fix(sql): restore SQL file schema for sidebar locate

This commit is contained in:
Melody Ruch authored and GitHub committed 2026-09-01 16:21:40 +08:00
1 parent 440c41f5b9
commit e9a90173a6
11 files changed
+175 -53

No files matched your search

@@ -427,6 +427,41 @@ describe("object tree node schema", () => {
expect(metadataSchemaForConnection(connection, resourcePath, resourcePath)).toBe("");
});
it("never uses the database name as the schema when a schema-tree engine has no schema yet", () => {
// Query tabs can reach locate/metadata paths before a schema is picked (a
// toolbar "new query" tab, or an external .sql file reopened from disk).
// PostgreSQL and its relatives keep tables under a separate schema level, so
// `schema || database` sends "dbx_test" as the schema and matches nothing —
// locate in the sidebar silently does nothing (issue #7648). The blank schema
// lets the backend resolve the session default instead.
for (const dbType of ["postgres", "gaussdb", "opengauss", "kingbase", "redshift", "duckdb"] as const) {
expect(connectionObjectTreeQuerySchema({ db_type: dbType }, "dbx_test")).toBe("");
expect(connectionObjectTreeNodeSchema({ db_type: dbType }, "dbx_test")).toBeUndefined();
expect(metadataSchemaForConnection({ db_type: dbType }, "dbx_test")).toBe("");
// An explicit schema still wins, and the database name is a legitimate
// schema name when the user really did select it.
expect(connectionObjectTreeQuerySchema({ db_type: dbType }, "dbx_test", "public")).toBe("public");
expect(connectionObjectTreeNodeSchema({ db_type: dbType }, "dbx_test", "public")).toBe("public");
}
// SQL Server keeps its own dbo default for metadata lookups.
expect(connectionObjectTreeQuerySchema({ db_type: "sqlserver" }, "dbx_test")).toBe("");
expect(connectionObjectTreeNodeSchema({ db_type: "sqlserver" }, "dbx_test")).toBeUndefined();
expect(metadataSchemaForConnection({ db_type: "sqlserver" }, "dbx_test")).toBe("dbo");
});
it("keeps the database-as-schema fallback for engines whose database is the schema", () => {
// Oracle, Dameng and the Hive family address objects as schema.table with no
// separate schema level in the tree, so the database name is the schema there
// and must survive the fix above.
expect(connectionObjectTreeNodeSchema({ db_type: "oracle" }, "DBX_TEST")).toBe("DBX_TEST");
expect(connectionObjectTreeQuerySchema({ db_type: "oracle" }, "DBX_TEST")).toBe("DBX_TEST");
expect(connectionObjectTreeNodeSchema({ db_type: "dameng" }, "DBX_TEST")).toBe("DBX_TEST");
expect(connectionObjectTreeNodeSchema({ db_type: "hive" }, "CS")).toBe("CS");
expect(connectionObjectTreeNodeSchema({ db_type: "spark" }, "CS")).toBe("CS");
// Flat engines keep returning no tree schema at all.
expect(connectionObjectTreeNodeSchema({ db_type: "mysql" }, "shop")).toBeUndefined();
});
it("keeps the database-as-schema fallback for flat engines and blanks it for Cloud Spanner", () => {
// Editor completion paths send the database name as the metadata schema when a
// connection has no schema level; only Spanner must send a blank schema instead.