Compare commits

...

24 commits

Author SHA1 Message Date
Emily Toop
daaff7074d Addressed Review comments rnewman.
* Use better grammar in comments.
* Test for which error returned from `not-join` validation.
2017-04-07 18:26:36 +01:00
Emily Toop
649eff08a5 Rebase against https://github.com/mozilla/mentat/pull/402.
This is because it contains the latest version of parser and affects the structure of the code
2017-04-07 12:23:17 +01:00
Emily Toop
60592a87e6 Address review comments rnewman.
* Remove `WhereNotClause` and populate `NotJoin` with `WhereClause`.
* Fix validation for `not` and `not-join`, removing tests that were invalid.
* Address rustification comments.
2017-04-07 12:18:41 +01:00
Emily Toop
114f71fa5e Part 2 - Validate not and not-join pre-algebrization 2017-04-07 12:18:41 +01:00
Emily Toop
76d0223c2a Part 1 - Parse not and not-join 2017-04-07 12:18:41 +01:00
Emily Toop
4b45827be7 Fix merge conflicts 2017-04-07 12:17:39 +01:00
Nick Alexander
71d3aa29ed Part 3: Start binding fulltext values. 2017-04-05 15:31:05 -07:00
Nick Alexander
6544ca1594 Pre: Handle SrcVar. 2017-04-05 15:30:22 -07:00
Nick Alexander
62fda71fbc Part 2: Implement apply_fulltext and column constraints.
No bindings, yet.
2017-04-05 11:02:09 -07:00
Nick Alexander
08cae4bc7c Review comment: Find, not Bind. 2017-04-05 11:01:47 -07:00
Nick Alexander
c203046c16 Part 1: Parse functions in where clauses. 2017-04-03 16:46:11 -07:00
Nick Alexander
678a116130 Review comment: .map() as late as possible. 2017-04-03 13:15:21 -07:00
Nick Alexander
f7fb22ae7e Review comment: Extract and use def_matches_* macros. 2017-04-03 13:15:21 -07:00
Nick Alexander
0165a842ef 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.
2017-04-03 13:15:21 -07:00
Nick Alexander
f8e75d817e Review comment: nits. 2017-04-03 13:15:21 -07:00
Nick Alexander
7135eeac49 Review comment: Make Span and SpanPosition Copy. 2017-04-03 13:15:21 -07:00
Nick Alexander
720fbf3d01 Part 4: Use value_and_span apparatus in root crate. 2017-04-03 13:15:21 -07:00
Nick Alexander
ff136b2546 Part 3: Use value_and_span apparatus in query-parser/. 2017-04-03 13:15:21 -07:00
Nick Alexander
e947a32c59 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.
2017-04-03 13:12:41 -07:00
Nick Alexander
f4e2a0471f 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.
2017-04-03 09:15:36 -07:00
Nick Alexander
dcc05c643c Part 1b: Add keyword_map() parsing an edn::Value::Vector into an edn::Value::map. 2017-04-03 09:15:36 -07:00
Nick Alexander
8754cf224b 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.
2017-04-03 09:15:36 -07:00
Nick Alexander
f807b16db1 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.
2017-04-03 09:15:36 -07:00
Nick Alexander
c1409078fa Pre: Expose more in edn. 2017-04-03 09:15:36 -07:00
15 changed files with 920 additions and 43 deletions

View file

@ -38,8 +38,20 @@ pub type ResultParser<O, I> = Expected<FnParser<I, fn(I) -> ParseResult<O, I>>>;
#[macro_export] #[macro_export]
macro_rules! assert_parses_to { macro_rules! assert_parses_to {
( $parser: expr, $input: expr, $expected: expr ) => {{ ( $parser: expr, $input: expr, $expected: expr ) => {{
let mut par = $parser(); let par = $parser();
let result = par.parse($input.with_spans().into_atom_stream()).map(|x| x.0); // TODO: check remainder of stream. let result = par.skip(eof()).parse($input.with_spans().into_atom_stream()).map(|x| x.0);
assert_eq!(result, Ok($expected));
}}
}
/// `assert_edn_parses_to!` simplifies some of the boilerplate around running a parser function
/// against string input and expecting a certain result.
#[macro_export]
macro_rules! assert_edn_parses_to {
( $parser: expr, $input: expr, $expected: expr ) => {{
let par = $parser();
let input = edn::parse::value($input).expect("to be able to parse input as EDN");
let result = par.skip(eof()).parse(input.into_atom_stream()).map(|x| x.0);
assert_eq!(result, Ok($expected)); assert_eq!(result, Ok($expected));
}} }}
} }

View file

@ -56,11 +56,15 @@ use types::{
}; };
mod or; mod or;
mod not;
mod pattern; mod pattern;
mod predicate; mod predicate;
mod resolve; mod resolve;
use validate::validate_or_join; use validate::{
validate_not_join,
validate_or_join,
};
// We do this a lot for errors. // We do this a lot for errors.
trait RcCloned<T> { trait RcCloned<T> {
@ -96,6 +100,7 @@ fn unit_type_set(t: ValueType) -> HashSet<ValueType> {
/// ///
/// - Ordinary pattern clauses turn into `FROM` parts and `WHERE` parts using `=`. /// - Ordinary pattern clauses turn into `FROM` parts and `WHERE` parts using `=`.
/// - Predicate clauses turn into the same, but with other functions. /// - Predicate clauses turn into the same, but with other functions.
/// - Function clauses turn `WHERE` parts using function-specific comparisons.
/// - `not` turns into `NOT EXISTS` with `WHERE` clauses inside the subquery to /// - `not` turns into `NOT EXISTS` with `WHERE` clauses inside the subquery to
/// bind it to the outer variables, or adds simple `WHERE` clauses to the outer /// bind it to the outer variables, or adds simple `WHERE` clauses to the outer
/// clause. /// clause.
@ -581,11 +586,17 @@ impl ConjoiningClauses {
WhereClause::Pred(p) => { WhereClause::Pred(p) => {
self.apply_predicate(schema, p) self.apply_predicate(schema, p)
}, },
WhereClause::WhereFn(f) => {
self.apply_where_fn(schema, f)
},
WhereClause::OrJoin(o) => { WhereClause::OrJoin(o) => {
validate_or_join(&o) validate_or_join(&o)
//?; //?;
//self.apply_or_join(schema, o) //self.apply_or_join(schema, o)
}, },
WhereClause::NotJoin(n) => {
validate_not_join(&n)
},
_ => unimplemented!(), _ => unimplemented!(),
} }
} }

View file

@ -8,8 +8,6 @@
// CONDITIONS OF ANY KIND, either express or implied. See the License for the // CONDITIONS OF ANY KIND, either express or implied. See the License for the
// specific language governing permissions and limitations under the License. // specific language governing permissions and limitations under the License.
use std::rc::Rc;
use mentat_core::{ use mentat_core::{
Schema, Schema,
TypedValue, TypedValue,
@ -268,6 +266,7 @@ mod testing {
use super::*; use super::*;
use std::collections::BTreeMap; use std::collections::BTreeMap;
use std::rc::Rc;
use mentat_core::attribute::Unique; use mentat_core::attribute::Unique;
use mentat_core::{ use mentat_core::{

