Merge branch 'main' into ci/cargo-audit

Author: LucaCappelletti94

Cargo.toml

@@ -36,6 +36,9 @@ edition = "2021"
 name = "sqlparser"
 path = "src/lib.rs"
 
+[lints.rust]
+unsafe_code = "forbid"
+
 [features]
 default = ["std", "recursive-protection"]
 std = []

derive/Cargo.toml

@@ -35,6 +35,9 @@ edition = "2021"
 [lib]
 proc-macro = true
 
+[lints.rust]
+unsafe_code = "forbid"
+
 [dependencies]
 syn = { version = "2.0", default-features = false, features = ["full", "printing", "parsing", "derive", "proc-macro", "clone-impls"] }
 proc-macro2 = "1.0"

sqlparser_bench/benches/sqlparser_bench.rs

@@ -16,7 +16,7 @@
 // under the License.
 
 use criterion::{criterion_group, criterion_main, Criterion};
-use sqlparser::dialect::GenericDialect;
+use sqlparser::dialect::{GenericDialect, PostgreSqlDialect, SQLiteDialect};
 use sqlparser::keywords::Keyword;
 use sqlparser::parser::Parser;
 use sqlparser::tokenizer::{Span, Word};
@@ -177,11 +177,111 @@ fn parse_compound_chain(c: &mut Criterion) {
     group.finish();
 }
 
+/// Benchmark parsing pathological compound chains with a reserved keyword in
+/// field position, like `SELECT x.not-b.not-b...`. The `.not-b` shape used to
+/// cause 2^N work in `parse_compound_expr` because `parse_prefix` descended
+/// into `parse_not` -> `parse_subexpr`, re-walking the remaining chain at
+/// every segment.
+fn parse_compound_keyword_chain(c: &mut Criterion) {
+    let mut group = c.benchmark_group("parse_compound_keyword_chain");
+    let dialect = GenericDialect {};
+
+    for &n in &[5usize, 10, 15] {
+        let body = std::iter::repeat_n(".not-b", n).collect::<String>();
+        let sql = format!("SELECT x{body}");
+
+        group.bench_function(format!("chain_{n}"), |b| {
+            b.iter(|| {
+                let _ = Parser::parse_sql(&dialect, std::hint::black_box(&sql));
+            });
+        });
+    }
+
+    group.finish();
+}
+
+/// Benchmark parsing pathological `IF(<keyword-fn>(<keyword-fn>(...x` chains
+/// that previously caused 2^N work in `parse_prefix`. Each nested
+/// `current_time(` segment used to be explored twice at every level (once via
+/// the speculative reserved-word arm, once via the unreserved-word fallback),
+/// doubling work per level. Post-fix the cost is linear in chain length.
+fn parse_prefix_keyword_call_chain(c: &mut Criterion) {
+    let mut group = c.benchmark_group("parse_prefix_keyword_call_chain");
+    let dialect = PostgreSqlDialect {};
+
+    for &n in &[10usize, 20, 30] {
+        let sql = String::from("if(") + &"current_time(".repeat(n) + "x";
+
+        group.bench_function(format!("chain_{n}"), |b| {
+            b.iter(|| {
+                let _ = Parser::parse_sql(&dialect, std::hint::black_box(&sql));
+            });
+        });
+    }
+
+    group.finish();
+}
+
+/// Benchmark parsing pathological `case-case-case-...c` chains that
+/// previously caused 2^N work in `parse_prefix`. Each `case` token used to
+/// trigger a speculative `parse_case_expr` that recursively descends the
+/// chain, but the unreserved-word fallback returns `Identifier(case)` so the
+/// overall `parse_prefix` succeeds and the failure cache never fires.
+/// Post-fix the per-arm cache short-circuits the speculative descent.
+fn parse_prefix_case_chain(c: &mut Criterion) {
+    let mut group = c.benchmark_group("parse_prefix_case_chain");
+    let dialect = SQLiteDialect {};
+
+    for &n in &[10usize, 20, 30] {
+        let sql = "case\t-".repeat(n) + "c";
+
+        group.bench_function(format!("chain_{n}"), |b| {
+            b.iter(|| {
+                let _ = Parser::parse_sql(&dialect, std::hint::black_box(&sql));
+            });
+        });
+    }
+
+    group.finish();
+}
+
+/// Benchmark parsing pathological paren chains that previously caused 2^N
+/// work in `parse_table_factor`. The input `SELECT 1 FROM ((((...` rejects
+/// at EOF, which used to force exponential backtracking through the chain.
+fn parse_table_factor_paren_chain(c: &mut Criterion) {
+    let mut group = c.benchmark_group("parse_table_factor_paren_chain");
+    let dialect = GenericDialect {};
+
+    for &n in &[10usize, 20, 30] {
+        let mut sql = String::from("SELECT 1 ");
+        for _ in 0..5 {
+            sql.push_str("FROM ");
+            sql.push_str(&"(".repeat(n));
+            sql.push(' ');
+        }
+
+        group.bench_function(format!("chain_{n}"), |b| {
+            b.iter(|| {
+                let _ = Parser::new(&dialect)
+                    .with_recursion_limit(256)
+                    .try_with_sql(std::hint::black_box(&sql))
+                    .and_then(|mut p| p.parse_statements());
+            });
+        });
+    }
+
+    group.finish();
+}
+
 criterion_group!(
     benches,
     basic_queries,
     word_to_ident,
     parse_many_identifiers,
-    parse_compound_chain
+    parse_compound_chain,
+    parse_compound_keyword_chain,
+    parse_prefix_keyword_call_chain,
+    parse_prefix_case_chain,
+    parse_table_factor_paren_chain
 );
 criterion_main!(benches);

