'a edn::ValueAndSpan, &'a edn::ValueAndSpan)) -> std::iter::Chain<std::iter::Once<&'a edn::ValueAndSpan>, std::iter::Once<&'a edn::ValueAndSpan>>>),
/// Iterates a map with vector values {:k1 [v11 v12 ...], :k2 [v21 v22 ...], ...} as a single
/// flattened map [k1, v11, v12, ..., k2, v21, v22, ...].
KeywordMap(std::iter::FlatMap<std::collections::btree_map::Iter<'a, edn::ValueAndSpan, edn::ValueAndSpan>,
std::iter::Chain<std::iter::Once<&'a edn::ValueAndSpan>, Box<Iter<'a>>>,
fn((&'a edn::ValueAndSpan, &'a edn::ValueAndSpan)) -> std::iter::Chain<std::iter::Once<&'a edn::ValueAndSpan>, Box<Iter<'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
// TODO: Support Set and Map more naturally. This is significantly more work because the
// existing BTreeSet and BTreeMap iterators do not implement Clone, and implementing Clone for
// them is involved. Since we don't really need to parse sets and maps at this time, this will
// do for now.
}
impl<'a> Iterator for Iter<'a> {
type Item = &'a edn::ValueAndSpan;
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
fn next(&mut self) -> Option<Self::Item> {
match *self {
Iter::Empty => None,
Iter::Atom(ref mut i) => i.next(),
Iter::Vector(ref mut i) => i.next(),
Iter::List(ref mut i) => i.next(),
Iter::Map(ref mut i) => i.next(),
Iter::KeywordMap(ref mut i) => i.next(),
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
}
}
}
/// A single `combine::Stream` implementation iterating `edn::ValueAndSpan` instances. Equivalent
/// to `combine::IteratorStream` as produced by `combine::from_iter`, but specialized to
/// `edn::ValueAndSpan`.
#[derive(Clone)]
pub struct Stream<'a>(Iter<'a>, SpanPosition);
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
/// Things specific to parsing with `combine` and our `Stream` that need a trait to live outside of
/// the `edn` crate.
pub trait Item<'a>: Clone + PartialEq + Sized {
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
/// Position could be specialized to `SpanPosition`.
type Position: Clone + Ord + std::fmt::Display;
/// A slight generalization of `combine::Positioner` that allows to set the position based on
/// the `edn::ValueAndSpan` being iterated.
fn start(&self) -> Self::Position;
fn update_position(&self, &mut Self::Position);
fn child_iter(&'a self) -> Iter<'a>;
fn child_stream(&'a self) -> Stream<'a>;
fn atom_iter(&'a self) -> Iter<'a>;
fn atom_stream(&'a self) -> Stream<'a>;
fn keyword_map_iter(&'a self) -> Iter<'a>;
fn keyword_map_stream(&'a self) -> Stream<'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
}
impl<'a> Item<'a> for edn::ValueAndSpan {
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
type Position = SpanPosition;
fn start(&self) -> Self::Position {
SpanPosition(self.span.clone())
}
fn update_position(&self, position: &mut Self::Position) {
*position = SpanPosition(self.span.clone())
}
fn keyword_map_iter(&'a self) -> Iter<'a> {
fn flatten_k_vector<'a>((k, v): (&'a edn::ValueAndSpan, &'a edn::ValueAndSpan)) -> std::iter::Chain<std::iter::Once<&'a edn::ValueAndSpan>, Box<Iter<'a>>> {
std::iter::once(k).chain(Box::new(v.child_iter()))
}
match self.inner.as_map() {
Some(ref map) => Iter::KeywordMap(map.iter().flat_map(flatten_k_vector)),
None => Iter::Empty
}
}
fn keyword_map_stream(&'a self) -> Stream<'a> {
let span = self.span.clone();
Stream(self.keyword_map_iter(), SpanPosition(span))
}
fn child_iter(&'a self) -> Iter<'a> {
fn flatten_k_v<'a>((k, v): (&'a edn::ValueAndSpan, &'a edn::ValueAndSpan)) -> std::iter::Chain<std::iter::Once<&'a edn::ValueAndSpan>, std::iter::Once<&'a edn::ValueAndSpan>> {
std::iter::once(k).chain(std::iter::once(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
match self.inner {
edn::SpannedValue::Vector(ref values) => Iter::Vector(values.iter()),
edn::SpannedValue::List(ref values) => Iter::List(values.iter()),
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
// Parsing pairs with `combine` is tricky; parsing sequences is easy.