View file

@ -10,35 +10,48 @@
use mentat_core::{ use mentat_core::{
Schema, Schema,
TypedValue,
ValueType,
}; };
use mentat_query::{ use mentat_query::{
Binding,
FnArg,
NonIntegerConstant,
Predicate, Predicate,
SrcVar,
VariableOrPlaceholder,
WhereFn,
}; };
use clauses::ConjoiningClauses; use clauses::ConjoiningClauses;
use errors::{ use errors::{
Result, Error,
ErrorKind, ErrorKind,
Result,
}; };
use types::{ use types::{
ColumnConstraint, ColumnConstraint,
DatomsColumn,
DatomsTable,
FulltextColumn,
FulltextQualifiedAlias,
NumericComparison, NumericComparison,
QualifiedAlias,
QueryValue,
}; };
/// Application of predicates. /// Application of predicates.
impl ConjoiningClauses { impl ConjoiningClauses {
/// There are several kinds of predicates/functions in our Datalog: /// There are several kinds of predicates in our Datalog:
/// - A limited set of binary comparison operators: < > <= >= !=. /// - A limited set of binary comparison operators: < > <= >= !=.
/// These are converted into SQLite binary comparisons and some type constraints. /// These are converted into SQLite binary comparisons and some type constraints.
/// - A set of predicates like `fulltext` and `get-else` that are translated into
/// SQL `MATCH`es or joins, yielding bindings.
/// - In the future, some predicates that are implemented via function calls in SQLite. /// - In the future, some predicates that are implemented via function calls in SQLite.
/// ///
/// At present we have implemented only the five built-in comparison binary operators. /// At present we have implemented only the five built-in comparison binary operators.
pub fn apply_predicate<'s, 'p>(&mut self, schema: &'s Schema, predicate: Predicate) -> Result<()> { pub fn apply_predicate<'s>(&mut self, schema: &'s Schema, predicate: Predicate) -> Result<()> {
// Because we'll be growing the set of built-in predicates, handling each differently, // Because we'll be growing the set of built-in predicates, handling each differently,
// and ultimately allowing user-specified predicates, we match on the predicate name first. // and ultimately allowing user-specified predicates, we match on the predicate name first.
if let Some(op) = NumericComparison::from_datalog_operator(predicate.operator.0.as_str()) { if let Some(op) = NumericComparison::from_datalog_operator(predicate.operator.0.as_str()) {
@ -53,7 +66,7 @@ impl ConjoiningClauses {
/// - Ensures that the predicate functions name a known operator. /// - Ensures that the predicate functions name a known operator.
/// - Accumulates a `NumericInequality` constraint into the `wheres` list. /// - Accumulates a `NumericInequality` constraint into the `wheres` list.
#[allow(unused_variables)] #[allow(unused_variables)]
pub fn apply_numeric_predicate<'s, 'p>(&mut self, schema: &'s Schema, comparison: NumericComparison, predicate: Predicate) -> Result<()> { pub fn apply_numeric_predicate<'s>(&mut self, schema: &'s Schema, comparison: NumericComparison, predicate: Predicate) -> Result<()> {
if predicate.args.len() != 2 { if predicate.args.len() != 2 {
bail!(ErrorKind::InvalidNumberOfArguments(predicate.operator.clone(), predicate.args.len(), 2)); bail!(ErrorKind::InvalidNumberOfArguments(predicate.operator.clone(), predicate.args.len(), 2));
} }
@ -81,6 +94,161 @@ impl ConjoiningClauses {
self.wheres.add_intersection(constraint); self.wheres.add_intersection(constraint);
Ok(()) Ok(())
} }
/// There are several kinds of functions binding variables in our Datalog:
/// - A set of functions like `fulltext` and `get-else` that are translated into
/// SQL `MATCH`es or joins, yielding bindings.
/// - In the future, some functions that are implemented via function calls in SQLite.
///
/// At present we have implemented only the `fulltext` operator.
pub fn apply_where_fn<'s>(&mut self, schema: &'s Schema, where_fn: WhereFn) -> Result<()> {
// Because we'll be growing the set of built-in functions, handling each differently, and
// ultimately allowing user-specified functions, we match on the function name first.
match where_fn.operator.0.as_str() {
"fulltext" => self.apply_fulltext(schema, where_fn),
_ => bail!(ErrorKind::UnknownFunction(where_fn.operator.clone())),
}
}
/// This function:
/// - Resolves variables and converts types to those more amenable to SQL.
/// - Ensures that the predicate functions name a known operator.
/// - Accumulates a `NumericInequality` constraint into the `wheres` list.
#[allow(unused_variables)]
pub fn apply_fulltext<'s>(&mut self, schema: &'s Schema, where_fn: WhereFn) -> Result<()> {
if where_fn.args.len() != 3 {
bail!(ErrorKind::InvalidNumberOfArguments(where_fn.operator.clone(), where_fn.args.len(), 3));
}
// TODO: binding-specific error messages.
let mut bindings = match where_fn.binding {
Binding::BindRel(bindings) => {
if bindings.len() > 4 {
bail!(ErrorKind::InvalidNumberOfArguments(where_fn.operator.clone(), bindings.len(), 4));
}
bindings.into_iter()
},
_ => bail!(ErrorKind::InvalidArgument(where_fn.operator.clone(), "bindings".into(), 999)),
};
// Go from arguments -- parser output -- to columns or values.
// Any variables that aren't bound by this point in the linear processing of clauses will
// cause the application of the predicate to fail.
let mut args = where_fn.args.into_iter();
// TODO: process source variables.
match args.next().unwrap() {
FnArg::SrcVar(SrcVar::DefaultSrc) => {},
_ => bail!(ErrorKind::InvalidArgument(where_fn.operator.clone(), "source variable".into(), 0)),
}
// TODO: accept placeholder and set of attributes. Alternately, consider putting the search
// term before the attribute arguments and collect the (variadic) attributes into a set.
// let a: Entid = self.resolve_attribute_argument(&where_fn.operator, 1, args.next().unwrap())?;
//
// TODO: allow non-constant attributes.
// TODO: improve the expression of this matching, possibly by using attribute_for_* uniformly.
let a = match args.next().unwrap() {
FnArg::Ident(i) => schema.get_entid(&i),
// Must be an entid.
FnArg::EntidOrInteger(e) => Some(e),
_ => None,
};
let a = a.ok_or(ErrorKind::InvalidArgument(where_fn.operator.clone(), "attribute".into(), 1))?;
let attribute = schema.attribute_for_entid(a).cloned().ok_or(ErrorKind::InvalidArgument(where_fn.operator.clone(), "attribute".into(), 1))?;
let fulltext_values = DatomsTable::FulltextValues;
let datoms_table = DatomsTable::Datoms;
let fulltext_values_alias = (self.aliaser)(fulltext_values);
let datoms_table_alias = (self.aliaser)(datoms_table);
// TODO: constrain types in more general cases?
self.constrain_attribute(datoms_table_alias.clone(), a);
self.wheres.add_intersection(ColumnConstraint::Equals(
QualifiedAlias(datoms_table_alias.clone(), DatomsColumn::Value),
QueryValue::FulltextColumn(FulltextQualifiedAlias(fulltext_values_alias.clone(), FulltextColumn::Rowid))));
// search is either text or a variable.
