csv-profiler

Profile and analyze CSV or other tabular data — column types, summary statistics, missing values, and anomalies. Use when the user needs to understand, clean, or sanity-check a dataset.

Turn a raw CSV (or TSV, or an existing DataFrame) into a clear profile of what the data actually contains.

Workflow

1. Load and detect. Read the file; detect delimiter, encoding, and header. Report row and column counts.

2. Classify columns. For each column, assign a type: numeric, datetime, categorical, boolean, or free text.

3. Summarize. For each column report:

   • Numeric: min, max, mean, median, missing count, distinct count.
   • Categorical: top values with counts, distinct count, missing count.
   • Datetime: min/max range and count of unparseable values.

4. Flag anomalies. Call out:

   • Missing or empty values, and all-null columns.
   • Mixed types within a single column.
   • Duplicate rows and near-duplicate keys.
   • Outliers beyond roughly 1.5×IQR.

5. Recommend. Suggest concrete cleaning steps (drop, fill, cast, dedupe) and note which columns look usable as identifiers or join keys.