Databricks SQL native parser

Add crates/sqlquack-syntax-databricks, a scanner and parser for
Databricks SQL that builds the shared sqlquack-syntax tree, with
parse_strict for callers that need every statement parsed in full.

The parser refuses deep nesting with a parse error instead of
exhausting the stack, and after each statement it applies the shared
tree-depth check, which refuses a tree nested past 8,000 levels.

The syntax-coverage checker links the parser as the local parser for
the Databricks dialect, and the deep-nesting probe covers it, with its
ARRAY<> type constructor and its pipe chain. Every Databricks construct
is clean at every depth in both profiles, so the allowlist has no
Databricks entries.

Author: kvlonge

Cargo.lock

@@ -21,6 +21,7 @@ members = [
     "crates/sqlquack-format",
     "crates/sqlquack-format-check",
     "crates/sqlquack-syntax-duckdb",
+    "crates/sqlquack-syntax-databricks",
     "crates/sqlquack-python",
 ]
 exclude = ["fuzz"]
@@ -78,6 +79,7 @@ sqlquack-syntax-derive = { path = "crates/sqlquack-syntax-derive" }
 sqlquack-syntax-postgres = { path = "crates/sqlquack-syntax-postgres" }
 sqlquack-syntax-bigquery = { path = "crates/sqlquack-syntax-bigquery" }
 sqlquack-syntax-duckdb = { path = "crates/sqlquack-syntax-duckdb" }
+sqlquack-syntax-databricks = { path = "crates/sqlquack-syntax-databricks" }
 stacker = "0.1"
 syn = { version = "2", features = ["full"] }
 tempfile = "3"

Cargo.toml

@@ -17,6 +17,7 @@ sha2.workspace = true
 sqlquack-syntax.workspace = true
 sqlquack-syntax-bigquery.workspace = true
 sqlquack-syntax-duckdb.workspace = true
+sqlquack-syntax-databricks.workspace = true
 sqlquack-syntax-postgres.workspace = true
 toml.workspace = true
 

crates/sqlquack-syntax-coverage/Cargo.toml

@@ -69,13 +69,19 @@ fn duckdb_verdict(sql: &str) -> LocalVerdict {
     native_verdict(sqlquack_syntax_duckdb::parse_strict(sql))
 }
 
+/// SQLQuack's strict Databricks SQL verdict.
+fn databricks_verdict(sql: &str) -> LocalVerdict {
+    native_verdict(sqlquack_syntax_databricks::parse_strict(sql))
+}
+
 /// The verdict of each dialect's local parser, or none when the dialect has no parser yet.
 pub(crate) fn local_verdict(parser: LocalParser) -> Option<fn(&str) -> LocalVerdict> {
     match parser {
         LocalParser::Snowflake => Some(snowflake_verdict),
         LocalParser::Postgres => Some(postgres_verdict),
         LocalParser::BigQuery => Some(bigquery_verdict),
         LocalParser::DuckDb => Some(duckdb_verdict),
+        LocalParser::Databricks => Some(databricks_verdict),
         LocalParser::None => None,
     }
 }

crates/sqlquack-syntax-coverage/src/cli/_helpers/local.rs

@@ -27,6 +27,7 @@ fn constructed_type(parser: LocalParser, depth: usize) -> Option<String> {
         LocalParser::Snowflake => Some(nest("ARRAY(", "INT", ")", depth)),
         LocalParser::BigQuery => Some(nest("STRUCT<x ", "INT64", ">", depth)),
         LocalParser::DuckDb => Some(nest("STRUCT(a ", "INT", ")", depth)),
+        LocalParser::Databricks => Some(nest("ARRAY<", "INT", ">", depth)),
         LocalParser::Postgres | LocalParser::None => None,
     }
 }
@@ -38,7 +39,10 @@ fn mixed_type(parser: LocalParser, depth: usize) -> Option<String> {
     let open: &str = match parser {
         LocalParser::Snowflake => "ARRAY(",
         LocalParser::DuckDb => "STRUCT(a ",
-        LocalParser::Postgres | LocalParser::BigQuery | LocalParser::None => return None,
+        LocalParser::Postgres
+        | LocalParser::BigQuery
+        | LocalParser::Databricks
+        | LocalParser::None => return None,
     };
     Some(format!(
         "{}INT{}{run}",
@@ -258,7 +262,9 @@ pub(crate) fn constructs() -> Vec<Construct> {
         Construct {
             name: "nested_pipe_chain",
             text: |parser, depth| match parser {
-                LocalParser::BigQuery => Some(nested_chain("FROM t", " |> WHERE TRUE", depth)),
+                LocalParser::BigQuery | LocalParser::Databricks => {
+                    Some(nested_chain("FROM t", " |> WHERE TRUE", depth))
+                }
                 _ => None,
             },
         },

crates/sqlquack-syntax-coverage/src/depth/_helpers/constructs.rs

@@ -17,9 +17,10 @@ pub const MIXED_SUFFIX_RUN: usize = 60;
 /// How many links each parenthesised level of a nested chain holds.
 pub const NESTED_CHAIN_LINKS: usize = 60;
 /// Every native parser this checker links, each probed in turn.
-pub const PROBED_PARSERS: [LocalParser; 4] = [
+pub const PROBED_PARSERS: [LocalParser; 5] = [
     LocalParser::Snowflake,
     LocalParser::Postgres,
     LocalParser::BigQuery,
     LocalParser::DuckDb,
+    LocalParser::Databricks,
 ];

crates/sqlquack-syntax-coverage/src/depth/constants.rs

@@ -90,6 +90,12 @@ fn given_a_parser_without_a_syntax_when_listing_constructs_then_only_its_own_are
             construct: "nested_pipe_chain",
             expected_probed: true,
         },
+        ConstructsTestCase {
+            description: "Databricks chains pipe operators",
+            parser: LocalParser::Databricks,
+            construct: "nested_pipe_chain",
+            expected_probed: true,
+        },
         ConstructsTestCase {
             description: "Postgres has no pipe operator",
             parser: LocalParser::Postgres,

crates/sqlquack-syntax-coverage/src/depth/tests/test_probe_construct.rs

@@ -650,6 +650,8 @@ pub enum LocalParser {
     #[serde(rename = "duckdb")]
     #[value(name = "duckdb")]
     DuckDb,
+    /// `sqlquack_syntax_databricks`, through the isolated worker.
+    Databricks,
     /// No parser yet: every case fails locally as `no_parser`.
     None,
 }

crates/sqlquack-syntax-coverage/src/formats/models.rs

@@ -0,0 +1,20 @@
+[package]
+name = "sqlquack-syntax-databricks"
+version.workspace = true
+description = "SQLQuack's Databricks SQL scanner and parser into the shared sqlquack-syntax tree."
+edition.workspace = true
+license.workspace = true
+repository.workspace = true
+homepage.workspace = true
+publish = false
+rust-version.workspace = true
+
+[dependencies]
+sqlquack-syntax.workspace = true
+stacker.workspace = true
+
+[dev-dependencies]
+serde_json.workspace = true
+
+[lints]
+workspace = true