perf: pin polyglot with faster column qualification (#923)

* perf(compile): bump polyglot for linear column qualification and commented set operations

Bump polyglot-sql to a revision that borrows source column sets during
schema-aware column qualification, so wide relations and SELECT * over
derived tables qualify in linear time, and that builds scopes for
set-operation chains with comments before the operators in linear time.
Analysis results are unchanged.

Extend the long UNION ALL fixture guard with an input fixture whose rows
are separated by a comment before each UNION ALL; 200-row commented
fixtures previously took about 17 times as long as 50-row fixtures.

* build: pin polyglot to the merged column-set and comment-scope changes

Author: kvlonge

Cargo.lock

@@ -17,7 +17,7 @@ globset = "0.4"
 mimalloc = { version = "0.1", default-features = false, features = ["local_dynamic_tls"] }
 proc-macro2 = { version = "1", features = ["span-locations"] }
 pyo3 = { version = "0.26", features = ["abi3-py312"] }
-polyglot-sql = { git = "https://github.com/chio-labs/polyglot.git", rev = "a7413439c520d7949924149b7ea02fa3817463fc", default-features = false, features = [
+polyglot-sql = { git = "https://github.com/chio-labs/polyglot.git", rev = "9ff0af282829d92efeeb36b6a820be2271e1da49", default-features = false, features = [
     "ast-tools",
     "builder",
     "dialect-bigquery",
@@ -29,7 +29,7 @@ polyglot-sql = { git = "https://github.com/chio-labs/polyglot.git", rev = "a7413
     "semantic",
     "stacker",
 ] }
-polyglot-sql-function-catalogs = { git = "https://github.com/chio-labs/polyglot.git", rev = "a7413439c520d7949924149b7ea02fa3817463fc", default-features = false, features = [
+polyglot-sql-function-catalogs = { git = "https://github.com/chio-labs/polyglot.git", rev = "9ff0af282829d92efeeb36b6a820be2271e1da49", default-features = false, features = [
     "dialect-duckdb",
 ] }
 quote = "1"

Cargo.toml

@@ -496,6 +496,7 @@ class UnionFixtureCompileScalingTestCase:
     expected_sql_tests: int
     expected_errors: int
     expected_max_scaling_ratio: float
+    input_fixture_row_separator: str = " UNION ALL\n"
 
 
 @dataclass(frozen=True)

tests/e2e/src/sqlbuild/cli/commands/main/compile/_test_types.py

@@ -2771,17 +2771,25 @@ class UnionFixtureCompileMeasurement(NamedTuple):
     errors: int
 
 
-def _union_all_fixture_rows(*, row_count: int, status: str) -> str:
-    return " UNION ALL\n".join(
+def _union_all_fixture_rows(*, row_count: int, status: str, separator: str) -> str:
+    return separator.join(
         f"  SELECT {row} AS order_id, CAST({row} AS DOUBLE) AS amount, '{status}' AS status"
         for row in range(row_count)
     )
 
 
 def write_union_fixture_test_project(
-    *, project_dir: Path, sql_test_count: int, fixture_row_count: int
+    *,
+    project_dir: Path,
+    sql_test_count: int,
+    fixture_row_count: int,
+    input_fixture_row_separator: str,
 ) -> None:
-    """Write a DuckDB project whose SQL tests use long UNION ALL fixture chains."""
+    """Write a DuckDB project whose SQL tests use long UNION ALL fixture chains.
+
+    `input_fixture_row_separator` joins the rows of the mocked input fixture; the
+    expected fixture always uses plain `UNION ALL`.
+    """
 
     files: dict[str, str] = {
         "sqlbuild_project.toml": (
@@ -2798,13 +2806,17 @@ def write_union_fixture_test_project(
         ),
     }
     for index in range(sql_test_count):
-        fixture_rows: str = _union_all_fixture_rows(
-            row_count=fixture_row_count, status=f"status_{index:02d}"
+        status: str = f"status_{index:02d}"
+        input_rows: str = _union_all_fixture_rows(
+            row_count=fixture_row_count, status=status, separator=input_fixture_row_separator
+        )
+        expected_rows: str = _union_all_fixture_rows(
+            row_count=fixture_row_count, status=status, separator=" UNION ALL\n"
         )
         files[f"tests/unit/orders_case_{index:02d}.sql"] = (
             f'TEST (name "orders_case_{index:02d}");\n\n'
-            f"WITH\n__ref__base_orders AS (\n{fixture_rows}\n),\n"
-            f"__expected__orders AS (\n{fixture_rows}\n)\nSELECT 1\n"
+            f"WITH\n__ref__base_orders AS (\n{input_rows}\n),\n"
+            f"__expected__orders AS (\n{expected_rows}\n)\nSELECT 1\n"
         )
     prepare_inline_project(
         tmp_path=project_dir.parent, project_name=project_dir.name, repo_files=files

tests/e2e/src/sqlbuild/cli/commands/main/compile/helpers.py

@@ -140,7 +140,18 @@ def test_given_wide_scan_heavy_projects_when_doubling_sql_size_then_compile_scal
             expected_sql_tests=4,
             expected_errors=0,
             expected_max_scaling_ratio=6.0,
-        )
+        ),
+        UnionFixtureCompileScalingTestCase(
+            description="quadrupling commented UNION ALL input fixture rows keeps SQL test analysis linear",
+            sql_test_count=4,
+            small_fixture_rows=50,
+            large_fixture_rows=200,
+            measured_runs=3,
+            expected_sql_tests=4,
+            expected_errors=0,
+            expected_max_scaling_ratio=6.0,
+            input_fixture_row_separator="\n  -- fixture row\n  UNION ALL\n",
+        ),
     ],
     ids=lambda case: case.description,
 )
@@ -154,11 +165,13 @@ def test_given_sql_tests_with_union_all_fixtures_when_quadrupling_rows_then_anal
         project_dir=small_project_dir,
         sql_test_count=test_case.sql_test_count,
         fixture_row_count=test_case.small_fixture_rows,
+        input_fixture_row_separator=test_case.input_fixture_row_separator,
     )
     write_union_fixture_test_project(
         project_dir=large_project_dir,
         sql_test_count=test_case.sql_test_count,
         fixture_row_count=test_case.large_fixture_rows,
+        input_fixture_row_separator=test_case.input_fixture_row_separator,
     )
     small: UnionFixtureCompileMeasurement = measure_union_fixture_compile(
         project_dir=small_project_dir, runs=test_case.measured_runs