edn::SpannedValue::Map(ref map) => Iter::Map(map.iter().flat_map(flatten_k_v)),
_ => Iter::Empty,
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
}
}
fn child_stream(&'a self) -> Stream<'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
let span = self.span.clone();
Stream(self.child_iter(), SpanPosition(span))
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
}
fn atom_iter(&'a self) -> Iter<'a> {
Iter::Atom(std::iter::once(self))
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
}
fn atom_stream(&'a self) -> Stream<'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
let span = self.span.clone();
Stream(self.atom_iter(), SpanPosition(span))
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
}
}
/// `OfExactly` and `of_exactly` allow us to express nested parsers naturally.
///
/// For example, `vector().of_exactly(many(list()))` parses a vector-of-lists, like [(1 2) (:a :b) ("test") ()].
///
/// The "outer" parser `P` and the "nested" parser `N` must be compatible: `P` must produce an
/// output `edn::ValueAndSpan` which can itself be turned into a stream of child elements; and `N`
/// must accept the resulting input `Stream`. This compatibility allows us to lift errors from the
/// nested parser to the outer parser, which is part of what has made parsing `&'a [edn::Value]`
/// difficult.
#[derive(Clone)]
pub struct OfExactly<P, N>(P, N);
pub trait Streaming<'a> {
fn as_stream(self) -> Stream<'a>;
}
impl<'a> Streaming<'a> for &'a edn::ValueAndSpan {
fn as_stream(self) -> Stream<'a> {
self.child_stream()
}
}
impl<'a> Streaming<'a> for Stream<'a> {
fn as_stream(self) -> Stream<'a> {
self
}
}
impl<'a, P, N, M, O> Parser for OfExactly<P, N>
where P: Parser<Input=Stream<'a>, Output=M>,
N: Parser<Input=Stream<'a>, Output=O>,
M: 'a + Streaming<'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
{
type Input = P::Input;
type Output = O;
#[inline]
fn parse_lazy(&mut self, input: Self::Input) -> ConsumedResult<Self::Output, Self::Input> {
use self::FastResult::*;
match self.0.parse_lazy(input) {
ConsumedOk((outer_value, outer_input)) => {
match self.1.parse_lazy(outer_value.as_stream()) {
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
ConsumedOk((inner_value, mut inner_input)) | EmptyOk((inner_value, mut inner_input)) => {
match inner_input.uncons() {
Err(ref err) if *err == primitives::Error::end_of_input() => ConsumedOk((inner_value, outer_input)),
_ => EmptyErr(ParseError::empty(inner_input.position())),
}
},
// TODO: Improve the error output to reference the nested value (or span) in
// some way. This seems surprisingly difficult to do, so we just surface the
// inner error message right now. See also the comment below.
EmptyErr(e) | ConsumedErr(e) => ConsumedErr(e),
}
},
EmptyOk((outer_value, outer_input)) => {
match self.1.parse_lazy(outer_value.as_stream()) {
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
ConsumedOk((inner_value, mut inner_input)) | EmptyOk((inner_value, mut inner_input)) => {
match inner_input.uncons() {
Err(ref err) if *err == primitives::Error::end_of_input() => EmptyOk((inner_value, outer_input)),
_ => EmptyErr(ParseError::empty(inner_input.position())),
}
},
// TODO: Improve the error output. See the comment above.
