fix: preserve extensionless document suffix on GaussDB (#18483)

### Summary

RAGFlow's "Create empty document" flow accepts names without a file
extension. The `POST /datasets/<dataset_id>/documents?type=empty` route
calls `_upload_empty_document()`, where `Path(name).suffix.lstrip(".")`
returns `""`.

In GaussDB's A/ORA compatibility mode, that empty string is persisted as
SQL `NULL`. Because `document.suffix` was defined as `NOT NULL`, the
insert failed with a constraint violation.

api/db/db_models.py

@@ -1321,7 +1321,7 @@ class Document(DataBaseModel):
     progress_msg = TextField(null=True, help_text="process message", default="")
     process_begin_at = DateTimeField(null=True, index=True)
     process_duration = FloatField(default=0)
-    suffix = CharField(max_length=32, null=False, help_text="The real file extension suffix", index=True)
+    suffix = EmptyStringCharField(max_length=32, null=False, help_text="The real file extension suffix", index=True)
 
     content_hash = CharField(max_length=32, null=True, help_text="xxhash128 of document content for change detection", default="", index=True)
 
@@ -1895,6 +1895,7 @@ GAUSSDB_EMPTY_STRING_COMPATIBLE_COLUMNS = (
     ("dialog", ("llm_id", "rerank_id")),
     ("memory", ("embd_id", "llm_id")),
     ("file", ("source_type",)),
+    ("document", ("suffix",)),
     ("system_settings", ("value",)),
     ("task", ("task_type",)),
     ("sync_logs", ("error_msg", "full_exception_trace")),
@@ -2395,7 +2396,7 @@ def migrate_db():
     alter_db_add_column(migrator, "mcp_server", "variables", JSONField(null=True, help_text="MCP Server variables", default=dict))
     alter_db_rename_column(migrator, "task", "process_duation", "process_duration")
     alter_db_rename_column(migrator, "document", "process_duation", "process_duration")
-    alter_db_add_column(migrator, "document", "suffix", CharField(max_length=32, null=False, default="", help_text="The real file extension suffix", index=True))
+    alter_db_add_column(migrator, "document", "suffix", EmptyStringCharField(max_length=32, null=False, default="", help_text="The real file extension suffix", index=True))
     alter_db_add_column(migrator, "api_4_conversation", "errors", TextField(null=True, help_text="errors"))
     alter_db_add_column(migrator, "dialog", "meta_data_filter", JSONField(null=True, default={}))
     alter_db_column_type(migrator, "canvas_template", "title", JSONField(null=True, default=dict, help_text="Canvas title"))

test/unit_test/api/db/test_gaussdb_adaptation_points.py

@@ -75,6 +75,7 @@ def test_gaussdb_empty_string_compatible_migration_drops_not_null_only_for_gauss
     assert len(FakeDB.queries) == expected_query_count
     assert ('ALTER TABLE "user" ALTER COLUMN "nickname" DROP NOT NULL', None) in FakeDB.queries
     assert ('ALTER TABLE "tenant" ALTER COLUMN "llm_id" DROP NOT NULL', None) in FakeDB.queries
+    assert ('ALTER TABLE "document" ALTER COLUMN "suffix" DROP NOT NULL', None) in FakeDB.queries
     assert ('ALTER TABLE "system_settings" ALTER COLUMN "value" DROP NOT NULL', None) in FakeDB.queries
     assert ('ALTER TABLE "task" ALTER COLUMN "task_type" DROP NOT NULL', None) in FakeDB.queries
     assert ('ALTER TABLE "sync_logs" ALTER COLUMN "error_msg" DROP NOT NULL', None) in FakeDB.queries
@@ -88,15 +89,14 @@ def test_gaussdb_empty_string_compatible_migration_drops_not_null_only_for_gauss
     assert len(FakeDB.queries) == expected_query_count
 
 
-def test_gaussdb_migration_adds_compatible_tags_before_relaxing_columns(monkeypatch):
+def test_gaussdb_migration_adds_compatible_fields_before_relaxing_columns(monkeypatch):
     events = []
-    migrated_tags_field = None
+    migrated_fields = {}
 
     def record_add_column(_migrator, table_name, column_name, column_type):
-        nonlocal migrated_tags_field
         events.append(("add", table_name, column_name))
-        if (table_name, column_name) == ("user_canvas", "tags"):
-            migrated_tags_field = column_type
+        if (table_name, column_name) in {("document", "suffix"), ("user_canvas", "tags")}:
+            migrated_fields[(table_name, column_name)] = column_type
 
     monkeypatch.setattr(settings, "DATABASE_TYPE", "gaussdb")
     monkeypatch.setattr(db_models, "alter_db_add_column", record_add_column)
@@ -111,11 +111,22 @@ def test_gaussdb_migration_adds_compatible_tags_before_relaxing_columns(monkeypa
 
     db_models.migrate_db()
 
-    assert isinstance(migrated_tags_field, db_models.EmptyStringCharField)
-    assert migrated_tags_field.null is True
+    for field in migrated_fields.values():
+        assert isinstance(field, db_models.EmptyStringCharField)
+        assert field.null is True
+    assert set(migrated_fields) == {("document", "suffix"), ("user_canvas", "tags")}
     assert events[-1] == ("relax",)
 
 
+def test_empty_string_char_field_keeps_non_gaussdb_constraints(monkeypatch):
+    for database_type in ("mysql", "postgres", "oceanbase"):
+        monkeypatch.setattr(settings, "DATABASE_TYPE", database_type)
+        field = db_models.EmptyStringCharField(null=False)
+
+        assert field.null is False
+        assert field.db_value("") == ""
+
+
 def test_gaussdb_unique_email_migration_checks_unique_email_index_not_fixed_name(monkeypatch):
     class Cursor:
         def fetchone(self):

test/unit_test/api/db/test_gaussdb_user_flow.py

@@ -77,7 +77,7 @@ print("gaussdb-adapter-ok")
 def test_gaussdb_empty_string_compatible_fields_are_nullable_only_for_gaussdb():
     gaussdb_result = run_isolated_flow(
         """
-from api.db.db_models import API4Conversation, Dialog, File, Knowledgebase, Memory, SyncLogs, SystemSettings, Task, Tenant, User, UserCanvas
+from api.db.db_models import API4Conversation, Dialog, Document, File, Knowledgebase, Memory, SyncLogs, SystemSettings, Task, Tenant, User, UserCanvas
 
 fields = [
     User.nickname,
@@ -92,6 +92,7 @@ fields = [
     Dialog.rerank_id,
     Memory.embd_id,
     Memory.llm_id,
+    Document.suffix,
     SystemSettings.value,
     Task.task_type,
     Task.progress_msg,
@@ -135,7 +136,7 @@ print("gaussdb-empty-string-compatible-ok")
 
     mysql_result = run_isolated_flow(
         """
-from api.db.db_models import API4Conversation, Dialog, File, Knowledgebase, Memory, SyncLogs, SystemSettings, Task, Tenant, User, UserCanvas
+from api.db.db_models import API4Conversation, Dialog, Document, File, Knowledgebase, Memory, SyncLogs, SystemSettings, Task, Tenant, User, UserCanvas
 
 fields = [
     (User.nickname, False),
@@ -150,6 +151,7 @@ fields = [
     (Dialog.rerank_id, False),
     (Memory.embd_id, False),
     (Memory.llm_id, False),
+    (Document.suffix, False),
     (SystemSettings.value, False),
     (Task.task_type, False),
     (Task.progress_msg, True),