Every row accounted for.

Source: messy_customers.csv. The source file was read, never edited.

8 input rows = 5 ready + 1 duplicates + 2 for review.

Files to use

cleaned.csv

Rows with the expected number of columns. Values stay text, including IDs with leading zeroes.

review.csv

Rows with missing or extra cells, with their original values and source record number.

audit.jsonl

One event for each changed, removed or quarantined record.

summary.json

Counts, settings and the source fingerprint.

What was applied

Trim outer whitespace: yes. Remove exact duplicate rows: yes. Cells changed: 3.

Duplicates use every cell, after optional trimming. The first matching record is kept. No names, dates, numbers or currencies were inferred.

View the source and quick start on GitHub