EmptyErr(e) | ConsumedErr(e) => EmptyErr(e),
}
},
ConsumedErr(e) => ConsumedErr(e),
EmptyErr(e) => EmptyErr(e),
}
}
fn add_error(&mut self, errors: &mut ParseError<Self::Input>) {
self.0.add_error(errors);
}
}
#[inline(always)]
pub fn of_exactly<'a, P, N, M, O>(p: P, n: N) -> OfExactly<P, N>
where P: Parser<Input=Stream<'a>, Output=M>,
N: Parser<Input=Stream<'a>, Output=O>,
M: 'a + Streaming<'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
{
OfExactly(p, n)
}
/// We need a trait to define `Parser.of` and have it live outside of the `combine` crate.
pub trait OfExactlyParsing: Parser + Sized {
fn of_exactly<N, O>(self, n: N) -> OfExactly<Self, N>
where Self: Sized,
N: Parser<Input = Self::Input, Output=O>;
}
impl<'a, P, M> OfExactlyParsing for P
where P: Parser<Input=Stream<'a>, Output=M>,
M: 'a + Streaming<'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
{
fn of_exactly<N, O>(self, n: N) -> OfExactly<P, N>
where N: Parser<Input = Self::Input, Output=O>
{
of_exactly(self, n)
}
}
/// Equivalent to `combine::IteratorStream`.
impl<'a> StreamOnce for Stream<'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
{
type Item = &'a edn::ValueAndSpan;
type Range = &'a edn::ValueAndSpan;
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
type Position = SpanPosition;
#[inline]
fn uncons(&mut self) -> std::result::Result<Self::Item, primitives::Error<Self::Item, Self::Item>> {
match self.0.next() {
Some(x) => {
x.update_position(&mut self.1);
Ok(x)
},
None => Err(primitives::Error::end_of_input()),
}
}
#[inline(always)]
fn position(&self) -> Self::Position {
self.1.clone()
}
}
/// Shorthands, just enough to convert the `mentat_db` crate for now. Written using `Box` for now:
/// it's simple and we can address allocation issues if and when they surface.
pub fn vector_<'a>(input: Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>> {
satisfy_map(|v: &'a edn::ValueAndSpan| {
if v.inner.is_vector() {
Some(v.child_stream())
} else {
None
}
})
.parse_lazy(input)
.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
}
pub fn vector<'a>() -> Expected<FnParser<Stream<'a>, fn(Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>>>> {
parser(vector_ as fn(Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>>).expected("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
}
pub fn list_<'a>(input: Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>> {
satisfy_map(|v: &'a edn::ValueAndSpan| {
if v.inner.is_list() {
Some(v.child_stream())
} else {
None
}
})
.parse_lazy(input)
.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
}
pub fn list<'a>() -> Expected<FnParser<Stream<'a>, fn(Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>>>> {
parser(list_ as fn(Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>>).expected("list")
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 seq_<'a>(input: Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>> {
satisfy_map(|v: &'a edn::ValueAndSpan| {
if v.inner.is_list() || v.inner.is_vector() {
Some(v.child_stream())
} else {
None
}
})
.parse_lazy(input)
.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
}
pub fn seq<'a>() -> Expected<FnParser<Stream<'a>, fn(Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>>>> {
parser(seq_ as fn(Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>>).expected("vector|list")
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 map_<'a>(input: Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>> {
satisfy_map(|v: &'a edn::ValueAndSpan| {
if v.inner.is_map() {
Some(v.child_stream())
} else {
None
}
})
.parse_lazy(input)
.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
}
pub fn map<'a>() -> Expected<FnParser<Stream<'a>, fn(Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>>>> {
parser(map_ as fn(Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>>).expected("map")
}
/// A `[k v]` pair in the map form of a keyword map must have the shape `[:k, [v1, v2, ...]]`, with
/// none of `v1`, `v2`, ... a keyword: without loss of generality, we cannot represent the case
/// where `vn` is a keyword `:l`, since `[:k v1 v2 ... :l]`, isn't a valid keyword map in vector
/// form. This function tests that a `[k v]` pair obeys these constraints.
