Merge pull request #592 from MarinPostma/query-filters

Implements query filters

CHANGELOG.md

@@ -1,5 +1,6 @@
 ## 0.9.1 (unreleased)
 
+  - Refined filtering (#592)
   - Add the number of hits in search result (#541)
   - Add support for aligned crop in search result (#543)
   - Sanitize the content displayed in the web interface (#539)

meilisearch-core/Cargo.toml

@@ -27,12 +27,15 @@ meilisearch-tokenizer = { path = "../meilisearch-tokenizer", version = "0.9.0" }
 meilisearch-types = { path = "../meilisearch-types", version = "0.9.0" }
 once_cell = "1.3.1"
 ordered-float = { version = "1.0.2", features = ["serde"] }
+pest = { git = "https://github.com/MarinPostma/pest.git", tag = "meilisearch-patch1" } 
+pest_derive = "2.0"
 regex = "1.3.6"
 sdset = "0.4.0"
 serde = { version = "1.0.105", features = ["derive"] }
 serde_json = "1.0.50"
 siphasher = "0.3.2"
 slice-group-by = "0.2.6"
+unicase = "2.6.0"
 zerocopy = "0.3.0"
 
 [dev-dependencies]

meilisearch-core/src/error.rs

@@ -1,10 +1,13 @@
 use crate::serde::{DeserializerError, SerializerError};
 use serde_json::Error as SerdeJsonError;
+use pest::error::Error as PestError;
+use crate::filters::Rule;
 use std::{error, fmt, io};
 
-pub use heed::Error as HeedError;
-pub use fst::Error as FstError;
 pub use bincode::Error as BincodeError;
+pub use fst::Error as FstError;
+pub use heed::Error as HeedError;
+pub use pest::error as pest_error;
 
 pub type MResult<T> = Result<T, Error>;
 
@@ -25,6 +28,7 @@ pub enum Error {
     Serializer(SerializerError),
     Deserializer(DeserializerError),
     UnsupportedOperation(UnsupportedOperation),
+    FilterParseError(PestError<Rule>)
 }
 
 impl From<io::Error> for Error {
@@ -33,6 +37,28 @@ impl From<io::Error> for Error {
     }
 }
 
+impl From<PestError<Rule>> for Error {
+    fn from(error: PestError<Rule>) -> Error {
+        Error::FilterParseError(error.renamed_rules(|r| {
+            let s  = match r {
+                Rule::or => "OR",
+                Rule::and => "AND",
+                Rule::not => "NOT",
+                Rule::string => "string",
+                Rule::word => "word",
+                Rule::greater => "field > value",
+                Rule::less => "field < value",
+                Rule::eq => "field = value",
+                Rule::leq => "field <= value",
+                Rule::geq => "field >= value",
+                Rule::key => "key",
+                _ => "other",
+            };
+            s.to_string()
+        }))
+    }
+} 
+
 impl From<meilisearch_schema::Error> for Error {
     fn from(error: meilisearch_schema::Error) -> Error {
         Error::Schema(error)
@@ -100,6 +126,7 @@ impl fmt::Display for Error {
             Serializer(e) => write!(f, "serializer error; {}", e),
             Deserializer(e) => write!(f, "deserializer error; {}", e),
             UnsupportedOperation(op) => write!(f, "unsupported operation; {}", op),
+            FilterParseError(e) => write!(f, "error parsing filter; {}", e),
         }
     }
 }