// TODO: should this just use `resolve_argument`? Should it add a new `resolve_*` function?
let search = match args.next().unwrap() {
FnArg::Variable(var) => {
self.column_bindings
.get(&var)
.and_then(|cols| cols.first().map(|col| QueryValue::Column(col.clone())))
.ok_or_else(|| Error::from_kind(ErrorKind::UnboundVariable(var.name())))?
},
FnArg::Constant(NonIntegerConstant::Text(s)) => {
QueryValue::TypedValue(TypedValue::typed_string(s.as_str()))
},
_ => bail!(ErrorKind::InvalidArgument(where_fn.operator.clone(), "string".into(), 2)),
};
// TODO: should we build the FQA in ::Matches, preventing nonsense like matching on ::Rowid?
let constraint = ColumnConstraint::Matches(FulltextQualifiedAlias(fulltext_values_alias.clone(), FulltextColumn::Text), search);
self.wheres.add_intersection(constraint);
if let Some(VariableOrPlaceholder::Variable(var)) = bindings.next() {
// TODO: can we just check for late binding here?
// Do we have, or will we have, an external binding for this variable?
if self.bound_value(&var).is_some() || self.input_variables.contains(&var) {
// That's a paddlin'!
bail!(ErrorKind::InvalidArgument(where_fn.operator.clone(), "illegal bound variable".into(), 999))
}
self.constrain_var_to_type(var.clone(), ValueType::Ref);
let entity_alias = QualifiedAlias(datoms_table_alias.clone(), DatomsColumn::Entity);
self.column_bindings.entry(var).or_insert(vec![]).push(entity_alias);
}
if let Some(VariableOrPlaceholder::Variable(var)) = bindings.next() {
// TODO: can we just check for late binding here?
// Do we have, or will we have, an external binding for this variable?
if self.bound_value(&var).is_some() || self.input_variables.contains(&var) {
// That's a paddlin'!
bail!(ErrorKind::InvalidArgument(where_fn.operator.clone(), "illegal bound variable".into(), 999))
}
self.constrain_var_to_type(var.clone(), ValueType::String);
// TODO: figure out how to represent a FulltextQualifiedAlias.
// let value_alias = FulltextQualifiedAlias(fulltext_values_alias.clone(), FulltextColumn::Text);
// self.column_bindings.entry(var).or_insert(vec![]).push(value_alias);
}
if let Some(VariableOrPlaceholder::Variable(var)) = bindings.next() {
// TODO: can we just check for late binding here?
// Do we have, or will we have, an external binding for this variable?
if self.bound_value(&var).is_some() || self.input_variables.contains(&var) {
// That's a paddlin'!
bail!(ErrorKind::InvalidArgument(where_fn.operator.clone(), "illegal bound variable".into(), 999))
}
self.constrain_var_to_type(var.clone(), ValueType::Ref);
let tx_alias = QualifiedAlias(datoms_table_alias.clone(), DatomsColumn::Tx);
self.column_bindings.entry(var).or_insert(vec![]).push(tx_alias);
}
if let Some(VariableOrPlaceholder::Variable(var)) = bindings.next() {
// TODO: can we just check for late binding here?
// Do we have, or will we have, an external binding for this variable?
if self.bound_value(&var).is_some() || self.input_variables.contains(&var) {
// That's a paddlin'!
bail!(ErrorKind::InvalidArgument(where_fn.operator.clone(), "illegal bound variable".into(), 999))
}
self.constrain_var_to_type(var.clone(), ValueType::Double);
// TODO: produce this using SQLite's matchinfo.
self.value_bindings.insert(var.clone(), TypedValue::Double(0.0.into()));
// TODO: figure out how to represent a constant binding.
// self.column_bindings.entry(var).or_insert(vec![]).push(score_alias);
}
Ok(())
}
} }
#[cfg(test)] #[cfg(test)]
@ -88,6 +256,8 @@ mod testing {
use super::*; use super::*;
use std::collections::HashSet; use std::collections::HashSet;
use std::rc::Rc;
use mentat_core::attribute::Unique; use mentat_core::attribute::Unique;
use mentat_core::{ use mentat_core::{
Attribute, Attribute,
@ -96,13 +266,16 @@ mod testing {
}; };
use mentat_query::{ use mentat_query::{
Binding,
FnArg, FnArg,
NamespacedKeyword, NamespacedKeyword,
Pattern, Pattern,
PatternNonValuePlace, PatternNonValuePlace,
PatternValuePlace, PatternValuePlace,
PlainSymbol, PlainSymbol,
SrcVar,
Variable, Variable,
VariableOrPlaceholder,
}; };
use clauses::{ use clauses::{
@ -229,4 +402,68 @@ mod testing {
.collect(), .collect(),
ValueType::String)); ValueType::String));
} }
#[test]
fn test_apply_fulltext() {
let mut cc = ConjoiningClauses::default();
let mut schema = Schema::default();
associate_ident(&mut schema, NamespacedKeyword::new("foo", "fts"), 100);
add_attribute(&mut schema, 100, Attribute {
value_type: ValueType::String,
index: true,
fulltext: true,
..Default::default()
});
let op = PlainSymbol::new("fulltext");
cc.apply_fulltext(&schema, WhereFn {
operator: op,
args: vec![
FnArg::SrcVar(SrcVar::DefaultSrc),
FnArg::Ident(NamespacedKeyword::new("foo", "fts")),
FnArg::Constant(NonIntegerConstant::Text(Rc::new("needle".into()))),
],
binding: Binding::BindRel(vec![VariableOrPlaceholder::Variable(Variable::from_valid_name("?entity")),
VariableOrPlaceholder::Variable(Variable::from_valid_name("?value")),
VariableOrPlaceholder::Variable(Variable::from_valid_name("?tx")),
VariableOrPlaceholder::Variable(Variable::from_valid_name("?score"))]),
}).expect("to be able to apply_fulltext");
assert!(!cc.is_known_empty);
// Finally, expand column bindings.
cc.expand_column_bindings();
assert!(!cc.is_known_empty);
let clauses = cc.wheres;
assert_eq!(clauses.len(), 3);
assert_eq!(clauses.0[0], ColumnConstraint::Equals(QualifiedAlias("datoms01".to_string(), DatomsColumn::Attribute),
QueryValue::Entid(100)).into());
assert_eq!(clauses.0[1], ColumnConstraint::Equals(QualifiedAlias("datoms01".to_string(), DatomsColumn::Value),
QueryValue::FulltextColumn(FulltextQualifiedAlias("fulltext_values00".to_string(), FulltextColumn::Rowid))).into());
assert_eq!(clauses.0[2], ColumnConstraint::Matches(FulltextQualifiedAlias("fulltext_values00".to_string(), FulltextColumn::Text),
QueryValue::TypedValue(TypedValue::String(Rc::new("needle".into())))).into());
let bindings = cc.column_bindings;
assert_eq!(bindings.len(), 2);
assert_eq!(bindings.get(&Variable::from_valid_name("?entity")).expect("column binding for ?entity").clone(),
vec![QualifiedAlias("datoms01".to_string(), DatomsColumn::Entity)]);
assert_eq!(bindings.get(&Variable::from_valid_name("?tx")).expect("column binding for ?tx").clone(),
vec![QualifiedAlias("datoms01".to_string(), DatomsColumn::Tx)]);
let known_types = cc.known_types;
assert_eq!(known_types.len(), 4);
assert_eq!(known_types.get(&Variable::from_valid_name("?entity")).expect("known types for ?entity").clone(),
vec![ValueType::Ref].into_iter().collect());
assert_eq!(known_types.get(&Variable::from_valid_name("?value")).expect("known types for ?value").clone(),
vec![ValueType::String].into_iter().collect());
assert_eq!(known_types.get(&Variable::from_valid_name("?tx")).expect("known types for ?tx").clone(),
vec![ValueType::Ref].into_iter().collect());
assert_eq!(known_types.get(&Variable::from_valid_name("?score")).expect("known types for ?score").clone(),
vec![ValueType::Double].into_iter().collect());
}
} }