///
/// If we didn't test this, then we might flatten a map `[:k [:l]] to `[:k :l]`, which isn't a valid
/// keyword map in vector form.
pub fn is_valid_keyword_map_k_v<'a>((k, v): (&'a edn::ValueAndSpan, &'a edn::ValueAndSpan)) -> bool {
if !k.inner.is_keyword() {
return false;
}
match v.inner.as_vector() {
None => {
return false;
},
Some(ref vs) => {
if !vs.iter().all(|vv| !vv.inner.is_keyword()) {
return false;
}
},
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
}
return true;
}
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 keyword_map_<'a>(input: Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>> {
satisfy_map(|v: &'a edn::ValueAndSpan| {
v.inner.as_map().and_then(|map| {
if map.iter().all(is_valid_keyword_map_k_v) {
println!("yes {:?}", map);
Some(v.keyword_map_stream())
} else {
println!("no {:?}", map);
None
}
})
})
.parse_lazy(input)
.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
pub fn keyword_map<'a>() -> Expected<FnParser<Stream<'a>, fn(Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>>>> {
parser(keyword_map_ as fn(Stream<'a>) -> ParseResult<Stream<'a>, Stream<'a>>).expected("keyword map")
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 integer_<'a>(input: Stream<'a>) -> ParseResult<i64, Stream<'a>> {
satisfy_map(|v: &'a edn::ValueAndSpan| v.inner.as_integer())
.parse_lazy(input)
.into()
}
pub fn integer<'a>() -> Expected<FnParser<Stream<'a>, fn(Stream<'a>) -> ParseResult<i64, Stream<'a>>>> {
parser(integer_ as fn(Stream<'a>) -> ParseResult<i64, Stream<'a>>).expected("integer")
}
pub fn namespaced_keyword_<'a>(input: Stream<'a>) -> ParseResult<&'a edn::NamespacedKeyword, Stream<'a>> {
satisfy_map(|v: &'a edn::ValueAndSpan| v.inner.as_namespaced_keyword())
.parse_lazy(input)
.into()
}
pub fn namespaced_keyword<'a>() -> Expected<FnParser<Stream<'a>, fn(Stream<'a>) -> ParseResult<&'a edn::NamespacedKeyword, Stream<'a>>>> {
parser(namespaced_keyword_ as fn(Stream<'a>) -> ParseResult<&'a edn::NamespacedKeyword, Stream<'a>>).expected("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
}
pub fn forward_keyword_<'a>(input: Stream<'a>) -> ParseResult<&'a edn::NamespacedKeyword, Stream<'a>> {
satisfy_map(|v: &'a edn::ValueAndSpan| v.inner.as_namespaced_keyword().and_then(|k| if k.is_forward() { Some(k) } else { None }))
.parse_lazy(input)
.into()
}
pub fn forward_keyword<'a>() -> Expected<FnParser<Stream<'a>, fn(Stream<'a>) -> ParseResult<&'a edn::NamespacedKeyword, Stream<'a>>>> {
parser(forward_keyword_ as fn(Stream<'a>) -> ParseResult<&'a edn::NamespacedKeyword, Stream<'a>>).expected("forward_keyword")
}
pub fn backward_keyword_<'a>(input: Stream<'a>) -> ParseResult<&'a edn::NamespacedKeyword, Stream<'a>> {
satisfy_map(|v: &'a edn::ValueAndSpan| v.inner.as_namespaced_keyword().and_then(|k| if k.is_backward() { Some(k) } else { None }))
.parse_lazy(input)
.into()
}
pub fn backward_keyword<'a>() -> Expected<FnParser<Stream<'a>, fn(Stream<'a>) -> ParseResult<&'a edn::NamespacedKeyword, Stream<'a>>>> {
parser(backward_keyword_ as fn(Stream<'a>) -> ParseResult<&'a edn::NamespacedKeyword, Stream<'a>>).expected("backward_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
/// Generate a `satisfy` expression that matches a `PlainSymbol` value with the given name.