meilisearch-core/src/filters/condition.rs

@@ -0,0 +1,267 @@
+use std::str::FromStr;
+use std::cmp::Ordering;
+
+use crate::error::Error;
+use crate::{store::Index, DocumentId, MainT};
+use heed::RoTxn;
+use meilisearch_schema::{FieldId, Schema};
+use pest::error::{Error as PestError, ErrorVariant};
+use pest::iterators::Pair;
+use serde_json::{Value, Number};
+use super::parser::Rule;
+
+#[derive(Debug)]
+enum ConditionType {
+    Greater,
+    Less,
+    Equal,
+    LessEqual,
+    GreaterEqual,
+    NotEqual,
+}
+
+/// We need to infer type when the filter is constructed
+/// and match every possible types it can be parsed into.
+#[derive(Debug)]
+struct ConditionValue<'a> {
+    string: &'a str,
+    boolean: Option<bool>,
+    number: Option<Number>
+}
+
+impl<'a> ConditionValue<'a> {
+    pub fn new(value: &Pair<'a, Rule>) -> Self {
+        let value = match value.as_rule() {
+            Rule::string | Rule::word => {
+                let string =  value.as_str();
+                let boolean = match value.as_str() {
+                    "true" => Some(true),
+                    "false" => Some(false),
+                    _ => None,
+                };
+                let number = Number::from_str(value.as_str()).ok();
+                ConditionValue { string, boolean, number }
+            },
+            _ => unreachable!(),
+        };
+        value
+    }
+
+    pub fn as_str(&self) -> &str {
+        self.string.as_ref()
+    }
+
+    pub fn as_number(&self) -> Option<&Number> {
+        self.number.as_ref()
+    }
+
+    pub fn as_bool(&self) -> Option<bool> {
+        self.boolean
+    }
+}
+
+#[derive(Debug)]
+pub struct Condition<'a> {
+    field: FieldId,
+    condition: ConditionType,
+    value: ConditionValue<'a>
+}
+
+fn get_field_value<'a>(schema: &Schema, pair: Pair<'a, Rule>) -> Result<(FieldId, ConditionValue<'a>), Error> {
+    let mut items = pair.into_inner();
+    // lexing ensures that we at least have a key
+    let key = items.next().unwrap();
+    let field = schema
+        .id(key.as_str())
+        .ok_or::<PestError<Rule>>(PestError::new_from_span(
+                ErrorVariant::CustomError {
+                    message: format!(
+                                 "attribute `{}` not found, available attributes are: {}",
+                                 key.as_str(),
+                                 schema.names().collect::<Vec<_>>().join(", ")
+                             ),
+                },
+                key.as_span()))?;
+    let value = ConditionValue::new(&items.next().unwrap());
+    Ok((field, value))
+}
+
+// undefined behavior with big numbers
+fn compare_numbers(lhs: &Number, rhs: &Number) -> Option<Ordering> {
+    match (lhs.as_i64(), lhs.as_u64(), lhs.as_f64(),
+        rhs.as_i64(), rhs.as_u64(), rhs.as_f64()) {
+    //    i64   u64  f64  i64  u64  f64
+        (Some(lhs), _, _, Some(rhs), _, _) => lhs.partial_cmp(&rhs),
+        (_, Some(lhs), _, _, Some(rhs), _) => lhs.partial_cmp(&rhs),
+        (_, _, Some(lhs), _, _, Some(rhs)) => lhs.partial_cmp(&rhs),
+        (_, _, _, _, _, _) => None,
+    }
+}
+
+impl<'a> Condition<'a> {