View file

@ -56,7 +56,7 @@ impl ConjoiningClauses {
Constant(NonIntegerConstant::Text(_)) | Constant(NonIntegerConstant::Text(_)) |
Constant(NonIntegerConstant::BigInteger(_)) => { Constant(NonIntegerConstant::BigInteger(_)) => {
self.mark_known_empty(EmptyBecause::NonNumericArgument); self.mark_known_empty(EmptyBecause::NonNumericArgument);
bail!(ErrorKind::NonNumericArgument(function.clone(), position)); bail!(ErrorKind::InvalidArgument(function.clone(), "numeric".into(), position));
}, },
Constant(NonIntegerConstant::Float(f)) => Ok(QueryValue::TypedValue(TypedValue::Double(f))), Constant(NonIntegerConstant::Float(f)) => Ok(QueryValue::TypedValue(TypedValue::Double(f))),
} }

View file

@ -35,9 +35,9 @@ error_chain! {
display("unbound variable: {}", name) display("unbound variable: {}", name)
} }
NonNumericArgument(function: PlainSymbol, position: usize) { InvalidArgument(function: PlainSymbol, expected_type: String, position: usize) {
description("invalid argument") description("invalid argument")
display("invalid argument to {}: expected numeric in position {}.", function, position) display("invalid argument to {}: expected {} in position {}.", function, expected_type, position)
} }
NonMatchingVariablesInOrClause { NonMatchingVariablesInOrClause {
@ -45,6 +45,12 @@ error_chain! {
description("non-matching variables in 'or' clause") description("non-matching variables in 'or' clause")
display("non-matching variables in 'or' clause") display("non-matching variables in 'or' clause")
} }
NonMatchingVariablesInNotClause {
// TODO: flesh out.
description("non-matching variables in 'not' clause")
display("non-matching variables in 'not' clause")
}
} }
} }