///
/// We do this rather than using `combine::token` so that we don't need to allocate a new `String`
/// inside a `PlainSymbol` inside a `SpannedValue` inside a `ValueAndSpan` just to match input.
#[macro_export]
macro_rules! def_matches_plain_symbol {
( $parser: ident, $name: ident, $input: expr ) => {
def_parser!($parser, $name, &'a edn::ValueAndSpan, {
satisfy(|v: &'a edn::ValueAndSpan| {
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
match v.inner {
edn::SpannedValue::PlainSymbol(ref s) => s.0.as_str() == $input,
_ => false,
}
})
});
}
}
/// Generate a `satisfy` expression that matches a `Keyword` value with the given name.
///
/// We do this rather than using `combine::token` to save allocations.
#[macro_export]
macro_rules! def_matches_keyword {
( $parser: ident, $name: ident, $input: expr ) => {
def_parser!($parser, $name, &'a edn::ValueAndSpan, {
satisfy(|v: &'a edn::ValueAndSpan| {
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
match v.inner {
edn::SpannedValue::Keyword(ref s) => s.0.as_str() == $input,
_ => false,
}
})
});
}
}
/// Generate a `satisfy` expression that matches a `NamespacedKeyword` value with the given
/// namespace and name.
///
/// We do this rather than using `combine::token` to save allocations.
#[macro_export]
macro_rules! def_matches_namespaced_keyword {
( $parser: ident, $name: ident, $input_namespace: expr, $input_name: expr ) => {
def_parser!($parser, $name, &'a edn::ValueAndSpan, {
satisfy(|v: &'a edn::ValueAndSpan| {
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
match v.inner {
edn::SpannedValue::NamespacedKeyword(ref s) => s.namespace.as_str() == $input_namespace && s.name.as_str() == $input_name,
_ => false,
}
})
});
}
}
use combine::primitives::{
Error,
Info,
};
use combine::primitives::FastResult::*;
/// Compare to `tuple_parser!` in `combine`.
///
/// This uses edge cases in Rust's hygienic macro system to represent arbitrary values. That is,
/// `$value: ident` represents both a type in the tuple parameterizing `KeywordMapParser` (since
/// `(A, B, C)` is a valid type declaration) and also a variable value extracted from the underlying
/// instance value. `$tmp: ident` represents an optional value to return.
///
/// This unrolls the cases. Each loop iteration reads a token. It then unrolls the known cases,
/// checking if any case matches the keyword string. If yes, we parse further. If no, we move on
/// to the next case. If no case matches, we fail.
macro_rules! keyword_map_parser {
($(($keyword:ident, $value:ident, $tmp:ident)),+) => {
impl <'a, $($value:),+> Parser for KeywordMapParser<($((&'static str, $value)),+)>
where $($value: Parser<Input=Stream<'a>>),+
{
type Input = Stream<'a>;
type Output = ($(Option<$value::Output>),+);
#[allow(non_snake_case)]
fn parse_lazy(&mut self,
mut input: Stream<'a>)
-> ConsumedResult<($(Option<$value::Output>),+), Stream<'a>> {
let ($((ref $keyword, ref mut $value)),+) = (*self).0;
let mut consumed = false;
$(
let mut $tmp = None;
)+
loop {
match input.uncons() {
Ok(value) => {
$(
if let Some(ref keyword) = value.inner.as_keyword() {
if keyword.0.as_str() == *$keyword {
if $tmp.is_some() {
// Repeated match -- bail out! Providing good error
// messages is hard; this will do for now.