+    pub fn less(
+        item: Pair<'a, Rule>,
+        schema: &'a Schema,
+    ) -> Result<Self, Error> {
+        let (field, value) = get_field_value(schema, item)?;
+        let condition = ConditionType::Less;
+        Ok(Self { field, condition, value })
+    }
+
+    pub fn greater(
+        item: Pair<'a, Rule>,
+        schema: &'a Schema,
+    ) -> Result<Self, Error> {
+        let (field, value) = get_field_value(schema, item)?;
+        let condition = ConditionType::Greater;
+        Ok(Self { field, condition, value })
+    }
+
+    pub fn neq(
+        item: Pair<'a, Rule>,
+        schema: &'a Schema,
+    ) -> Result<Self, Error> {
+        let (field, value) = get_field_value(schema, item)?;
+        let condition = ConditionType::NotEqual;
+        Ok(Self { field, condition, value })
+    }
+
+    pub fn geq(
+        item: Pair<'a, Rule>,
+        schema: &'a Schema,
+    ) -> Result<Self, Error> {
+        let (field, value) = get_field_value(schema, item)?;
+        let condition = ConditionType::GreaterEqual;
+        Ok(Self { field, condition, value })
+    }
+
+    pub fn leq(
+        item: Pair<'a, Rule>,
+        schema: &'a Schema,
+    ) -> Result<Self, Error> {
+        let (field, value) = get_field_value(schema, item)?;
+        let condition = ConditionType::LessEqual;
+        Ok(Self { field, condition, value })
+    }
+
+    pub fn eq(
+        item: Pair<'a, Rule>,
+        schema: &'a Schema,
+    ) -> Result<Self, Error> {
+        let (field, value) = get_field_value(schema, item)?;
+        let condition = ConditionType::Equal;
+        Ok(Self { field, condition, value })
+    }
+
+    pub fn test(
+        &self,
+        reader: &RoTxn<MainT>,
+        index: &Index,
+        document_id: DocumentId,
+    ) -> Result<bool, Error> {
+        match index.document_attribute::<Value>(reader, document_id, self.field)? {
+            Some(Value::String(s)) => {
+                let value = self.value.as_str();
+                match self.condition {
+                    ConditionType::Equal => Ok(unicase::eq(value, &s)),
+                    ConditionType::NotEqual => Ok(!unicase::eq(value, &s)),
+                    _ => Ok(false)
+                }
+            },
+            Some(Value::Number(n)) => { 
+                if let Some(value) = self.value.as_number() {
+                    if let Some(ord) = compare_numbers(&n, value) {
+                        let res =  match self.condition {
+                            ConditionType::Equal => ord == Ordering::Equal,
+                            ConditionType::NotEqual => ord != Ordering::Equal,
+                            ConditionType::GreaterEqual => ord != Ordering::Less,
+                            ConditionType::LessEqual => ord != Ordering::Greater,
+                            ConditionType::Greater => ord == Ordering::Greater,
+                            ConditionType::Less => ord == Ordering::Less,
+                        };
+                        return Ok(res)
+                    } 
+                } 
+                Ok(false)
+            },
+            Some(Value::Bool(b)) => {
