feat: Allow regex operators

Uses the new custom operators feature from `sqlparser-rs` apache/datafusion-sqlparser-rs#868

Author: max-sixty

CHANGELOG.md

@@ -22,7 +22,9 @@
     REGEXP(name, 'Love')
   ```
 
-  ...though the exact form differs by dialect.
+  ...though the exact form differs by dialect; see the
+  [Regex docs](https://prql-lang.org/book/language-features/regex.html) for more
+  details.
 
 **Fixes**:
 

prql-compiler/src/sql/dialect.rs

@@ -162,49 +162,50 @@ pub(super) trait DialectHandler: Any + Debug {
         true
     }
 
-    fn regex_function(&self) -> Option<&'static str> {
-        Some("REGEXP")
-    }
-
     fn translate_regex(
         &self,
         search: sql_ast::Expr,
         target: sql_ast::Expr,
     ) -> anyhow::Result<sql_ast::Expr> {
-        // When
-        // https://github.com/sqlparser-rs/sqlparser-rs/issues/863#issuecomment-1537272570
-        // is fixed, we can implement the infix for the other dialects. Until
-        // then, we still only have the `regex_function` implementations. (But
-        // I'd done the work to allow any Expr to be created before realizing
-        // sqlparser-rs didn't support custom operators, so may as well leave it
-        // in for when they do / we find another approach.)
-
-        let Some(regex_function) = self.regex_function() else {
-            // TODO: name the dialect, but not immediately obvious how to actually
-            // get the dialect string from a `DialectHandler` (though we could
-            // add it to each impl...)
-            //
-            // MSSQL doesn't support them, MySQL & SQLite have a different construction.
-            return Err(Error::new(
-                crate::Reason::Simple("regex functions are not supported by this dialect (or PRQL doesn't yet implement this dialect)".to_string())
-            ).into())
-        };
+        self.translate_regex_with_function(search, target, "REGEXP")
+    }
 
+    fn translate_regex_with_function(
+        // This `self` isn't actually used, but it's require because of object
+        // safety (open to better ways of doing this...)
+        &self,
+        search: sql_ast::Expr,
+        target: sql_ast::Expr,
+        function_name: &str,
+    ) -> anyhow::Result<sql_ast::Expr> {
         let args = [search, target]
             .into_iter()
             .map(FunctionArgExpr::Expr)
             .map(FunctionArg::Unnamed)
             .collect();
 
         Ok(sql_ast::Expr::Function(Function {
-            name: ObjectName(vec![sql_ast::Ident::new(regex_function)]),
+            name: ObjectName(vec![sql_ast::Ident::new(function_name)]),
             args,
             over: None,
             distinct: false,
             special: false,
             order_by: vec![],
         }))
     }
+
+    fn translate_regex_with_operator(
+        &self,
+        search: sql_ast::Expr,
+        target: sql_ast::Expr,
+        operator: sql_ast::BinaryOperator,
+    ) -> anyhow::Result<sql_ast::Expr> {
+        Ok(sql_ast::Expr::BinaryOp {
+            left: Box::new(search),
+            op: operator,
+            right: Box::new(target),
+        })
+    }
 }
 
 impl dyn DialectHandler {
@@ -220,8 +221,12 @@ impl DialectHandler for PostgresDialect {
     fn requires_quotes_intervals(&self) -> bool {
         true
     }
-    fn regex_function(&self) -> std::option::Option<&'static str> {
-        Some("REGEXP_LIKE")
+    fn translate_regex(
+        &self,
+        search: sql_ast::Expr,
+        target: sql_ast::Expr,
+    ) -> anyhow::Result<sql_ast::Expr> {
+        self.translate_regex_with_operator(search, target, sql_ast::BinaryOperator::PGRegexMatch)
     }
 }
 
@@ -242,21 +247,32 @@ impl DialectHandler for SQLiteDialect {
         false
     }
 
-    fn regex_function(&self) -> Option<&'static str> {
-        // Sqlite has a different construction, using `REGEXP` as an operator.
-        // (`foo REGEXP 'bar').
-        //
-        // TODO: change the construction of the function to allow this.
-        None
+    fn translate_regex(
+        &self,
+        search: sql_ast::Expr,
+        target: sql_ast::Expr,