src/ast/comments.rs

@@ -33,7 +33,7 @@ impl Comments {
     /// last accepted comment.  In other words, this method will skip the
     /// comment if its comming out of order (as encountered in the parsed
     /// source code.)
-    pub(crate) fn offer(&mut self, comment: CommentWithSpan) {
+    pub fn offer(&mut self, comment: CommentWithSpan) {
         if self
             .0
             .last()
@@ -71,7 +71,7 @@ impl Comments {
     /// // all comments appearing before line seven, i.e. before the first statement itself
     /// assert_eq!(
     ///    &comments.find(..Location::new(7, 1)).map(|c| c.as_str()).collect::<Vec<_>>(),
-    ///    &["\n header comment ...\n ... spanning multiple lines\n", " first statement\n"]);
+    ///    &["\n header comment ...\n ... spanning multiple lines\n", " first statement"]);
     ///
     /// // all comments appearing within the first statement
     /// assert_eq!(
@@ -81,7 +81,7 @@ impl Comments {
     /// // all comments appearing within or after the first statement
     /// assert_eq!(
     ///    &comments.find(Location::new(7, 1)..).map(|c| c.as_str()).collect::<Vec<_>>(),
-    ///    &[" world ", " second statement\n", " trailing comment\n"]);
+    ///    &[" world ", " second statement", " trailing comment"]);
     /// ```
     ///
     /// The [Spanned](crate::ast::Spanned) trait allows you to access location