+                if let Some(value) = self.value.as_bool() {
+                    return match self.condition {
+                        ConditionType::Equal => Ok(b == value),
+                        ConditionType::NotEqual => Ok(b != value),
+                        _ => Ok(false)
+                    }
+                }
+                Ok(false)
+            },
+            _ => Ok(false),
+        }
+    }
+}
+
+#[cfg(test)]
+mod test {
+    use super::*;
+    use serde_json::Number;
+    use std::cmp::Ordering;
+
+    #[test]
+    fn test_number_comp() {
+        // test both u64
+        let n1 = Number::from(1u64);
+        let n2 = Number::from(2u64);
+        assert_eq!(Some(Ordering::Less), compare_numbers(&n1, &n2));
+        assert_eq!(Some(Ordering::Greater), compare_numbers(&n2, &n1));
+        let n1 = Number::from(1u64);
+        let n2 = Number::from(1u64);
+        assert_eq!(Some(Ordering::Equal), compare_numbers(&n1, &n2));
+
+        // test both i64
+        let n1 = Number::from(1i64);
+        let n2 = Number::from(2i64);
+        assert_eq!(Some(Ordering::Less), compare_numbers(&n1, &n2));
+        assert_eq!(Some(Ordering::Greater), compare_numbers(&n2, &n1));
+        let n1 = Number::from(1i64);
+        let n2 = Number::from(1i64);
+        assert_eq!(Some(Ordering::Equal), compare_numbers(&n1, &n2));
+
+        // test both f64
+        let n1 = Number::from_f64(1f64).unwrap();
+        let n2 = Number::from_f64(2f64).unwrap();
+        assert_eq!(Some(Ordering::Less), compare_numbers(&n1, &n2));
+        assert_eq!(Some(Ordering::Greater), compare_numbers(&n2, &n1));
+        let n1 = Number::from_f64(1f64).unwrap();
+        let n2 = Number::from_f64(1f64).unwrap();
+        assert_eq!(Some(Ordering::Equal), compare_numbers(&n1, &n2));
+
+        // test one u64 and one f64
+        let n1 = Number::from_f64(1f64).unwrap();
+        let n2 = Number::from(2u64);
+        assert_eq!(Some(Ordering::Less), compare_numbers(&n1, &n2));
+        assert_eq!(Some(Ordering::Greater), compare_numbers(&n2, &n1));
+
+        // equality
+        let n1 = Number::from_f64(1f64).unwrap();
+        let n2 = Number::from(1u64);
+        assert_eq!(Some(Ordering::Equal), compare_numbers(&n1, &n2));
+        assert_eq!(Some(Ordering::Equal), compare_numbers(&n2, &n1));
+
+        // float is neg
+        let n1 = Number::from_f64(-1f64).unwrap();
+        let n2 = Number::from(1u64);
+        assert_eq!(Some(Ordering::Less), compare_numbers(&n1, &n2));
+        assert_eq!(Some(Ordering::Greater), compare_numbers(&n2, &n1));
+
+        // float is too big
+        let n1 = Number::from_f64(std::f64::MAX).unwrap();
+        let n2 = Number::from(1u64);
+        assert_eq!(Some(Ordering::Greater), compare_numbers(&n1, &n2));
+        assert_eq!(Some(Ordering::Less), compare_numbers(&n2, &n1));
+
+        // misc
+        let n1 = Number::from_f64(std::f64::MAX).unwrap();
+        let n2 = Number::from(std::u64::MAX);
+        assert_eq!(Some(Ordering::Greater), compare_numbers(&n1, &n2));
+        assert_eq!(Some( Ordering::Less ), compare_numbers(&n2, &n1));
+    }
+}

