Synthetic CSV cleanup: explicit rules and a checked change log

Rule-based CSV cleanup — synthetic demonstration

Fictional data, not client work. AI-assisted preparation with checked arithmetic. This demonstrates explicit transformations, not customer results or production throughput.

Agreed rules: trim surrounding whitespace in ID and label fields; preserve ID text, case and leading zeros; remove only rows identical across all three fields after trimming, keeping the first occurrence; preserve negative quantities. Retain the original and a change log.

Input (quotes show whitespace): row 1: ID " 001 ", label " Widget A ", quantity 10. Row 2: "002", "Widget B", 7. Row 3: "002", "Widget B", 7. Row 4: "003", " Widget C", -2. Row 5: "0007", "Zero Seven ", 1.

Output: 001 / Widget A / 10; 002 / Widget B / 7; 003 / Widget C / -2; 0007 / Zero Seven / 1. Retained original rows: 1, 2, 4 and 5.

Verification: 5 input rows become 4. Row 3 is the only removal and exactly duplicates row 2 after normalization. Whitespace changes affect rows 1, 4 and 5. No quantity is edited. Original total 23 minus duplicate quantity 7 equals output total 16. ID 0007 and quantity -2 remain intact.

Without explicit permission to remove duplicates, the duplicate would be flagged rather than removed. No fuzzy merging, inferred IDs or silent corrections.