Debugging silent pandas pipeline failures with dframe-trace

dframe-trace is a pip-installable package that requires only pandas (or Polars) as a dependency [Dev.to]. After calling autopatch.install(), the library patches common DataFrame methods—merge, astype, dropna, etc.—logging before/after snapshots of row count, column list, null totals, and dtype changes. The trace() context manager collects these snapshots, and helper methods like where_null_introduced(col) or where_rows_lost() return the exact step that introduced the anomaly [GitHub]. The tool also provides guard functions (assert_no_row_loss, assert_no_new_nulls) that can be added to CI pipelines to fail builds on regressions. The initial release includes an autopatch layer with under 1 ms per operation overhead, a report() command that prints a step-by-step diff, and guard utilities that translate recorded violations into structured build errors. dframe-trace fills the gap between proactive schema checks and reactive debugging, as it records everything first, then lets you query the trace after the fact. Because the library stores only structural metadata, it can remain enabled in development environments without noticeable performance penalties [Dev.to].