View file

@ -99,6 +99,8 @@ pub use types::{
ColumnIntersection, ColumnIntersection,
DatomsColumn, DatomsColumn,
DatomsTable, DatomsTable,
FulltextColumn,
FulltextQualifiedAlias,
QualifiedAlias, QualifiedAlias,
QueryValue, QueryValue,
SourceAlias, SourceAlias,

View file

@ -71,6 +71,23 @@ impl DatomsColumn {
} }
} }
/// One of the named columns of our fulltext values table.
#[derive(PartialEq, Eq, Clone, Debug)]
pub enum FulltextColumn {
Rowid,
Text,
}
impl FulltextColumn {
pub fn as_str(&self) -> &'static str {
use self::FulltextColumn::*;
match *self {
Rowid => "rowid",
Text => "text",
}
}
}
/// A specific instance of a table within a query. E.g., "datoms123". /// A specific instance of a table within a query. E.g., "datoms123".
pub type TableAlias = String; pub type TableAlias = String;
@ -94,6 +111,16 @@ impl Debug for QualifiedAlias {
} }
} }
/// A particular column of a particular aliased fulltext table. E.g., "fulltext_values123", Rowid.
#[derive(PartialEq, Eq, Clone)]
pub struct FulltextQualifiedAlias(pub TableAlias, pub FulltextColumn);
impl Debug for FulltextQualifiedAlias {
fn fmt(&self, f: &mut Formatter) -> Result {
write!(f, "{}.{}", self.0, self.1.as_str())
}
}
impl QualifiedAlias { impl QualifiedAlias {
pub fn for_type_tag(&self) -> QualifiedAlias { pub fn for_type_tag(&self) -> QualifiedAlias {
QualifiedAlias(self.0.clone(), DatomsColumn::ValueTypeTag) QualifiedAlias(self.0.clone(), DatomsColumn::ValueTypeTag)
@ -103,6 +130,7 @@ impl QualifiedAlias {
#[derive(PartialEq, Eq)] #[derive(PartialEq, Eq)]
pub enum QueryValue { pub enum QueryValue {
Column(QualifiedAlias), Column(QualifiedAlias),
FulltextColumn(FulltextQualifiedAlias),
Entid(Entid), Entid(Entid),
TypedValue(TypedValue), TypedValue(TypedValue),
@ -120,6 +148,9 @@ impl Debug for QueryValue {
&Column(ref qa) => { &Column(ref qa) => {
write!(f, "{:?}", qa) write!(f, "{:?}", qa)
}, },
&FulltextColumn(ref qa) => {
write!(f, "{:?}", qa)
},
&Entid(ref entid) => { &Entid(ref entid) => {
write!(f, "entity({:?})", entid) write!(f, "entity({:?})", entid)
}, },
@ -192,6 +223,9 @@ pub enum ColumnConstraint {
right: QueryValue, right: QueryValue,
}, },
HasType(TableAlias, ValueType), HasType(TableAlias, ValueType),
// TODO: Merge this with NumericInequality? I expect the fine-grained information to be
// valuable when optimizing.
Matches(FulltextQualifiedAlias, QueryValue),
} }
#[derive(PartialEq, Eq, Debug)] #[derive(PartialEq, Eq, Debug)]
@ -290,6 +324,10 @@ impl Debug for ColumnConstraint {
write!(f, "{:?} {:?} {:?}", left, operator, right) write!(f, "{:?} {:?} {:?}", left, operator, right)
}, },
&Matches(ref qa, ref thing) => {
write!(f, "{:?} MATCHES {:?}", qa, thing)
},
&HasType(ref qa, value_type) => { &HasType(ref qa, value_type) => {
write!(f, "{:?}.value_type_tag = {:?}", qa, value_type) write!(f, "{:?}.value_type_tag = {:?}", qa, value_type)
}, },
@ -301,6 +339,7 @@ impl Debug for ColumnConstraint {
pub enum EmptyBecause { pub enum EmptyBecause {
// Var, existing, desired. // Var, existing, desired.
TypeMismatch(Variable, HashSet<ValueType>, ValueType), TypeMismatch(Variable, HashSet<ValueType>, ValueType),
NonAttributeArgument,
NonNumericArgument, NonNumericArgument,
NonStringFulltextValue, NonStringFulltextValue,
UnresolvedIdent(NamespacedKeyword), UnresolvedIdent(NamespacedKeyword),
@ -319,6 +358,9 @@ impl Debug for EmptyBecause {
write!(f, "Type mismatch: {:?} can't be {:?}, because it's already {:?}", write!(f, "Type mismatch: {:?} can't be {:?}, because it's already {:?}",
var, desired, existing) var, desired, existing)
}, },
&NonAttributeArgument => {
write!(f, "Non-attribute argument in attribute place")
},
&NonNumericArgument => { &NonNumericArgument => {
write!(f, "Non-numeric argument in numeric place") write!(f, "Non-numeric argument in numeric place")
}, },

View file

@ -13,6 +13,7 @@ use std::collections::BTreeSet;
use mentat_query::{ use mentat_query::{
ContainsVariables, ContainsVariables,
OrJoin, OrJoin,
NotJoin,
Variable, Variable,
UnifyVars, UnifyVars,
}; };
@ -74,6 +75,23 @@ pub fn validate_or_join(or_join: &OrJoin) -> Result<()> {
} }
} }
pub fn validate_not_join(not_join: &NotJoin) -> Result<()> {
// Grab our mentioned variables and ensure that the rules are followed.
match not_join.unify_vars {
UnifyVars::Implicit => {
Ok(())
},
UnifyVars::Explicit(ref vars) => {
// The joined vars must each appear somewhere in the clauses' mentioned variables.
let var_set: BTreeSet<Variable> = vars.iter().cloned().collect();
if !var_set.is_subset(&not_join.collect_mentioned_variables()) {
bail!(ErrorKind::NonMatchingVariablesInNotClause);
}
Ok(())
},
}
}
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
extern crate mentat_core; extern crate mentat_core;
@ -96,7 +114,15 @@ mod tests {
use clauses::ident; use clauses::ident;
use super::validate_or_join; use errors::{
Error,
ErrorKind,
};
use super::{
validate_not_join,
validate_or_join,
};
fn value_ident(ns: &str, name: &str) -> PatternValuePlace { fn value_ident(ns: &str, name: &str) -> PatternValuePlace {
PatternValuePlace::IdentOrKeyword(::std::rc::Rc::new(NamespacedKeyword::new(ns, name))) PatternValuePlace::IdentOrKeyword(::std::rc::Rc::new(NamespacedKeyword::new(ns, name)))
@ -229,4 +255,131 @@ mod tests {
_ => panic!(), _ => panic!(),
}; };
} }
/// Tests that the top-level form is a valid `not`, returning the clauses.
fn valid_not_join(parsed: FindQuery, expected_unify: UnifyVars) -> Vec<WhereClause> {
// Filter out all the clauses that are not `not`s.
let mut nots = parsed.where_clauses.into_iter().filter(|x| match x {
&WhereClause::NotJoin(_) => true,
_ => false,
});
// There should be only one not clause.
let clause = nots.next().unwrap();
assert_eq!(None, nots.next());
match clause {
WhereClause::NotJoin(not_join) => {
// It's valid: the variables are the same in each branch.
assert_eq!((), validate_not_join(&not_join).unwrap());
assert_eq!(expected_unify, not_join.unify_vars);
not_join.clauses
},
_ => panic!(),
}
}
/// Test that a `not` is valid if it is implicit.
#[test]
fn test_success_not() {
let query = r#"[:find ?name
:where [?id :artist/name ?name]
(not [?id :artist/country :country/CA]
[?id :artist/country :country/GB])]"#;
let parsed = parse_find_string(query).expect("expected successful parse");
let clauses = valid_not_join(parsed, UnifyVars::Implicit);
// Check each part of the body
let mut parts = clauses.into_iter();
match (parts.next(), parts.next(), parts.next()) {
(Some(clause1), Some(clause2), None) => {
assert_eq!(
clause1,
WhereClause::Pattern(Pattern {
source: None,
entity: PatternNonValuePlace::Variable(Variable::from_valid_name("?id")),
attribute: ident("artist", "country"),
value: value_ident("country", "CA"),
tx: PatternNonValuePlace::Placeholder,
}));
assert_eq!(
clause2,
WhereClause::Pattern(Pattern {
source: None,
entity: PatternNonValuePlace::Variable(Variable::from_valid_name("?id")),
attribute: ident("artist", "country"),
value: value_ident("country", "GB"),
tx: PatternNonValuePlace::Placeholder,
}));
},
_ => panic!(),
};
}
#[test]
fn test_success_not_join() {
let query = r#"[:find ?artist
:where [?artist :artist/name]
(not-join [?artist]
[?release :release/artists ?artist]
[?release :release/year 1970])]"#;
let parsed = parse_find_string(query).expect("expected successful parse");
let clauses = valid_not_join(parsed, UnifyVars::Explicit(vec![Variable::from_valid_name("?artist")]));
// Let's do some detailed parse checks.
let mut parts = clauses.into_iter();
match (parts.next(), parts.next(), parts.next()) {
(Some(clause1), Some(clause2), None) => {
assert_eq!(
clause1,
WhereClause::Pattern(Pattern {
source: None,
entity: PatternNonValuePlace::Variable(Variable::from_valid_name("?release")),
attribute: ident("release", "artists"),
value: PatternValuePlace::Variable(Variable::from_valid_name("?artist")),
tx: PatternNonValuePlace::Placeholder,
}));
assert_eq!(
clause2,
WhereClause::Pattern(Pattern {
source: None,
entity: PatternNonValuePlace::Variable(Variable::from_valid_name("?release")),
attribute: ident("release", "year"),
value: PatternValuePlace::EntidOrInteger(1970),
tx: PatternNonValuePlace::Placeholder,
}));
},
_ => panic!(),
};
}
/// Test that a `not-join` that does not use the joining var fails to validate.
#[test]
fn test_invalid_explicit_not_join_non_matching_join_vars() {
let query = r#"[:find ?artist
:where [?artist :artist/name]
(not-join [?artist]
[?release :release/artists "Pink Floyd"]
[?release :release/year 1970])]"#;
let parsed = parse_find_string(query).expect("expected successful parse");
let mut nots = parsed.where_clauses.iter().filter(|&x| match *x {
WhereClause::NotJoin(_) => true,
_ => false,
});
let clause = nots.next().unwrap().clone();
assert_eq!(None, nots.next());
match clause {
WhereClause::NotJoin(not_join) => { let result = validate_not_join(&not_join);
assert!(result.is_err());
match result.err().unwrap() {
Error(ErrorKind::NonMatchingVariablesInNotClause, _) => { assert!(true); },
x => panic!("expected NonMatchingVariablesInNotClause error, got {:?}", x),
}
},
_ => panic!(),
}
}
} }

