Build an interactive debugging environment that explains why a transformed dataset differs from its original source.
Users can upload source files, transformation scripts, SQL queries, mapping rules, and expected output examples. Represent each transformation as an inspectable stage, from filtering and type conversion to joins, aggregation, deduplication, and enrichment.
For every output value, allow users to trace its lineage back to the source records and transformation rules that produced it.
When discrepancies appear, generate competing explanations such as dropped rows, incorrect join keys, timezone conversion, null handling, rounding, duplicate aggregation, or mismatched units.
Provide a side-by-side comparison of expected and actual results, minimal reproducible examples, and proposed regression tests.
Never declare a transformation incorrect without examining the intended business rules. Preserve source data, display intermediate results, and clearly distinguish proven defects from plausible explanations.
0 Comments