(superset) branch master updated: fix(sql): rebase Dremio dialect on sqlglot's native dialect (#43099)

[jl...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

sadpandajoe pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/superset.git


The following commit(s) were added to refs/heads/master by this push:
     new a63483b9b1a fix(sql): rebase Dremio dialect on sqlglot's native dialect (#43099)
a63483b9b1a is described below

commit a63483b9b1ade3cc009bac85c46d61b363f00c41
Author: Evan Rusackas <ev...@preset.io>
AuthorDate: Thu Aug 13 13:16:06 2026 -0700

    fix(sql): rebase Dremio dialect on sqlglot's native dialect (#43099)
    
    Co-authored-by: Superset Dev <de...@superset.apache.org>
    Co-authored-by: Claude Sonnet 5 <no...@anthropic.com>
    Co-authored-by: Joe Li <jo...@preset.io>
---
 superset/sql/dialects/dremio.py               | 28 ++++++++++++++++++++-------
 tests/unit_tests/sql/dialects/dremio_tests.py | 27 ++++++++++++++++++++++++++
 2 files changed, 48 insertions(+), 7 deletions(-)

diff --git a/superset/sql/dialects/dremio.py b/superset/sql/dialects/dremio.py
index 317882c695e..cd5c428dffe 100644
--- a/superset/sql/dialects/dremio.py
+++ b/superset/sql/dialects/dremio.py
@@ -15,14 +15,20 @@
 # specific language governing permissions and limitations
 # under the License.
 
-from sqlglot import exp, generator, parser
-from sqlglot.dialects.dialect import Dialect, rename_func
+from sqlglot import exp
+from sqlglot.dialects.dialect import rename_func
+from sqlglot.dialects.dremio import Dremio as SqlglotDremio
+from sqlglot.generators.dremio import DremioGenerator
+from sqlglot.parsers.dremio import DremioParser
 
 
 class DremioRegexpSplit(exp.Expression, exp.Func):
     """
     Custom REGEXP_SPLIT function for Dremio that supports 4 arguments.
 
+    sqlglot's native Dremio dialect does not parse this form; Superset adds
+    it on top until it lands upstream.
+
     In sqlglot 30 the Expression class was split from the Func base class; custom
     Func subclasses must now explicitly inherit from both exp.Expression and exp.Func
     so that exp.Expression._set_parent is reachable via the MRO.
@@ -36,15 +42,23 @@ class DremioRegexpSplit(exp.Expression, exp.Func):
     }
 
 
-class Dremio(Dialect):
-    class Parser(parser.Parser):
+class Dremio(SqlglotDremio):
+    """
+    Superset's Dremio dialect, built on sqlglot's native Dremio dialect.
+
+    Adds the 4-argument REGEXP_SPLIT form on top of everything sqlglot
+    already gets right for Dremio (type mapping, DATE_ADD/DATE_SUB
+    semantics, TO_CHAR, CURRENT_DATE_UTC, timestamp-timezone rejection).
+    """
+
+    class Parser(DremioParser):
         FUNCTIONS = {
-            **parser.Parser.FUNCTIONS,
+            **DremioParser.FUNCTIONS,
             "REGEXP_SPLIT": DremioRegexpSplit.from_arg_list,
         }
 
-    class Generator(generator.Generator):
+    class Generator(DremioGenerator):
         TRANSFORMS = {
-            **generator.Generator.TRANSFORMS,
+            **DremioGenerator.TRANSFORMS,
             DremioRegexpSplit: rename_func("REGEXP_SPLIT"),
         }
diff --git a/tests/unit_tests/sql/dialects/dremio_tests.py b/tests/unit_tests/sql/dialects/dremio_tests.py
index 7d7aa9b6702..0de042f3e71 100644
--- a/tests/unit_tests/sql/dialects/dremio_tests.py
+++ b/tests/unit_tests/sql/dialects/dremio_tests.py
@@ -30,3 +30,30 @@ def test_regexp_split() -> None:
     regenerated = ast.sql(dialect=Dremio)
 
     assert regenerated == "SELECT REGEXP_SPLIT(tags, ',', 'ALL', 1000) AS t"
+
+
+def test_inherits_sqlglot_native_type_mapping() -> None:
+    """
+    Superset's Dremio dialect is built on sqlglot's native Dremio dialect,
+    not a bare Dialect base, so it should get correct type mapping for free
+    (e.g. TINYINT, which Dremio doesn't support, maps to INT).
+    """
+    sql = "SELECT CAST(x AS TINYINT)"
+
+    ast = parse_one(sql, dialect=Dremio)
+    regenerated = ast.sql(dialect=Dremio)
+
+    assert regenerated == "SELECT CAST(x AS INT)"
+
+
+def test_inherits_sqlglot_native_current_date_utc() -> None:
+    """
+    sqlglot's native Dremio dialect round-trips CURRENT_DATE_UTC(); a bare
+    Dialect base does not understand this Dremio-specific construct.
+    """
+    sql = "SELECT CURRENT_DATE_UTC()"
+
+    ast = parse_one(sql, dialect=Dremio)
+    regenerated = ast.sql(dialect=Dremio)
+
+    assert regenerated == "SELECT CURRENT_DATE_UTC"