The GitHub Actions job "Lint PR" on texera.git/feat/dataguard-mvp has failed.
Run started by GitHub user eugenegujing (triggered by eugenegujing).

Head commit for run:
87c8744d100029c7cc964285ac61e638c5ab8a92 / eugenegujing 
<[email protected]>
fix(dataguard): skip no-op writes in standardize, same as replace_value

After iterative Apply rounds normalize a string column to its canonical
form (e.g., `region: "South"` everywhere), the inconsistent_label detector
can keep re-flagging the column whenever the LLM proposes a mapping that
includes the canonical entry itself (`{south: "South"}` against rows that
are already `"South"`). The standardize branch was incrementing
`rowsAffected` on every mapping-key hit regardless of whether
`mapping[v] === v`, so the frontend pushed a byte-identical CSV/JSONL to
LakeFS and got "No changes detected in dataset. Version creation aborted"
— the same convergence failure the round-yesterday `cellEquals` guard
fixed for replace_value.

Add the same guard to `case "standardize"`. `affected` increments only
when the cell genuinely changes.

trim_whitespace already guards (`trimmed !== v`); impute is safe by
construction in the normal path (only missing cells are visited); the
all-missing-column edge case is a known follow-up and not in scope here.

Report URL: https://github.com/apache/texera/actions/runs/25963547357

With regards,
GitHub Actions via GitBox

Reply via email to