meilisearch-core/src/filters/mod.rs

@@ -0,0 +1,125 @@
+mod parser;
+mod condition;
+
+pub(crate) use parser::Rule;
+
+use std::ops::Not;
+
+use condition::Condition;
+use crate::error::Error;
+use crate::{DocumentId, MainT, store::Index};
+use heed::RoTxn;
+use meilisearch_schema::Schema;
+use parser::{PREC_CLIMBER, FilterParser};
+use pest::iterators::{Pair, Pairs};
+use pest::Parser;
+
+type FilterResult<'a> = Result<Filter<'a>, Error>;
+
+#[derive(Debug)]
+pub enum Filter<'a> {
+    Condition(Condition<'a>),
+    Or(Box<Self>, Box<Self>),
+    And(Box<Self>, Box<Self>),
+    Not(Box<Self>),
+}
+
+impl<'a> Filter<'a> {
+    pub fn parse(expr: &'a str, schema: &'a Schema) -> FilterResult<'a> {
+        let mut lexed = FilterParser::parse(Rule::prgm, expr.as_ref())?;
+        Self::build(lexed.next().unwrap().into_inner(), schema)
+    }
+
+    pub fn test(
+        &self,
+        reader: &RoTxn<MainT>,
+        index: &Index,
+        document_id: DocumentId,
+    ) -> Result<bool, Error> {
+        use Filter::*;
+        match self {
+            Condition(c) => c.test(reader, index, document_id),
+            Or(lhs, rhs) => Ok(
+                lhs.test(reader, index, document_id)? || rhs.test(reader, index, document_id)?
+            ),
+            And(lhs, rhs) => Ok(
+                lhs.test(reader, index, document_id)? && rhs.test(reader, index, document_id)?
+            ),
+            Not(op) => op.test(reader, index, document_id).map(bool::not),
+        }
+    }
+
+    fn build(expression: Pairs<'a, Rule>, schema: &'a Schema) -> FilterResult<'a> {
+        PREC_CLIMBER.climb(
+            expression,
+            |pair: Pair<Rule>| match pair.as_rule() {
+                Rule::eq => Ok(Filter::Condition(Condition::eq(pair, schema)?)),
+                Rule::greater => Ok(Filter::Condition(Condition::greater(pair, schema)?)),
+                Rule::less => Ok(Filter::Condition(Condition::less(pair, schema)?)),
+                Rule::neq => Ok(Filter::Condition(Condition::neq(pair, schema)?)),
+                Rule::geq => Ok(Filter::Condition(Condition::geq(pair, schema)?)),
+                Rule::leq => Ok(Filter::Condition(Condition::leq(pair, schema)?)),
+                Rule::prgm => Self::build(pair.into_inner(), schema),
+                Rule::not => Ok(Filter::Not(Box::new(Self::build(
+                    pair.into_inner(),
+                    schema,
+                )?))),
+                _ => unreachable!(),
+            },
+            |lhs: FilterResult, op: Pair<Rule>, rhs: FilterResult| match op.as_rule() {
+                Rule::or => Ok(Filter::Or(Box::new(lhs?), Box::new(rhs?))),
+                Rule::and => Ok(Filter::And(Box::new(lhs?), Box::new(rhs?))),
+                _ => unreachable!(),
+            },
+        )
+    }
+}
+
+#[cfg(test)]
+mod test {
+    use super::*;
+
+    #[test]
+    fn invalid_syntax() {
+        assert!(FilterParser::parse(Rule::prgm, "field : id").is_err());
+        assert!(FilterParser::parse(Rule::prgm, "field=hello hello").is_err());
+        assert!(FilterParser::parse(Rule::prgm, "field=hello OR OR").is_err());
+        assert!(FilterParser::parse(Rule::prgm, "OR field:hello").is_err());
+        assert!(FilterParser::parse(Rule::prgm, r#"field="hello world"#).is_err());
+        assert!(FilterParser::parse(Rule::prgm, r#"field='hello world"#).is_err());
+        assert!(FilterParser::parse(Rule::prgm, "NOT field=").is_err());
+        assert!(FilterParser::parse(Rule::prgm, "N").is_err());
+        assert!(FilterParser::parse(Rule::prgm, "(field=1").is_err());
+        assert!(FilterParser::parse(Rule::prgm, "(field=1))").is_err());
+        assert!(FilterParser::parse(Rule::prgm, "field=1ORfield=2").is_err());
+        assert!(FilterParser::parse(Rule::prgm, "field=1 ( OR field=2)").is_err());
+        assert!(FilterParser::parse(Rule::prgm, "hello world=1").is_err());
+        assert!(FilterParser::parse(Rule::prgm, "").is_err());
+        assert!(FilterParser::parse(Rule::prgm, r#"((((((hello=world)))))"#).is_err());
+    }
+
+    #[test]
+    fn valid_syntax() {
+        assert!(FilterParser::parse(Rule::prgm, "field = id").is_ok());
+        assert!(FilterParser::parse(Rule::prgm, "field=id").is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"field >= 10"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"field <= 10"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"field="hello world""#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"field='hello world'"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"field > 10"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"field < 10"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"field < 10 AND NOT field=5"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"field=true OR NOT field=5"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"NOT field=true OR NOT field=5"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"field='hello world' OR ( NOT field=true OR NOT field=5 )"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"field='hello \'worl\'d' OR ( NOT field=true OR NOT field=5 )"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"field="hello \"worl\"d" OR ( NOT field=true OR NOT field=5 )"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"((((((hello=world))))))"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#""foo bar" > 10"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#""foo bar" = 10"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"'foo bar' = 10"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"'foo bar' <= 10"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"'foo bar' != 10"#).is_ok());
+        assert!(FilterParser::parse(Rule::prgm, r#"bar != 10"#).is_ok());
+    }
+}