return ConsumedErr(ParseError::new(input.position(), Error::Unexpected(Info::Token(value))));
}
consumed = true;
$tmp = match $value.parse_lazy(input.clone()) {
ConsumedOk((x, new_input)) => {
input = new_input;
Some(x)
}
EmptyErr(mut err) => {
if let Ok(t) = input.uncons() {
err.add_error(Error::Unexpected(Info::Token(t)));
}
if consumed {
return ConsumedErr(err)
} else {
return EmptyErr(err)
}
}
ConsumedErr(err) => return ConsumedErr(err),
EmptyOk((x, new_input)) => {
input = new_input;
Some(x)
}
};
continue
}
}
)+
// No keyword matched! Bail out.
return ConsumedErr(ParseError::new(input.position(), Error::Unexpected(Info::Token(value))));
},
Err(err) => {
if consumed {
return ConsumedOk((($($tmp),+), input))
} else {
if err == Error::end_of_input() {
return EmptyOk((($($tmp),+), input));
}
return EmptyErr(ParseError::new(input.position(), err))
}
},
}
}
}
}
}
}
keyword_map_parser!((Ak, Av, At), (Bk, Bv, Bt));
keyword_map_parser!((Ak, Av, At), (Bk, Bv, Bt), (Ck, Cv, Ct));
keyword_map_parser!((Ak, Av, At), (Bk, Bv, Bt), (Ck, Cv, Ct), (Dk, Dv, Dt));
keyword_map_parser!((Ak, Av, At), (Bk, Bv, Bt), (Ck, Cv, Ct), (Dk, Dv, Dt), (Ek, Ev, Et));
keyword_map_parser!((Ak, Av, At), (Bk, Bv, Bt), (Ck, Cv, Ct), (Dk, Dv, Dt), (Ek, Ev, Et), (Fk, Fv, Ft));
keyword_map_parser!((Ak, Av, At), (Bk, Bv, Bt), (Ck, Cv, Ct), (Dk, Dv, Dt), (Ek, Ev, Et), (Fk, Fv, Ft), (Gk, Gv, Gt));
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
#[cfg(test)]
mod tests {
use combine::{
eof,
many,
satisfy,
};
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 super::*;
use macros::{
ResultParser,
};
/// A little test parser.
pub struct Test<'a>(std::marker::PhantomData<&'a ()>);
def_matches_namespaced_keyword!(Test, add, "db", "add");
def_parser!(Test, entid, i64, {
integer()
.map(|x| x)
.or(namespaced_keyword().map(|_| -1))
});
#[test]
#[should_panic(expected = r#"keyword map has repeated key: "x""#)]
fn test_keyword_map_of() {
keyword_map_of!(("x", Test::entid()),
("x", Test::entid()));
}
#[test]
fn test_iter() {
// A vector and a map iterated as a keyword map produce the same elements.
let input = edn::parse::value("[:y 3 4 :x 1 2]").expect("to be able to parse input as EDN");
assert_eq!(input.child_iter().cloned().map(|x| x.without_spans()).into_iter().collect::<Vec<_>>(),
edn::parse::value("[:y 3 4 :x 1 2]").expect("to be able to parse input as EDN").without_spans().into_vector().expect("an EDN vector"));
let input = edn::parse::value("{:x [1 2] :y [3 4]}").expect("to be able to parse input as EDN");
assert_eq!(input.keyword_map_iter().cloned().map(|x| x.without_spans()).into_iter().collect::<Vec<_>>(),
edn::parse::value("[:y 3 4 :x 1 2]").expect("to be able to parse input as EDN").without_spans().into_vector().expect("an EDN vector"));
// Parsing a keyword map in map and vector form produces the same elements. The order (:y
// before :x) is a foible of our EDN implementation and could be easily changed.