View file

@ -35,6 +35,7 @@ use self::mentat_parser_utils::value_and_span::{
}; };
use self::mentat_query::{ use self::mentat_query::{
Binding,
Element, Element,
FindQuery, FindQuery,
FindSpec, FindSpec,
@ -42,6 +43,7 @@ use self::mentat_query::{
FromValue, FromValue,
OrJoin, OrJoin,
OrWhereClause, OrWhereClause,
NotJoin,
Pattern, Pattern,
PatternNonValuePlace, PatternNonValuePlace,
PatternValuePlace, PatternValuePlace,
@ -50,7 +52,9 @@ use self::mentat_query::{
SrcVar, SrcVar,
UnifyVars, UnifyVars,
Variable, Variable,
VariableOrPlaceholder,
WhereClause, WhereClause,
WhereFn,
}; };
error_chain! { error_chain! {
@ -129,11 +133,56 @@ def_parser!(Where, pattern_non_value_place, PatternNonValuePlace, {
satisfy_map(PatternNonValuePlace::from_value) satisfy_map(PatternNonValuePlace::from_value)
}); });
def_matches_plain_symbol!(Where, and, "and");
def_matches_plain_symbol!(Where, or, "or"); def_parser!(Where, and, edn::ValueAndSpan, {
satisfy(|v: edn::ValueAndSpan| {
if let edn::SpannedValue::PlainSymbol(ref s) = v.inner {
s.0.as_str() == "and"
} else {
false
}
})
});
def_matches_plain_symbol!(Where, or_join, "or-join"); def_parser!(Where, or, edn::ValueAndSpan, {
satisfy(|v: edn::ValueAndSpan| {
if let edn::SpannedValue::PlainSymbol(ref s) = v.inner {
s.0.as_str() == "or"
} else {
false
}
})
});
def_parser!(Where, or_join, edn::ValueAndSpan, {
satisfy(|v: edn::ValueAndSpan| {
if let edn::SpannedValue::PlainSymbol(ref s) = v.inner {
s.0.as_str() == "or-join"
} else {
false
}
})
});
def_parser!(Where, not, edn::ValueAndSpan, {
satisfy(|v: edn::ValueAndSpan| {
if let edn::SpannedValue::PlainSymbol(ref s) = v.inner {
s.0.as_str() == "not"
} else {
false
}
})
});
def_parser!(Where, not_join, edn::ValueAndSpan, {
satisfy(|v: edn::ValueAndSpan| {
if let edn::SpannedValue::PlainSymbol(ref s) = v.inner {
s.0.as_str() == "not-join"
} else {
false
}
})
});
def_parser!(Where, rule_vars, Vec<Variable>, { def_parser!(Where, rule_vars, Vec<Variable>, {
seq() seq()
@ -182,6 +231,33 @@ def_parser!(Where, or_join_clause, WhereClause, {
})) }))
}); });
def_parser!(Where, not_clause, WhereClause, {
seq()
.of_exactly(Where::not()
.with(many1(Where::clause()))
.map(|clauses| {
WhereClause::NotJoin(
NotJoin {
unify_vars: UnifyVars::Implicit,
clauses: clauses,
})
}))
});
def_parser!(Where, not_join_clause, WhereClause, {
seq()
.of_exactly(Where::not_join()
.with(Where::rule_vars())
.and(many1(Where::clause()))
.map(|(vars, clauses)| {
WhereClause::NotJoin(
NotJoin {
unify_vars: UnifyVars::Explicit(vars),
clauses: clauses,
})
}))
});
/// A vector containing just a parenthesized filter expression. /// A vector containing just a parenthesized filter expression.
def_parser!(Where, pred, WhereClause, { def_parser!(Where, pred, WhereClause, {
// Accept either a nested list or a nested vector here: // Accept either a nested list or a nested vector here:
@ -198,6 +274,25 @@ def_parser!(Where, pred, WhereClause, {
}))) })))
}); });
/// A vector containing a parenthesized function expression and a binding.
def_parser!(Where, where_fn, WhereClause, {
// Accept either a nested list or a nested vector here:
// `[(foo ?x ?y) binding]` or `[[foo ?x ?y] binding]`
vector()
.of_exactly(
(seq().of_exactly(
(Query::predicate_fn(), Query::arguments())),
Bind::binding())
.map(|((f, args), binding)| {
WhereClause::WhereFn(
WhereFn {
operator: f.0,
args: args,
binding: binding,
})
}))
});
def_parser!(Where, pattern, WhereClause, { def_parser!(Where, pattern, WhereClause, {
vector() vector()
.of_exactly( .of_exactly(
@ -246,8 +341,11 @@ def_parser!(Where, clause, WhereClause, {
// We don't yet handle source vars. // We don't yet handle source vars.
try(Where::or_join_clause()), try(Where::or_join_clause()),
try(Where::or_clause()), try(Where::or_clause()),
try(Where::not_join_clause()),
try(Where::not_clause()),
try(Where::pred()), try(Where::pred()),
try(Where::where_fn()),
]) ])
}); });
@ -262,6 +360,8 @@ def_matches_plain_symbol!(Find, period, ".");
def_matches_plain_symbol!(Find, ellipsis, "..."); def_matches_plain_symbol!(Find, ellipsis, "...");
def_matches_plain_symbol!(Find, placeholder, "_");
def_parser!(Find, find_scalar, FindSpec, { def_parser!(Find, find_scalar, FindSpec, {
Query::variable() Query::variable()
.skip(Find::period()) .skip(Find::period())
@ -366,6 +466,47 @@ def_parser!(Find, query, FindQuery, {
}) })
}); });
pub struct Bind;
def_parser!(Bind, bind_scalar, Binding, {
Query::variable()
.skip(eof())
.map(|var: Variable| -> Binding { Binding::BindScalar(var) })
});
def_parser!(Bind, variable_or_placeholder, VariableOrPlaceholder, {
Query::variable().map(VariableOrPlaceholder::Variable)
.or(Find::placeholder().map(|_| VariableOrPlaceholder::Placeholder))
});
def_parser!(Bind, bind_coll, Binding, {
vector()
.of_exactly(Query::variable()
.skip(Find::ellipsis()))
.map(Binding::BindColl)
});
def_parser!(Bind, bind_rel, Binding, {
vector().of_exactly(
many1::<Vec<VariableOrPlaceholder>, _>(Bind::variable_or_placeholder())
.map(Binding::BindRel))
});
def_parser!(Bind, bind_tuple, Binding, {
many1::<Vec<VariableOrPlaceholder>, _>(Bind::variable_or_placeholder())
.skip(eof())
.map(Binding::BindTuple)
});
def_parser!(Bind, binding, Binding, {
// Any one of the four binding types might apply, so we combine them with `choice`. Our parsers
// consume input, so we need to wrap them in `try` so that they operate independently.
choice([try(Bind::bind_scalar()),
try(Bind::bind_coll()),
try(Bind::bind_tuple()),
try(Bind::bind_rel())])
});
pub fn parse_find_string(string: &str) -> Result<FindQuery> { pub fn parse_find_string(string: &str) -> Result<FindQuery> {
let expr = edn::parse::value(string)?; let expr = edn::parse::value(string)?;
Find::query() Find::query()
@ -385,6 +526,7 @@ mod test {
use self::combine::Parser; use self::combine::Parser;
use self::edn::OrderedFloat; use self::edn::OrderedFloat;
use self::mentat_query::{ use self::mentat_query::{
Binding,
Element, Element,
FindSpec, FindSpec,
NonIntegerConstant, NonIntegerConstant,
@ -393,6 +535,7 @@ mod test {
PatternValuePlace, PatternValuePlace,
SrcVar, SrcVar,
Variable, Variable,
VariableOrPlaceholder,
}; };
use super::*; use super::*;
@ -548,6 +691,49 @@ mod test {
})); }));
} }
#[test]
fn test_not() {
let e = edn::PlainSymbol::new("?e");
let a = edn::PlainSymbol::new("?a");
let v = edn::PlainSymbol::new("?v");
assert_edn_parses_to!(Where::not_clause,
"(not [?e ?a ?v])",
WhereClause::NotJoin(
NotJoin {
unify_vars: UnifyVars::Implicit,
clauses: vec![
WhereClause::Pattern(Pattern {
source: None,
entity: PatternNonValuePlace::Variable(variable(e)),
attribute: PatternNonValuePlace::Variable(variable(a)),
value: PatternValuePlace::Variable(variable(v)),
tx: PatternNonValuePlace::Placeholder,
})],
}));
}
#[test]
fn test_not_join() {
let e = edn::PlainSymbol::new("?e");
let a = edn::PlainSymbol::new("?a");
let v = edn::PlainSymbol::new("?v");
assert_edn_parses_to!(Where::not_join_clause,
"(not-join [?e] [?e ?a ?v])",
WhereClause::NotJoin(
NotJoin {
unify_vars: UnifyVars::Explicit(vec![variable(e.clone())]),
clauses: vec![WhereClause::Pattern(Pattern {
source: None,
entity: PatternNonValuePlace::Variable(variable(e)),
attribute: PatternNonValuePlace::Variable(variable(a)),
value: PatternValuePlace::Variable(variable(v)),
tx: PatternNonValuePlace::Placeholder,
})],
}));
}
#[test] #[test]
fn test_find_sp_variable() { fn test_find_sp_variable() {
let sym = edn::PlainSymbol::new("?x"); let sym = edn::PlainSymbol::new("?x");
@ -598,4 +784,86 @@ mod test {
FindSpec::FindTuple(vec![Element::Variable(variable(vx)), FindSpec::FindTuple(vec![Element::Variable(variable(vx)),
Element::Variable(variable(vy))])); Element::Variable(variable(vy))]));
} }
#[test]
fn test_bind_scalar() {
let vx = edn::PlainSymbol::new("?x");
assert_edn_parses_to!(|| vector().of_exactly(Bind::binding()),
"[?x]",
Binding::BindScalar(variable(vx)));
}
#[test]
fn test_bind_coll() {
let vx = edn::PlainSymbol::new("?x");
assert_edn_parses_to!(|| vector().of_exactly(Bind::binding()),
"[[?x ...]]",
Binding::BindColl(variable(vx)));
}
#[test]
fn test_bind_rel() {
let vx = edn::PlainSymbol::new("?x");
let vy = edn::PlainSymbol::new("?y");
let vw = edn::PlainSymbol::new("?w");
assert_edn_parses_to!(|| vector().of_exactly(Bind::binding()),
"[[?x ?y _ ?w]]",
Binding::BindRel(vec![VariableOrPlaceholder::Variable(variable(vx)),
VariableOrPlaceholder::Variable(variable(vy)),
VariableOrPlaceholder::Placeholder,
VariableOrPlaceholder::Variable(variable(vw)),
]));
}
#[test]
fn test_bind_tuple() {
let vx = edn::PlainSymbol::new("?x");
let vy = edn::PlainSymbol::new("?y");
let vw = edn::PlainSymbol::new("?w");
assert_edn_parses_to!(|| vector().of_exactly(Bind::binding()),
"[?x ?y _ ?w]",
Binding::BindTuple(vec![VariableOrPlaceholder::Variable(variable(vx)),
VariableOrPlaceholder::Variable(variable(vy)),
VariableOrPlaceholder::Placeholder,
VariableOrPlaceholder::Variable(variable(vw)),
]));
}
#[test]
fn test_where_fn() {
assert_edn_parses_to!(Where::where_fn,
"[(f ?x 1) ?y]",
WhereClause::WhereFn(WhereFn {
operator: edn::PlainSymbol::new("f"),
args: vec![FnArg::Variable(Variable::from_valid_name("?x")),
FnArg::EntidOrInteger(1)],
binding: Binding::BindScalar(Variable::from_valid_name("?y")),
}));
assert_edn_parses_to!(Where::where_fn,
"[(f ?x) [?y ...]]",
WhereClause::WhereFn(WhereFn {
operator: edn::PlainSymbol::new("f"),
args: vec![FnArg::Variable(Variable::from_valid_name("?x"))],
binding: Binding::BindColl(Variable::from_valid_name("?y")),
}));
// assert_edn_parses_to!(Where::where_fn,
// "[(f) [?y _]]",
// WhereClause::WhereFn(WhereFn {
// operator: edn::PlainSymbol::new("f"),
// args: vec![],
// binding: Binding::BindRel(vec![VariableOrPlaceholder::Variable(Variable::from_valid_name("?y")),
// VariableOrPlaceholder::Placeholder]),
// }));
assert_edn_parses_to!(Where::where_fn,
"[(f) _ ?y]",
WhereClause::WhereFn(WhereFn {
operator: edn::PlainSymbol::new("f"),
args: vec![],
binding: Binding::BindTuple(vec![VariableOrPlaceholder::Placeholder,
VariableOrPlaceholder::Variable(Variable::from_valid_name("?y"))]),
}));
}
} }

View file

@ -210,7 +210,7 @@ fn project_elements<'a, I: IntoIterator<Item = &'a Element>>(
let columns = query.cc let columns = query.cc
.column_bindings .column_bindings
.get(var) .get(var)
.expect("Every variable has a binding"); .expect(format!("Every variable should have a binding, but {} does not", var.as_str()).as_str());
let qa = columns[0].clone(); let qa = columns[0].clone();
let name = column_name(var); let name = column_name(var);

