Dtdiff

dtdiff — Tabular Data Comparison

Compare two tabular data files of the same format and report added, removed, and modified rows.

Quick Reference

# Compare two CSV files (positional mode)
dtdiff old.csv new.csv

# Compare by key column
dtdiff old.parquet new.parquet --key ID

# Composite key
dtdiff old.csv new.csv --key Date,Ticker

# Float tolerance (differences ` | auto | Override format detection (applies to both files) |
| `--sheet ` | first | Select sheet (Excel only) |
| `--key ` | *(none)* | Key columns for matching (comma-separated names) |
| `--tolerance ` | `1e-10` | Float comparison tolerance |
| `--json` | false | Output as JSON |
| `--csv` | false | Output as CSV |
| `--no-color` | false | Disable ANSI colors |

## Exit Codes (diff convention)

| Code | Meaning |
|------|---------|
| 0 | No differences |
| 1 | Differences found |
| 2 | Error |

## Workflow

1. Identify the two files to compare (must be same format)
2. Use `--key` if rows have a natural identity column (ID, date, etc.)
3. Use `--tolerance` for financial/scientific data with float rounding
4. Use `--json` for structured output, default text for quick review

## Common Patterns

**What changed between versions:**
```bash
dtdiff old.csv new.csv --key ID

Ignore small rounding differences:

dtdiff old.parquet new.parquet --key ID --tolerance 0.001

Pipe structured results for processing:

dtdiff a.csv b.csv --key ID --json | jq '.modified'

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: LouLouLibs
• Source: LouLouLibs/claude-skills
• License: MIT

Install and usage instructions live in the source repository linked above.