meilisearch-core/src/filters/parser/grammar.pest

@@ -0,0 +1,28 @@
+key = _{quoted | word}
+value = _{quoted | word}
+quoted = _{ (PUSH("'") | PUSH("\"")) ~ string ~ POP  }
+string = {char*}
+word = ${(LETTER | NUMBER | "_" | "-")+}
+
+char =  _{ !(PEEK | "\\") ~ ANY
+    | "\\" ~ (PEEK | "\\" | "/" | "b" | "f" | "n" | "r" | "t")
+    | "\\" ~ ("u" ~ ASCII_HEX_DIGIT{4})}
+
+condition = _{eq | greater | less | geq | leq | neq}
+geq = {key ~ ">=" ~ value}
+leq = {key ~ "<=" ~ value}
+neq = {key ~ "!=" ~ value}
+eq = {key ~ "=" ~ value}
+greater = {key ~ ">" ~ value}
+less = {key ~ "<" ~ value}
+
+prgm = {SOI ~ expr ~ EOI}
+expr = _{ ( term ~ (operation ~ term)* ) }
+term = _{ ("(" ~ expr ~ ")") | condition | not }
+operation = _{ and | or }
+	and = {"AND"}
+	or = {"OR"}
+
+not = {"NOT" ~ term}
+
+WHITESPACE = _{ " " }

meilisearch-core/src/filters/parser/mod.rs

@@ -0,0 +1,12 @@
+use once_cell::sync::Lazy;
+use pest::prec_climber::{Operator, Assoc, PrecClimber};
+
+pub static PREC_CLIMBER: Lazy<PrecClimber<Rule>> = Lazy::new(|| {
+    use Assoc::*;
+    use Rule::*;
+    pest::prec_climber::PrecClimber::new(vec![Operator::new(or, Left), Operator::new(and, Left)])
+});
+
+#[derive(Parser)]
+#[grammar = "filters/parser/grammar.pest"]
+pub struct FilterParser;

meilisearch-core/src/lib.rs

@@ -1,12 +1,15 @@
 #[cfg(test)]
 #[macro_use]
 extern crate assert_matches;
+#[macro_use]
+extern crate pest_derive;
 
 mod automaton;
 mod bucket_sort;
 mod database;
 mod distinct_map;
 mod error;
+mod filters;
 mod levenshtein;
 mod number;
 mod query_builder;
@@ -23,7 +26,8 @@ pub mod serde;
 pub mod store;
 
 pub use self::database::{BoxUpdateFn, Database, MainT, UpdateT};
-pub use self::error::{Error, HeedError, FstError, MResult};
+pub use self::error::{Error, HeedError, FstError, MResult, pest_error};
+pub use self::filters::Filter;
 pub use self::number::{Number, ParseNumberError};
 pub use self::ranked_map::RankedMap;
 pub use self::raw_document::RawDocument;

meilisearch-http/src/error.rs

@@ -19,6 +19,7 @@ pub enum ResponseError {
     IndexNotFound(String),
     DocumentNotFound(String),
     MissingHeader(String),
+    FilterParsing(String),
     BadParameter(String, String),
     OpenIndex(String),
     CreateIndex(String),
@@ -73,11 +74,15 @@ impl IntoResponse for ResponseError {
         match self {
             ResponseError::Internal(err) => {
                 error!("internal server error: {}", err);
-                error(
-                    String::from("Internal server error"),
+                error("Internal server error".to_string(),
                     StatusCode::INTERNAL_SERVER_ERROR,
                 )
             }
+            ResponseError::FilterParsing(err) => {
+                warn!("error paring filter: {}", err);
+                error(format!("parsing error: {}", err),
+                StatusCode::BAD_REQUEST)
+            }
             ResponseError::BadRequest(err) => {
                 warn!("bad request: {}", err);
                 error(err, StatusCode::BAD_REQUEST)
@@ -159,7 +164,10 @@ impl From<FstError> for ResponseError {
 
 impl From<SearchError> for ResponseError {
     fn from(err: SearchError) -> ResponseError {
-        ResponseError::internal(err)
+        match err {
+            SearchError::FilterParsing(s) => ResponseError::FilterParsing(s),
+            _ => ResponseError::internal(err),
+        }
     }
 }
 

meilisearch-http/src/helpers/meilisearch.rs

@@ -8,6 +8,7 @@ use std::time::{Duration, Instant};
 
