1b788732da
The query processing was implicitly relying on the order provided by the catalog. This had two issues: - this ordering was not defined in the API contract (neither via docs nor via typing) - the order was based on chunk IDs which is not adequate in some cases (e.g. when chunks are created while a persistence operations is in progress) Now we explicitly sort chunks by `(order, ID)`. Fixes #1963. |
||
|---|---|---|
| .. | ||
| all_chunks_dropped.expected | ||
| all_chunks_dropped.sql | ||
| chunk_order.expected | ||
| chunk_order.sql | ||
| duplicates.expected | ||
| duplicates.sql | ||
| pushdown.expected | ||
| pushdown.sql | ||
| stats_plans.expected | ||
| stats_plans.sql | ||