+    ) -> anyhow::Result<sql_ast::Expr> {
+        self.translate_regex_with_operator(
+            search,
+            target,
+            sql_ast::BinaryOperator::Custom("REGEXP".to_string()),
+        )
     }
 }
 
 impl DialectHandler for MsSqlDialect {
     fn use_top(&self) -> bool {
         true
     }
-    fn regex_function(&self) -> Option<&'static str> {
-        None
+
+    fn translate_regex(
+        &self,
+        _search: sql_ast::Expr,
+        _target: sql_ast::Expr,
+    ) -> anyhow::Result<sql_ast::Expr> {
+        Err(Error::new(crate::Reason::Simple(
+            "regex functions are not supported by MsSql".to_string(),
+        )))?
     }
 
     // https://learn.microsoft.com/en-us/sql/t-sql/language-elements/set-operators-except-and-intersect-transact-sql?view=sql-server-ver16
@@ -279,12 +295,16 @@ impl DialectHandler for MySqlDialect {
         true
     }
 
-    fn regex_function(&self) -> Option<&'static str> {
-        // MySQL has a different construction, using `REGEXP` as an operator.
-        // (`foo REGEXP 'bar'). So
-        //
-        // TODO: change the construction of the function to allow this.
-        None
+    fn translate_regex(
+        &self,
+        search: sql_ast::Expr,
+        target: sql_ast::Expr,
+    ) -> anyhow::Result<sql_ast::Expr> {
+        self.translate_regex_with_operator(
+            search,
+            target,
+            sql_ast::BinaryOperator::Custom("REGEXP".to_string()),
+        )
     }
 }
 
@@ -311,8 +331,12 @@ impl DialectHandler for BigQueryDialect {
         true
     }
 
-    fn regex_function(&self) -> Option<&'static str> {
-        Some("REGEXP_CONTAINS")
+    fn translate_regex(
+        &self,
+        search: sql_ast::Expr,
+        target: sql_ast::Expr,
+    ) -> anyhow::Result<sql_ast::Expr> {
+        self.translate_regex_with_function(search, target, "REGEXP_CONTAINS")
     }
 }
 
@@ -339,8 +363,12 @@ impl DialectHandler for DuckDbDialect {
         false
     }
 
-    fn regex_function(&self) -> Option<&'static str> {
-        Some("REGEXP_MATCHES")
+    fn translate_regex(
+        &self,
+        search: sql_ast::Expr,
+        target: sql_ast::Expr,
+    ) -> anyhow::Result<sql_ast::Expr> {
+        self.translate_regex_with_function(search, target, "REGEXP_MATCHES")
     }
 }
 

prql-compiler/src/tests/test_error_messages.rs

@@ -130,7 +130,7 @@ fn test_hint_missing_args() {
 #[test]
 fn test_regex_dialect() {
     assert_display_snapshot!(compile(r###"
-    prql target:sql.sqlite
+    prql target:sql.mssql
     from foo
     filter bar ~= 'love'
     "###).unwrap_err(), @r###"
@@ -139,7 +139,7 @@ fn test_regex_dialect() {
        │
      4 │     filter bar ~= 'love'
        │            ──────┬──────
-       │                  ╰──────── regex functions are not supported by this dialect (or PRQL doesn't yet implement this dialect)
+       │                  ╰──────── regex functions are not supported by MsSql
     ───╯
     "###)
 }

web/book/src/language-features/regex.md

@@ -34,3 +34,17 @@ prql target:sql.postgres
 from tracks
 filter (name ~= "\\(I Can't Help\\) Falling")
 ```
+
+```prql no-fmt
+prql target:sql.mysql
+
+from tracks
+filter (name ~= "With You")
+```
+
+```prql no-fmt
+prql target:sql.sqlite
+
+from tracks
+filter (name ~= "But Why Isn't Your Syntax More Similar\\?")
+```

web/book/tests/snapshots/snapshot__language-features__regex-3.snap

@@ -7,5 +7,5 @@ SELECT
 FROM
   tracks
 WHERE
-  REGEXP_LIKE(name, '\(I Can''t Help\) Falling')
+  name ~ '\(I Can''t Help\) Falling'
 

web/book/tests/snapshots/snapshot__language-features__regex-4.snap

@@ -0,0 +1,11 @@
+---
+source: web/book/tests/snapshot.rs
+expression: "prql target:sql.mysql\n\nfrom tracks\nfilter (name ~= \"With You\")\n"
+---
+SELECT
+  *
+FROM
+  tracks
+WHERE
+  name REGEXP 'With You'
+

web/book/tests/snapshots/snapshot__language-features__regex-5.snap

@@ -0,0 +1,11 @@
+---
+source: web/book/tests/snapshot.rs
+expression: "prql target:sql.sqlite\n\nfrom tracks\nfilter (name ~= \"But Why Isn't Your Syntax More Similar\\\\?\")\n"
+---
+SELECT
+  *
+FROM
+  tracks
+WHERE
+  name REGEXP 'But Why Isn''t Your Syntax More Similar\?'
+