 use indexmap::IndexMap;
 use log::error;
+use meilisearch_core::Filter;
 use meilisearch_core::criterion::*;
 use meilisearch_core::settings::RankingRule;
 use meilisearch_core::{Highlight, Index, MainT, RankedMap};
@@ -23,6 +24,7 @@ pub enum Error {
     RetrieveDocument(u64, String),
     DocumentNotFound(u64),
     CropFieldWrongType(String),
+    FilterParsing(String),
     AttributeNotFoundOnDocument(String),
     AttributeNotFoundOnSchema(String),
     MissingFilterValue,
@@ -56,13 +58,26 @@ impl fmt::Display for Error {
                 f.write_str("a filter is specifying an unknown schema attribute")
             }
             Internal(err) => write!(f, "internal error; {}", err),
+            FilterParsing(err) => write!(f, "filter parsing error: {}", err),
         }
     }
 }
 
 impl From<meilisearch_core::Error> for Error {
     fn from(error: meilisearch_core::Error) -> Self {
-        Error::Internal(error.to_string())
+        use meilisearch_core::pest_error::LineColLocation::*;
+        match error {
+            meilisearch_core::Error::FilterParseError(e) => {
+                let (line, column) = match e.line_col {
+                    Span((line, _), (column, _)) => (line, column),
+                    Pos((line, column)) => (line, column),
+                };
+                let message = format!("parsing error on line {} at column {}: {}", line, column, e.variant.message());
+
+                Error::FilterParsing(message) 
+            },
+            _ => Error::Internal(error.to_string()),
+        }
     }
 }
 
@@ -171,40 +186,21 @@ impl<'a> SearchBuilder<'a> {
             None => self.index.query_builder(),
         };
 
-        if let Some(filters) = &self.filters {
-            let mut split = filters.split(':');
-            match (split.next(), split.next()) {
-                (Some(_), None) | (Some(_), Some("")) => return Err(Error::MissingFilterValue),
-                (Some(attr), Some(value)) => {
-                    let ref_reader = reader;
-                    let ref_index = &self.index;
-                    let value = value.trim().to_lowercase();
-
-                    let attr = match schema.id(attr) {
-                        Some(attr) => attr,
-                        None => return Err(Error::UnknownFilteredAttribute),
-                    };
-
+        if let Some(filter_expression) = &self.filters {
+            let filter = Filter::parse(filter_expression, &schema)?;
             query_builder.with_filter(move |id| {
-                        let attr = attr;
-                        let index = ref_index;
-                        let reader = ref_reader;
-
-                        match index.document_attribute::<Value>(reader, id, attr) {
-                            Ok(Some(Value::String(s))) => s.to_lowercase() == value,
-                            Ok(Some(Value::Bool(b))) => {
-                                (value == "true" && b) || (value == "false" && !b)
+                let index = &self.index;
+                let reader = &reader;
+                let filter = &filter;
+                match filter.test(reader, index, id) {
+                    Ok(res) => res,
+                    Err(e) => {
+                        log::warn!("unexpected error during filtering: {}", e);
+                        false
                     }
-                            Ok(Some(Value::Array(a))) => {
-                                a.into_iter().any(|s| s.as_str() == Some(&value))
-                            }
-                            _ => false,
                 }
             });
         }
-                (_, _) => (),
-            }
-        }
 
         query_builder.with_fetch_timeout(self.timeout);
 

meilisearch-http/tests/search.rs

@@ -438,10 +438,10 @@ fn basic_search() {
 
     // 1 - Simple search with filter
     // q: Captain
-    // limit: 1
+    // limit: 3
     // filters: director:Anthony%20Russo
 
-    let query = "q=captain&limit=3&filters=director:Anthony%20Russo";
+    let query = "q=captain&filters=director%20%3D%20%22Anthony%20Russo%22&limit=3";
 
     let expect = json!([
       {

meilisearch-schema/src/schema.rs

@@ -79,6 +79,10 @@ impl Schema {
         self.fields_map.name(id)
     }
 
+    pub fn names(&self) -> impl Iterator<Item = &str> {
+        self.fields_map.iter().map(|(k, _)| k.as_ref())
+    }
+
     pub fn contains(&self, name: &str) -> bool {
         self.fields_map.id(name).is_some()
     }