View file

@ -20,6 +20,8 @@ use mentat_core::{
use mentat_query_algebrizer::{ use mentat_query_algebrizer::{
DatomsColumn, DatomsColumn,
FulltextColumn,
FulltextQualifiedAlias,
QualifiedAlias, QualifiedAlias,
QueryValue, QueryValue,
SourceAlias, SourceAlias,
@ -44,6 +46,7 @@ use mentat_sql::{
/// implementation for each storage backend. Passing `TypedValue`s here allows for that. /// implementation for each storage backend. Passing `TypedValue`s here allows for that.
pub enum ColumnOrExpression { pub enum ColumnOrExpression {
Column(QualifiedAlias), Column(QualifiedAlias),
FulltextColumn(FulltextQualifiedAlias),
Entid(Entid), // Because it's so common. Entid(Entid), // Because it's so common.
Integer(i32), // We use these for type codes etc. Integer(i32), // We use these for type codes etc.
Long(i64), Long(i64),
@ -55,6 +58,7 @@ impl From<QueryValue> for ColumnOrExpression {
fn from(v: QueryValue) -> Self { fn from(v: QueryValue) -> Self {
match v { match v {
QueryValue::Column(c) => ColumnOrExpression::Column(c), QueryValue::Column(c) => ColumnOrExpression::Column(c),
QueryValue::FulltextColumn(c) => ColumnOrExpression::FulltextColumn(c),
QueryValue::Entid(e) => ColumnOrExpression::Entid(e), QueryValue::Entid(e) => ColumnOrExpression::Entid(e),
QueryValue::PrimitiveLong(v) => ColumnOrExpression::Long(v), QueryValue::PrimitiveLong(v) => ColumnOrExpression::Long(v),
QueryValue::TypedValue(v) => ColumnOrExpression::Value(v), QueryValue::TypedValue(v) => ColumnOrExpression::Value(v),
@ -109,6 +113,14 @@ impl Constraint {
right: right, right: right,
} }
} }
pub fn fulltext_match(left: ColumnOrExpression, right: ColumnOrExpression) -> Constraint {
Constraint::Infix {
op: Op("MATCH"), // SQLite specific!
left: left,
right: right,
}
}
} }
#[allow(dead_code)] #[allow(dead_code)]
@ -157,6 +169,11 @@ fn push_column(qb: &mut QueryBuilder, col: &DatomsColumn) {
qb.push_sql(col.as_str()); qb.push_sql(col.as_str());
} }
// We know that FulltextColumns are safe to serialize.
fn push_fulltext_column(qb: &mut QueryBuilder, col: &FulltextColumn) {
qb.push_sql(col.as_str());
}
//--------------------------------------------------------- //---------------------------------------------------------
// Turn that representation into SQL. // Turn that representation into SQL.
@ -199,6 +216,12 @@ impl QueryFragment for ColumnOrExpression {
push_column(out, column); push_column(out, column);
Ok(()) Ok(())
}, },
&FulltextColumn(FulltextQualifiedAlias(ref table, ref column)) => {
out.push_identifier(table.as_str())?;
out.push_sql(".");
push_fulltext_column(out, column);
Ok(())
},
&Entid(entid) => { &Entid(entid) => {
out.push_sql(entid.to_string().as_str()); out.push_sql(entid.to_string().as_str());
Ok(()) Ok(())
@ -406,13 +429,20 @@ impl SelectQuery {
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::*; use super::*;
use std::rc::Rc;
use mentat_query_algebrizer::DatomsTable; use mentat_query_algebrizer::DatomsTable;
fn build_constraint(c: Constraint) -> String { fn build_constraint_query(c: Constraint) -> SQLQuery {
let mut builder = SQLiteQueryBuilder::new(); let mut builder = SQLiteQueryBuilder::new();
c.push_sql(&mut builder) c.push_sql(&mut builder)
.map(|_| builder.finish()) .map(|_| builder.finish())
.unwrap().sql .expect("to produce a query for the given constraint")
}
fn build_constraint(c: Constraint) -> String {
build_constraint_query(c).sql
} }
#[test] #[test]
@ -469,6 +499,25 @@ mod tests {
assert_eq!("((123 = 456 AND 789 = 246))", build_constraint(c)); assert_eq!("((123 = 456 AND 789 = 246))", build_constraint(c));
} }
#[test]
fn test_matches_constraint() {
let c = Constraint::Infix {
op: Op("MATCHES"),
left: ColumnOrExpression::FulltextColumn(FulltextQualifiedAlias("fulltext01".to_string(), FulltextColumn::Text)),
right: ColumnOrExpression::Value(TypedValue::String(Rc::new("needle".to_string()))),
};
let q = build_constraint_query(c);
assert_eq!("`fulltext01`.text MATCHES $v0", q.sql);
assert_eq!(vec![("$v0".to_string(), Rc::new("needle".to_string()))], q.args);
let c = Constraint::Infix {
op: Op("="),
left: ColumnOrExpression::FulltextColumn(FulltextQualifiedAlias("fulltext01".to_string(), FulltextColumn::Rowid)),
right: ColumnOrExpression::Column(QualifiedAlias("datoms02".to_string(), DatomsColumn::Value)),
};
assert_eq!("`fulltext01`.rowid = `datoms02`.v", build_constraint(c));
}
#[test] #[test]
fn test_end_to_end() { fn test_end_to_end() {
// [:find ?x :where [?x 65537 ?v] [?x 65536 ?v]] // [:find ?x :where [?x 65537 ?v] [?x 65536 ?v]]

View file

@ -32,6 +32,8 @@ use mentat_query_algebrizer::{
ConjoiningClauses, ConjoiningClauses,
DatomsColumn, DatomsColumn,
DatomsTable, DatomsTable,
FulltextColumn,
FulltextQualifiedAlias,
QualifiedAlias, QualifiedAlias,
QueryValue, QueryValue,
SourceAlias, SourceAlias,
@ -69,6 +71,12 @@ impl ToColumn for QualifiedAlias {
} }
} }
impl ToColumn for FulltextQualifiedAlias {
fn to_column(self) -> ColumnOrExpression {
ColumnOrExpression::FulltextColumn(self)
}
}
impl ToConstraint for ColumnIntersection { impl ToConstraint for ColumnIntersection {
fn to_constraint(self) -> Constraint { fn to_constraint(self) -> Constraint {
Constraint::And { Constraint::And {
@ -108,6 +116,11 @@ impl ToConstraint for ColumnConstraint {
Equals(left, QueryValue::Column(right)) => Equals(left, QueryValue::Column(right)) =>
Constraint::equal(left.to_column(), right.to_column()), Constraint::equal(left.to_column(), right.to_column()),
Equals(left, QueryValue::FulltextColumn(right)) =>
// TODO: figure out if this is the correct abstraction. Can we make it so that
// FulltextColumns::Text is not accepted here?
Constraint::equal(left.to_column(), right.to_column()),
Equals(qa, QueryValue::PrimitiveLong(value)) => { Equals(qa, QueryValue::PrimitiveLong(value)) => {
let tag_column = qa.for_type_tag().to_column(); let tag_column = qa.for_type_tag().to_column();
let value_column = qa.to_column(); let value_column = qa.to_column();
@ -148,6 +161,14 @@ impl ToConstraint for ColumnConstraint {
} }
}, },
Matches(left, right) => {
Constraint::Infix {
op: Op("MATCH"),
left: ColumnOrExpression::FulltextColumn(left),
right: right.into(),
}
},
HasType(table, value_type) => { HasType(table, value_type) => {
let column = QualifiedAlias(table, DatomsColumn::ValueTypeTag).to_column(); let column = QualifiedAlias(table, DatomsColumn::ValueTypeTag).to_column();
Constraint::equal(column, ColumnOrExpression::Integer(value_type.value_type_tag())) Constraint::equal(column, ColumnOrExpression::Integer(value_type.value_type_tag()))

