Troubleshoot SQL
Este conteúdo não está disponível em sua língua ainda.
Table or column not found
Section titled “Table or column not found”Confirm the selected organization, connection, and schema. Inspect the catalog for the exact name and casing. For Lake, use the logical name shown in FastBI, such as uploads.orders, and wait for ingestion to complete.
Function not found
Section titled “Function not found”Check the connection’s dialect. In Lake, use SUM(CASE WHEN ... THEN 1 ELSE 0 END) for conditional counts, TRY_CAST for tolerant conversion, and COALESCE for null fallback. Do not paste BigQuery-specific functions into Lake unchanged.
Type mismatch
Section titled “Type mismatch”Inspect the column type and sample values. A numeric-looking imported field may be text. Use explicit casts and inspect conversion failures before aggregating. For string operations on numeric IDs, cast to VARCHAR first.
Unexpectedly high totals
Section titled “Unexpectedly high totals”Check the row granularity and join cardinality. Joining orders to several customer-history records or order lines can multiply rows. Aggregate at the intended level and use a unique dimension key.
Missing days or an empty result
Section titled “Missing days or an empty result”Check source freshness, ingestion status, and the date filter. Verify whether timestamps are stored in UTC and whether your question uses a different reporting timezone. A filter ending at midnight can exclude the rest of that day.
Query takes too long or scans too much
Section titled “Query takes too long or scans too much”Filter on a date or partition column, select fewer fields, and avoid unnecessary joins. A low LIMIT caps returned rows but may still require a large scan. If FastBI pauses for scan confirmation, review the estimate before continuing.
Read-only query rejected
Section titled “Read-only query rejected”Lake analysis accepts read-only queries. Use supported upload or ingestion workflows to change data instead of UPDATE, DELETE, or schema changes in an analysis query.
Workbench function rejected before execution
Section titled “Workbench function rejected before execution”The workbench uses an explicit function supported list plus SQL functions. A function supported by a connected database can still be unavailable through this workbench. Check workbench compatibility instead of repeatedly renaming the function.
Ambiguous table
Section titled “Ambiguous table”Qualify the table as connection.dataset.table using the exact Data explorer name. If several connections expose public.customers, writing only customers is not enough to select one.
Cross-source transfer is incomplete
Section titled “Cross-source transfer is incomplete”The workbench refuses to join a result that is known to be incomplete. Inspect source or proxy row limits and retry with source-side filters or aggregation in independent CTEs. A smaller final LIMIT alone may not reduce transfer. See cross-source examples.
Result expired or save destination exists
Section titled “Result expired or save destination exists”Temporary workbench results expire after 24 hours. Rerun an expired query before saving. Save to Lake creates a new private table; it does not overwrite an existing destination. Choose a new dataset/table name if the destination already exists.