refactor(parser): use v2 parser for cast/extract/substring/position/o…

…verlay (#17053)

Signed-off-by: TennyZhuang <[email protected]>

src/sqlparser/src/parser.rs

@@ -170,7 +170,7 @@ type ColumnsDefTuple = (
 
 /// Reference:
 /// <https://www.postgresql.org/docs/current/sql-syntax-lexical.html#SQL-PRECEDENCE>
-#[derive(Debug, PartialEq, Eq, PartialOrd, Ord)]
+#[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord)]
 pub enum Precedence {
     Zero = 0,
     LogicalOr, // 5 in upstream
@@ -1009,28 +1009,12 @@ impl Parser<'_> {
 
     /// Parse a SQL CAST function e.g. `CAST(expr AS FLOAT)`
     pub fn parse_cast_expr(&mut self) -> PResult<Expr> {
-        self.expect_token(&Token::LParen)?;
-        let expr = self.parse_expr()?;
-        self.expect_keyword(Keyword::AS)?;
-        let data_type = self.parse_data_type()?;
-        self.expect_token(&Token::RParen)?;
-        Ok(Expr::Cast {
-            expr: Box::new(expr),
-            data_type,
-        })
+        parser_v2::expr_cast(self)
     }
 
     /// Parse a SQL TRY_CAST function e.g. `TRY_CAST(expr AS FLOAT)`
     pub fn parse_try_cast_expr(&mut self) -> PResult<Expr> {
-        self.expect_token(&Token::LParen)?;
-        let expr = self.parse_expr()?;
-        self.expect_keyword(Keyword::AS)?;
-        let data_type = self.parse_data_type()?;
-        self.expect_token(&Token::RParen)?;
-        Ok(Expr::TryCast {
-            expr: Box::new(expr),
-            data_type,
-        })
+        parser_v2::expr_try_cast(self)
     }
 
     /// Parse a SQL EXISTS expression e.g. `WHERE EXISTS(SELECT ...)`.
@@ -1042,83 +1026,21 @@ impl Parser<'_> {
     }
 
     pub fn parse_extract_expr(&mut self) -> PResult<Expr> {
-        self.expect_token(&Token::LParen)?;
-        let field = self.parse_date_time_field_in_extract()?;
-        self.expect_keyword(Keyword::FROM)?;
-        let expr = self.parse_expr()?;
-        self.expect_token(&Token::RParen)?;
-        Ok(Expr::Extract {
-            field,
-            expr: Box::new(expr),
-        })
+        parser_v2::expr_extract(self)
     }
 
     pub fn parse_substring_expr(&mut self) -> PResult<Expr> {
-        // PARSE SUBSTRING (EXPR [FROM 1] [FOR 3])
-        self.expect_token(&Token::LParen)?;
-        let expr = self.parse_expr()?;
-        let mut from_expr = None;
-        if self.parse_keyword(Keyword::FROM) || self.consume_token(&Token::Comma) {
-            from_expr = Some(self.parse_expr()?);
-        }
-
-        let mut to_expr = None;
-        if self.parse_keyword(Keyword::FOR) || self.consume_token(&Token::Comma) {
-            to_expr = Some(self.parse_expr()?);
-        }
-        self.expect_token(&Token::RParen)?;
-
-        Ok(Expr::Substring {
-            expr: Box::new(expr),
-            substring_from: from_expr.map(Box::new),
-            substring_for: to_expr.map(Box::new),
-        })
+        parser_v2::expr_substring(self)
     }
 
     /// `POSITION(<expr> IN <expr>)`
     pub fn parse_position_expr(&mut self) -> PResult<Expr> {
-        self.expect_token(&Token::LParen)?;
-
-        // Logically `parse_expr`, but limited to those with precedence higher than `BETWEEN`/`IN`,
-        // to avoid conflict with general IN operator, for example `position(a IN (b) IN (c))`.
-        // https://github.com/postgres/postgres/blob/REL_15_2/src/backend/parser/gram.y#L16012
-        let substring = self.parse_subexpr(Precedence::Between)?;
-        self.expect_keyword(Keyword::IN)?;
-        let string = self.parse_subexpr(Precedence::Between)?;
-
-        self.expect_token(&Token::RParen)?;
-
-        Ok(Expr::Position {
-            substring: Box::new(substring),
-            string: Box::new(string),
-        })
+        parser_v2::expr_position(self)
     }
 
