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
{{ message }}
Repository navigation
Commit a4423dc
Browse filesBrowse the repository at this point in the historyBrowse files
[opt](column) reuse nullable column metadata during function execution (apache#66031)
The default nullable function path independently inspected the same null
maps when checking for all-NULL arguments, unnesting nullable inputs,
and wrapping function results. This caused redundant full null-map scans
and made type-level nullable checks difficult to distinguish from
runtime NULL-value checks.
This change introduces `NullableColumnInfo` to collect the nested
column, typed null-map column, constness, and non-NULL count once per
input column. The information is reused across nullable unnesting, CAST
handling, and result wrapping. It also renames the type-level helper to
`has_nullable_argument_type` and preserves the existing copy-on-write
behavior by copying nested data only when NULL payloads must be replaced
with defaults.
0 commit comments