You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
codebase-memory-mcp reports a best-effort indexing gap for project Users-gautamparab-Downloads-wisdomai_cs_large. index_status (diagnostics: full) lists sql/wisdom_cs/dim_domain_crm.sql under parse_partial with error ranges 51-52,54-54 (1-based). The rest of sql/wisdom_cs is not flagged. Status was ready (15,688 nodes / 49,797 edges).
The flagged lines are a BigQuery QUALIFY filter on a window function. Lines 51–52 are QUALIFY ROW_NUMBER() OVER ( and the PARTITION BY line. Line 54 is the closing ) = 1. The ORDER BY ... DESC NULLS LAST line between them (line 53) is not in the error ranges, so the parser recovers the inner order-by and fails again on the comparison after the window.
QUALIFY is valid BigQuery (and Snowflake) SQL. It is not in the core SQL grammar, so the tree-sitter parse records error nodes there. The file is still indexed; symbols inside those ranges may be missing from the graph.
Reduced stand-in for the flagged construct (generic names, not the warehouse view):
SELECT id
FROM items
QUALIFY ROW_NUMBER() OVER (
PARTITION BY id
ORDER BY updated_at DESC NULLS LAST, id DESC
) =1;
Environment: codebase-memory-mcp dev on macOS Apple Silicon (Darwin 25.6.0).
Activity