View file

@ -58,6 +58,13 @@ fn prepopulated_schema() -> Schema {
value_type: ValueType::String, value_type: ValueType::String,
..Default::default() ..Default::default()
}); });
associate_ident(&mut schema, NamespacedKeyword::new("foo", "fts"), 100);
add_attribute(&mut schema, 100, Attribute {
value_type: ValueType::String,
index: true,
fulltext: true,
..Default::default()
});
schema schema
} }
@ -242,3 +249,13 @@ fn test_numeric_not_equals_known_attribute() {
assert_eq!(sql, "SELECT `datoms00`.e AS `?x` FROM `datoms` AS `datoms00` WHERE `datoms00`.a = 99 AND `datoms00`.v <> 12 LIMIT 1"); assert_eq!(sql, "SELECT `datoms00`.e AS `?x` FROM `datoms` AS `datoms00` WHERE `datoms00`.a = 99 AND `datoms00`.v <> 12 LIMIT 1");
assert_eq!(args, vec![]); assert_eq!(args, vec![]);
} }
#[test]
fn test_fulltext() {
let schema = prepopulated_schema();
let input = r#"[:find ?entity ?value ?tx ?score :where [(fulltext $ :foo/fts "needle") [?entity ?value ?tx ?score]]]"#;
let SQLQuery { sql, args } = translate(&schema, input, None);
assert_eq!(sql, "SELECT `datoms00`.e AS `?entity`, `datoms00`.v AS `?value`, `datoms00`.tx AS `?tx`, 0.0 AS `?score` FROM `datoms` AS `datoms00` WHERE `datoms00`.a = 99 AND `datoms00`.v = $v0 LIMIT 1");
assert_eq!(args, vec![make_arg("$v0", "needle")]);
}

View file

@ -147,7 +147,11 @@ impl FromValue<SrcVar> for SrcVar {
impl SrcVar { impl SrcVar {
pub fn from_symbol(sym: &PlainSymbol) -> Option<SrcVar> { pub fn from_symbol(sym: &PlainSymbol) -> Option<SrcVar> {
if sym.is_src_symbol() { if sym.is_src_symbol() {
if sym.0 == "$" {
Some(SrcVar::DefaultSrc)
} else {
Some(SrcVar::NamedSrc(sym.plain_name().to_string())) Some(SrcVar::NamedSrc(sym.plain_name().to_string()))
}
} else { } else {
None None
} }
@ -185,16 +189,34 @@ pub enum FnArg {
impl FromValue<FnArg> for FnArg { impl FromValue<FnArg> for FnArg {
fn from_value(v: edn::ValueAndSpan) -> Option<FnArg> { fn from_value(v: edn::ValueAndSpan) -> Option<FnArg> {
// TODO: support SrcVars. use edn::SpannedValue::*;
Variable::from_value(v.clone()) // TODO: don't clone!
.and_then(|v| Some(FnArg::Variable(v)))
.or_else(|| {
println!("from_value {}", v.inner);
match v.inner { match v.inner {
edn::SpannedValue::Integer(i) => Some(FnArg::EntidOrInteger(i)), Integer(x) =>
edn::SpannedValue::Float(f) => Some(FnArg::Constant(NonIntegerConstant::Float(f))), Some(FnArg::EntidOrInteger(x)),
_ => unimplemented!(), PlainSymbol(ref x) if x.is_src_symbol() =>
}}) SrcVar::from_symbol(x).map(FnArg::SrcVar),
PlainSymbol(ref x) if x.is_var_symbol() =>
Variable::from_symbol(x).map(FnArg::Variable),
PlainSymbol(_) => None,
NamespacedKeyword(ref x) =>
Some(FnArg::Ident(x.clone())),
Boolean(x) =>
Some(FnArg::Constant(NonIntegerConstant::Boolean(x))),
Float(x) =>
Some(FnArg::Constant(NonIntegerConstant::Float(x))),
BigInteger(ref x) =>
Some(FnArg::Constant(NonIntegerConstant::BigInteger(x.clone()))),
Text(ref x) =>
// TODO: intern strings. #398.
Some(FnArg::Constant(NonIntegerConstant::Text(Rc::new(x.clone())))),
Nil |
NamespacedSymbol(_) |
Keyword(_) |
Vector(_) |
List(_) |
Set(_) |
Map(_) => None,
}
} }
} }
@ -455,6 +477,25 @@ impl FindSpec {
} }
} }
// Datomic accepts variable or placeholder. DataScript accepts recursive bindings. Mentat sticks
// to the non-recursive form Datomic accepts, which is much simpler to process.
#[derive(Clone, Debug, Eq, PartialEq)]
pub enum VariableOrPlaceholder {
Placeholder,
Variable(Variable),
}
#[derive(Clone,Debug,Eq,PartialEq)]
pub enum Binding {
BindRel(Vec<VariableOrPlaceholder>),
BindColl(Variable),
BindTuple(Vec<VariableOrPlaceholder>),
BindScalar(Variable),
}
// Note that the "implicit blank" rule applies. // Note that the "implicit blank" rule applies.
// A pattern with a reversed attribute — :foo/_bar — is reversed // A pattern with a reversed attribute — :foo/_bar — is reversed
// at the point of parsing. These `Pattern` instances only represent // at the point of parsing. These `Pattern` instances only represent
@ -510,6 +551,13 @@ pub struct Predicate {
pub args: Vec<FnArg>, pub args: Vec<FnArg>,
} }
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct WhereFn {
pub operator: PlainSymbol,
pub args: Vec<FnArg>,
pub binding: Binding,
}
#[derive(Clone, Debug, Eq, PartialEq)] #[derive(Clone, Debug, Eq, PartialEq)]
pub enum UnifyVars { pub enum UnifyVars {
/// `Implicit` means the variables in an `or` or `not` are derived from the enclosed pattern. /// `Implicit` means the variables in an `or` or `not` are derived from the enclosed pattern.
@ -570,14 +618,19 @@ pub struct OrJoin {
pub clauses: Vec<OrWhereClause>, pub clauses: Vec<OrWhereClause>,
} }
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct NotJoin {
pub unify_vars: UnifyVars,
pub clauses: Vec<WhereClause>,
}
#[allow(dead_code)] #[allow(dead_code)]
#[derive(Clone, Debug, Eq, PartialEq)] #[derive(Clone, Debug, Eq, PartialEq)]
pub enum WhereClause { pub enum WhereClause {
Not, NotJoin(NotJoin),
NotJoin,
OrJoin(OrJoin), OrJoin(OrJoin),
Pred(Predicate), Pred(Predicate),
WhereFn, WhereFn(WhereFn),
RuleExpr, RuleExpr,
Pattern(Pattern), Pattern(Pattern),
} }
@ -628,9 +681,8 @@ impl ContainsVariables for WhereClause {
&OrJoin(ref o) => o.accumulate_mentioned_variables(acc), &OrJoin(ref o) => o.accumulate_mentioned_variables(acc),
&Pred(ref p) => p.accumulate_mentioned_variables(acc), &Pred(ref p) => p.accumulate_mentioned_variables(acc),
&Pattern(ref p) => p.accumulate_mentioned_variables(acc), &Pattern(ref p) => p.accumulate_mentioned_variables(acc),
&Not => (), &NotJoin(ref n) => n.accumulate_mentioned_variables(acc),
&NotJoin => (), &WhereFn(_) => (),
&WhereFn => (),
&RuleExpr => (), &RuleExpr => (),
} }
} }
@ -654,6 +706,14 @@ impl ContainsVariables for OrJoin {
} }
} }
impl ContainsVariables for NotJoin {
fn accumulate_mentioned_variables(&self, acc: &mut BTreeSet<Variable>) {
for clause in &self.clauses {
clause.accumulate_mentioned_variables(acc);
}
}
}
impl ContainsVariables for Predicate { impl ContainsVariables for Predicate {
fn accumulate_mentioned_variables(&self, acc: &mut BTreeSet<Variable>) { fn accumulate_mentioned_variables(&self, acc: &mut BTreeSet<Variable>) {
for arg in &self.args { for arg in &self.args {