assert_edn_parses_to!(|| keyword_map().or(vector()).map(|x| x.0.map(|x| x.clone().without_spans()).into_iter().collect::<Vec<_>>()),
"{:x [1] :y [2]}",
edn::parse::value("[:y 2 :x 1]").expect("to be able to parse input as EDN").without_spans().into_vector().expect("an EDN vector"));
assert_edn_parses_to!(|| keyword_map().or(vector()).map(|x| x.0.map(|x| x.clone().without_spans()).into_iter().collect::<Vec<_>>()),
"[:y 2 :x 1]",
edn::parse::value("[:y 2 :x 1]").expect("to be able to parse input as EDN").without_spans().into_vector().expect("an EDN 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
}
#[test]
fn test_keyword_map() {
assert_edn_parses_to!(|| vector().of_exactly(keyword_map_of!(("x", Test::entid()), ("y", Test::entid()))),
"[:y 2 :x 1]",
(Some(1), Some(2)));
assert_edn_parses_to!(|| vector().of_exactly(keyword_map_of!(("x", Test::entid()), ("y", Test::entid()))),
"[:x 1 :y 2]",
(Some(1), Some(2)));
assert_edn_parses_to!(|| vector().of_exactly(keyword_map_of!(("x", Test::entid()), ("y", Test::entid()))),
"[:x 1]",
(Some(1), None));
assert_edn_parses_to!(|| vector().of_exactly(keyword_map_of!(("x", vector().of_exactly(many::<Vec<_>, _>(Test::entid()))),
("y", vector().of_exactly(many::<Vec<_>, _>(Test::entid()))))),
"[:x [] :y [1 2]]",
(Some(vec![]), Some(vec![1, 2])));
assert_edn_parses_to!(|| vector().of_exactly(keyword_map_of!(("x", vector().of_exactly(many::<Vec<_>, _>(Test::entid()))),
("y", vector().of_exactly(many::<Vec<_>, _>(Test::entid()))))),
"[]",
(None, None));
}
#[test]
fn test_keyword_map_failures() {
assert_parse_failure_contains!(|| vector().of_exactly(keyword_map_of!(("x", Test::entid()), ("y", Test::entid()))),
"[:x 1 :x 2]",
r#"errors: [Unexpected(Token(ValueAndSpan { inner: Keyword(Keyword("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
}
// assert_edn_parses_to!(|| keyword_map().or(vector()).map(|x| x.0.map(|x| x.clone().without_spans()).into_iter().collect::<Vec<_>>()), "{:x [1] :y [2]}", vec![]);
// assert_edn_parses_to!(|| keyword_map().or(vector()).of_exactly((Test::entid(), Test::entid())), "{:x [1] :y [2]}", (-1, 1));
// assert_edn_parses_to!(|| kw_map().of_exactly((Test::entid(), Test::entid())), "[:a 0 :b 0 1]", (1, 1));
// assert_edn_parses_to!(|| keyword_map_of(&[(":kw1", Test::entid()),
// (":kw2", (Test::entid(), Test::entid())),]),
// "{:kw1 0 :kw2 1 :x/y}", ((Some(0), Some((0, 1)))));
// let input = edn::parse::value("[:x/y]").expect("to be able to parse input as EDN");
// let par = vector().of_exactly(Test::entid());
// let stream: Stream = (&input).atom_stream();
// let result = par.skip(eof()).parse(stream).map(|x| x.0);
// assert_eq!(result, Ok(1));
// }
// #[test]
// fn test_keyword_map() {
// assert_keyword_map_eq!(
// "[:foo 1 2 3 :bar 4]",
// Some("{:foo [1 2 3] :bar [4]}"));
// // Trailing keywords aren't allowed.
// assert_keyword_map_eq!(
// "[:foo]",
// None);
// assert_keyword_map_eq!(
// "[:foo 2 :bar]",
// None);
// // Duplicate keywords aren't allowed.
// assert_keyword_map_eq!(
// "[:foo 2 :foo 1]",
// None);
// // Starting with anything but a keyword isn't allowed.
// assert_keyword_map_eq!(
// "[2 :foo 1]",
// None);
// // Consecutive keywords aren't allowed.
// assert_keyword_map_eq!(
// "[:foo :bar 1]",
// None);
// // Empty lists return an empty map.
// assert_keyword_map_eq!(
// "[]",
// Some("{}"));
// }
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
}