     /// `OVERLAY(<expr> PLACING <expr> FROM <expr> [ FOR <expr> ])`
     pub fn parse_overlay_expr(&mut self) -> PResult<Expr> {
-        self.expect_token(&Token::LParen)?;
-
-        let expr = self.parse_expr()?;
-
-        self.expect_keyword(Keyword::PLACING)?;
-        let new_substring = self.parse_expr()?;
-
-        self.expect_keyword(Keyword::FROM)?;
-        let start = self.parse_expr()?;
-
-        let mut count = None;
-        if self.parse_keyword(Keyword::FOR) {
-            count = Some(self.parse_expr()?);
-        }
-
-        self.expect_token(&Token::RParen)?;
-
-        Ok(Expr::Overlay {
-            expr: Box::new(expr),
-            new_substring: Box::new(new_substring),
-            start: Box::new(start),
-            count: count.map(Box::new),
-        })
+        parser_v2::expr_overlay(self)
     }
 
     /// `TRIM ([WHERE] ['text'] FROM 'text')`\

src/sqlparser/src/parser_v2/expr.rs

@@ -9,14 +9,17 @@
 // 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.
-use winnow::combinator::{cut_err, opt, preceded, repeat, trace};
+use winnow::combinator::{alt, cut_err, opt, preceded, repeat, seq, trace};
+use winnow::error::ContextError;
 use winnow::{PResult, Parser};
 
-use super::TokenStream;
+use super::{data_type, token, ParserExt, TokenStream};
 use crate::ast::Expr;
 use crate::keywords::Keyword;
+use crate::parser::Precedence;
+use crate::tokenizer::Token;
 
-fn expr<S>(input: &mut S) -> PResult<Expr>
+fn expr_parse<S>(input: &mut S) -> PResult<Expr>
 where
     S: TokenStream,
 {
@@ -27,22 +30,32 @@ where
     .parse_next(input)
 }
 
+fn subexpr<S>(precedence: Precedence) -> impl Parser<S, Expr, ContextError>
+where
+    S: TokenStream,
+{
+    // TODO: implement this function using combinator style.
+    trace("subexpr", move |input: &mut S| {
+        input.parse_v1(|parser| parser.parse_subexpr(precedence))
+    })
+}
+
 pub fn expr_case<S>(input: &mut S) -> PResult<Expr>
 where
     S: TokenStream,
 {
     let parse = (
-        opt(expr),
+        opt(expr_parse),
         repeat(
             1..,
             (
                 Keyword::WHEN,
-                cut_err(expr),
+                cut_err(expr_parse),
                 cut_err(Keyword::THEN),
-                cut_err(expr),
+                cut_err(expr_parse),
             ),
         ),
-        opt(preceded(Keyword::ELSE, cut_err(expr))),
+        opt(preceded(Keyword::ELSE, cut_err(expr_parse))),
         cut_err(Keyword::END),
     )
         .map(|(operand, branches, else_result, _)| {
@@ -58,3 +71,126 @@ where
 
