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 b764040
Browse filesBrowse the repository at this point in the historyBrowse files
fix(engine): bound table gap analysis memory, analyse expression columns (#549)
* fix(engine): bound table gap analysis memory, analyse expression columns
Missing-case detection built the full atom/row partition for every column
before charging the work budget. Nested ranges (`> 1`, `> 2`, ...) make that
partition quadratic, so large tables (tens of thousands of rows) allocated
gigabytes before giving up, exhausting the 1GB WASM memory. The partition
size is now counted first and the budget checked before building it; results
are unchanged (50k rows: ~10GB/28s -> ~650MB/1s natively).
Columns without a field whose filled cells are all conditions on one field
(`age > 50`, `age > 30`) are now analysed like field columns on that field:
duplicates, unreachable rows, missing cases and compression. Compressed and
missing-case cells are written back as full expressions (`age > 30`).
Columns mixing fields or using non-analysable expressions stay opaque.
* chore(engine): remove comments from table analysis changes
0 commit comments