src/ast/ddl.rs

@@ -5632,6 +5632,161 @@ impl Spanned for AlterFunction {
     }
 }
 
+/// Text search object kind.
+///
+/// See [PostgreSQL](https://www.postgresql.org/docs/current/textsearch-intro.html).
+#[derive(Debug, Clone, Copy, PartialEq, PartialOrd, Eq, Ord, Hash)]
+#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
+#[cfg_attr(feature = "visitor", derive(Visit, VisitMut))]
+pub enum TextSearchObjectType {
+    /// `DICTIONARY`
+    Dictionary,
+    /// `CONFIGURATION`
+    Configuration,
+    /// `TEMPLATE`
+    Template,
+    /// `PARSER`
+    Parser,
+}
+
+impl fmt::Display for TextSearchObjectType {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        match self {
+            TextSearchObjectType::Dictionary => write!(f, "DICTIONARY"),
+            TextSearchObjectType::Configuration => write!(f, "CONFIGURATION"),
+            TextSearchObjectType::Template => write!(f, "TEMPLATE"),
+            TextSearchObjectType::Parser => write!(f, "PARSER"),
+        }
+    }
+}
+
+/// `CREATE TEXT SEARCH ...` statement.
+///
+/// See [PostgreSQL](https://www.postgresql.org/docs/current/sql-createtsdictionary.html).
+#[derive(Debug, Clone, PartialEq, PartialOrd, Eq, Ord, Hash)]
+#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
+#[cfg_attr(feature = "visitor", derive(Visit, VisitMut))]
+pub struct CreateTextSearch {
+    /// The specific text search object type.
+    pub object_type: TextSearchObjectType,
+    /// Object name.
+    pub name: ObjectName,
+    /// Parenthesized options.
+    pub options: Vec<SqlOption>,
+}
+
+impl fmt::Display for CreateTextSearch {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(
+            f,
+            "CREATE TEXT SEARCH {} {} ({})",
+            self.object_type,
+            self.name,
+            display_comma_separated(&self.options)
+        )
+    }
+}
+
+impl Spanned for CreateTextSearch {
+    fn span(&self) -> Span {
+        Span::empty()
+    }
+}
+
+/// Option assignment used by `ALTER TEXT SEARCH ... ( ... )`.
+///
+/// See [PostgreSQL](https://www.postgresql.org/docs/current/sql-altertsdictionary.html).
+#[derive(Debug, Clone, PartialEq, PartialOrd, Eq, Ord, Hash)]
+#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
+#[cfg_attr(feature = "visitor", derive(Visit, VisitMut))]
+pub struct AlterTextSearchOption {
+    /// Option name.
+    pub key: Ident,
+    /// Optional value (`option [= value]`).
+    pub value: Option<Expr>,
+}
+
+impl fmt::Display for AlterTextSearchOption {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        match &self.value {
+            Some(value) => write!(f, "{} = {}", self.key, value),
+            None => write!(f, "{}", self.key),
+        }
+    }
+}
+
+/// Operation for `ALTER TEXT SEARCH ...`.
+///
+/// See [PostgreSQL](https://www.postgresql.org/docs/current/sql-altertsdictionary.html).
+#[derive(Debug, Clone, PartialEq, PartialOrd, Eq, Ord, Hash)]
+#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
+#[cfg_attr(feature = "visitor", derive(Visit, VisitMut))]
+pub enum AlterTextSearchOperation {
+    /// `RENAME TO new_name`
+    RenameTo {
+        /// New name.
+        new_name: Ident,
+    },
+    /// `OWNER TO ...`
+    OwnerTo(Owner),
+    /// `SET SCHEMA schema_name`
+    SetSchema {
+        /// Target schema.
+        schema_name: ObjectName,
+    },
+    /// `( option [= value] [, ...] )`
+    SetOptions {
+        /// Text search options to apply.
+        options: Vec<AlterTextSearchOption>,
+    },
+}
+
+impl fmt::Display for AlterTextSearchOperation {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        match self {
+            AlterTextSearchOperation::RenameTo { new_name } => write!(f, "RENAME TO {new_name}"),
+            AlterTextSearchOperation::OwnerTo(owner) => write!(f, "OWNER TO {owner}"),
+            AlterTextSearchOperation::SetSchema { schema_name } => {
+                write!(f, "SET SCHEMA {schema_name}")
+            }
+            AlterTextSearchOperation::SetOptions { options } => {
+                write!(f, "({})", display_comma_separated(options))
+            }
+        }
+    }
+}
+
+/// `ALTER TEXT SEARCH ...` statement.
+///
+/// See [PostgreSQL](https://www.postgresql.org/docs/current/sql-altertsdictionary.html).
+#[derive(Debug, Clone, PartialEq, PartialOrd, Eq, Ord, Hash)]
+#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
+#[cfg_attr(feature = "visitor", derive(Visit, VisitMut))]
+pub struct AlterTextSearch {
+    /// The specific text search object type.
+    pub object_type: TextSearchObjectType,
+    /// Object name.
+    pub name: ObjectName,
+    /// Operation to apply.
+    pub operation: AlterTextSearchOperation,
+}
+
+impl fmt::Display for AlterTextSearch {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(
+            f,
+            "ALTER TEXT SEARCH {} {} {}",
+            self.object_type, self.name, self.operation
+        )
+    }
+}
+
+impl Spanned for AlterTextSearch {
+    fn span(&self) -> Span {
+        Span::empty()
+    }
+}
+
 /// CREATE POLICY statement.
 ///
 /// See [PostgreSQL](https://www.postgresql.org/docs/current/sql-createpolicy.html)