     trace("expr_case", parse).parse_next(input)
 }
+
+/// Consume a SQL CAST function e.g. `CAST(expr AS FLOAT)`
+pub fn expr_cast<S>(input: &mut S) -> PResult<Expr>
+where
+    S: TokenStream,
+{
+    let parse = cut_err(seq! {Expr::Cast {
+        _: Token::LParen,
+        expr: expr_parse.map(Box::new),
+        _: Keyword::AS,
+        data_type: data_type,
+        _: Token::RParen,
+    }});
+
+    trace("expr_cast", parse).parse_next(input)
+}
+
+/// Consume a SQL TRY_CAST function e.g. `TRY_CAST(expr AS FLOAT)`
+pub fn expr_try_cast<S>(input: &mut S) -> PResult<Expr>
+where
+    S: TokenStream,
+{
+    let parse = cut_err(seq! {Expr::TryCast {
+        _: Token::LParen,
+        expr: expr_parse.map(Box::new),
+        _: Keyword::AS,
+        data_type: data_type,
+        _: Token::RParen,
+    }});
+
+    trace("expr_try_cast", parse).parse_next(input)
+}
+
+/// Consume a SQL EXTRACT function e.g. `EXTRACT(YEAR FROM expr)`
+pub fn expr_extract<S>(input: &mut S) -> PResult<Expr>
+where
+    S: TokenStream,
+{
+    let mut date_time_field = token
+        .verify_map(|token| match token.token {
+            Token::Word(w) => Some(w.value.to_uppercase()),
+            Token::SingleQuotedString(s) => Some(s.to_uppercase()),
+            _ => None,
+        })
+        .expect("date/time field");
+
+    let parse = cut_err(seq! {Expr::Extract {
+        _: Token::LParen,
+        field: date_time_field,
+        _: Keyword::FROM,
+        expr: expr_parse.map(Box::new),
+        _: Token::RParen,
+    }});
+
+    trace("expr_extract", parse).parse_next(input)
+}
+
+/// Consume `SUBSTRING (EXPR [FROM 1] [FOR 3])`
+pub fn expr_substring<S>(input: &mut S) -> PResult<Expr>
+where
+    S: TokenStream,
+{
+    let mut substring_from = opt(preceded(
+        alt((Token::Comma.void(), Keyword::FROM.void())),
+        cut_err(expr_parse).map(Box::new),
+    ));
+    let mut substring_for = opt(preceded(
+        alt((Token::Comma.void(), Keyword::FOR.void())),
+        cut_err(expr_parse).map(Box::new),
+    ));
+    let parse = cut_err(seq! {Expr::Substring {
+        _: Token::LParen,
+        expr: expr_parse.map(Box::new),
+        substring_from: substring_from,
+        substring_for: substring_for,
+        _: Token::RParen,
+    }});
+
+    trace("expr_substring", parse).parse_next(input)
+}
+
+/// `POSITION(<expr> IN <expr>)`
+pub fn expr_position<S>(input: &mut S) -> PResult<Expr>
+where
+    S: TokenStream,
+{
+    let parse = cut_err(seq! {Expr::Position {
+        _: Token::LParen,
+        // Logically `parse_expr`, but limited to those with precedence higher than `BETWEEN`/`IN`,
+        // to avoid conflict with general IN operator, for example `position(a IN (b) IN (c))`.
+        // https://github.com/postgres/postgres/blob/REL_15_2/src/backend/parser/gram.y#L16012
+        substring: subexpr(Precedence::Between).map(Box::new),
+        _: Keyword::IN,
+        string: subexpr(Precedence::Between).map(Box::new),
+        _: Token::RParen,
+    }});
+
+    trace("expr_position", parse).parse_next(input)
+}
+
+/// `OVERLAY(<expr> PLACING <expr> FROM <expr> [ FOR <expr> ])`
+pub fn expr_overlay<S>(input: &mut S) -> PResult<Expr>
+where
+    S: TokenStream,
+{
+    let mut count_parse = opt(preceded(
+        Keyword::FOR.void(),
+        cut_err(expr_parse).map(Box::new),
+    ));
+
+    let parse = cut_err(seq! {Expr::Overlay {
+        _: Token::LParen,
+        expr: expr_parse.map(Box::new),
+        _: Keyword::PLACING,
+        new_substring: expr_parse.map(Box::new),
+        _: Keyword::FROM,
+        start: expr_parse.map(Box::new),
+        count: count_parse,
+        _: Token::RParen,
+    }});
+
+    trace("expr_overlay", parse).parse_next(input)
+}

src/sqlparser/tests/sqlparser_common.rs

@@ -1268,9 +1268,6 @@ fn parse_extract() {
     verified_stmt("SELECT EXTRACT(HOUR FROM d)");
     verified_stmt("SELECT EXTRACT(MINUTE FROM d)");
     verified_stmt("SELECT EXTRACT(SECOND FROM d)");
-
-    let res = parse_sql_statements("SELECT EXTRACT(0 FROM d)");
-    assert!(format!("{}", res.unwrap_err()).contains("expected date/time field, found: 0"));
 }
 
 #[test]
@@ -2919,38 +2916,6 @@ fn parse_substring() {
     one_statement_parses_to("SELECT SUBSTRING('1' FOR 3)", "SELECT SUBSTRING('1' FOR 3)");
 }
 
-#[test]
-fn parse_overlay() {
-    one_statement_parses_to(
-        "SELECT OVERLAY('abc' PLACING 'xyz' FROM 1)",
-        "SELECT OVERLAY('abc' PLACING 'xyz' FROM 1)",
-    );
-
-    one_statement_parses_to(
-        "SELECT OVERLAY('abc' PLACING 'xyz' FROM 1 FOR 2)",
-        "SELECT OVERLAY('abc' PLACING 'xyz' FROM 1 FOR 2)",
-    );
-
-    for (sql, err_msg) in [
-        ("SELECT OVERLAY('abc', 'xyz')", "expected PLACING, found: ,"),
-        (
-            "SELECT OVERLAY('abc' PLACING 'xyz')",
-            "expected FROM, found: )",
-        ),
-        (
-            "SELECT OVERLAY('abc' PLACING 'xyz' FOR 2)",
-            "expected FROM, found: FOR",
-        ),
-        (
-            "SELECT OVERLAY('abc' PLACING 'xyz' FOR 2 FROM 1)",
-            "expected FROM, found: FOR",
-        ),
-    ] {
-        let res = parse_sql_statements(sql);
-        assert!(format!("{}", res.unwrap_err()).contains(err_msg));
-    }
-}
-
 #[test]
 fn parse_trim() {
     one_statement_parses_to(

src/sqlparser/tests/testdata/extract.yaml

@@ -0,0 +1,6 @@
+# This file is automatically generated by `src/sqlparser/tests/parser_test.rs`.
+- input: SELECT EXTRACT(0 FROM d)
+  error_msg: |-
+    sql parser error: expected date/time field
+    LINE 1: SELECT EXTRACT(0 FROM d)
+                           ^

src/sqlparser/tests/testdata/overlay.yaml

@@ -0,0 +1,13 @@
+# This file is automatically generated by `src/sqlparser/tests/parser_test.rs`.
+- input: SELECT OVERLAY('abc' PLACING 'xyz' FROM 1)
+  formatted_sql: SELECT OVERLAY('abc' PLACING 'xyz' FROM 1)
+- input: SELECT OVERLAY('abc' PLACING 'xyz' FROM 1 FOR 2)
+  formatted_sql: SELECT OVERLAY('abc' PLACING 'xyz' FROM 1 FOR 2)
+- input: SELECT OVERLAY('abc', 'xyz')
+  error_msg: "sql parser error: \nLINE 1: SELECT OVERLAY('abc', 'xyz')\n                            ^"
+- input: SELECT OVERLAY('abc' PLACING 'xyz')
+  error_msg: "sql parser error: \nLINE 1: SELECT OVERLAY('abc' PLACING 'xyz')\n                                          ^"
+- input: SELECT OVERLAY('abc' PLACING 'xyz' FOR 2)
+  error_msg: "sql parser error: \nLINE 1: SELECT OVERLAY('abc' PLACING 'xyz' FOR 2)\n                                           ^"
+- input: SELECT OVERLAY('abc' PLACING 'xyz' FOR 2 FROM 1)
+  error_msg: "sql parser error: \nLINE 1: SELECT OVERLAY('abc' PLACING 'xyz' FOR 2 FROM 1)\n                                           ^"