csvsmith

Small, focused CSV utilities for common data wrangling tasks.

csvsmith provides a handful of practical tools for working with CSV files, including cleaning numeric values, filtering rows, deduplicating records, classifying files, converting Excel spreadsheets to CSV, moving files by suffix, and finding matches inside CSV content.

Documentation

Read the full documentation at:

https://csvsmith.readthedocs.io/en/latest/

Features

• Clean numeric strings into normalized values
• Filter CSV rows by substring matching
• Deduplicate row data and generate reports
• Classify CSV files into folders based on headers/signatures
• Convert Excel workbooks to CSV
• Move files by suffix
• Find matching values inside CSV files
• Concatenate CSV files with identical headers
• Tokenize repeated CSV values and restore them with a versioned map
• Use the tools either from Python or from the command line

Installation

Install the package in your environment as usual for your project setup.

Example:

pip install csvsmith

If you are developing locally, install it in editable mode from the project root:

pip install -e .

Quick start

You can use the library from Python:

from csvsmith.utils.clean_numeric import clean_currency_numeric

print(clean_currency_numeric("$1,234.56"))

For command-line usage, use single quotes around values containing $:

csvsmith --help

Command-line usage

The package provides a CLI with several subcommands.

Clean numeric values:

csvsmith clean-numeric "1,234.56" --sep "," --decimal "."

Clean currency-prefixed numeric values:

csvsmith clean-currency-numeric '$1,234.56' --sep "," --decimal "."

Note

Use single quotes for values containing $. Double quotes may trigger shell expansion and change the input unexpectedly.

Filter rows in a CSV:

csvsmith drop-rows input.csv notes spam --case-insensitive --drop-header

Deduplicate rows:

csvsmith dedupe input.csv -o out.csv --subset id --keep first

Classify CSV files:

csvsmith classify src_dir dst_dir --mode relaxed --match subset --auto --dry-run

Convert Excel to CSV:

csvsmith excel-to-csv input.xlsx

Move files by suffix:

csvsmith move-files src_dir dst_dir --suffixes .csv,.pdf

Find matches in a CSV:

csvsmith find-matches input.csv target --ignore-case --ignore-whitespace

Concatenate CSV files:

csvsmith strict-concat incoming/ -o combined.csv

Concentrate repeated values and restore them later:

csvsmith concentrate input.csv
csvsmith rehydrate input.dense.csv -m input.dense-map.json -o restored.csv

Dense CSV scope and performance

The dense CSV workflow is intended for spreadsheet-oriented and medium-sized CSV files, roughly in the 100 MB class. Concentration makes two passes over the input and keeps value counts in memory, so memory use depends on the number and size of distinct values in the selected columns.

For substantially larger datasets, especially gigabyte-scale pipelines, a flat CSV plus a separate JSON map is usually the wrong interchange format. Consider a binary columnar format such as Apache Parquet instead.

Dense CSV replaces repeated values with tokens containing a 64-character SHA-256 digest. The complete token includes the csvsmith:sha256: prefix and is therefore 79 ASCII characters. The JSON map also stores each original value.

The workflow can still reduce expensive downstream work even when the files become larger. A consumer can process each mapped value once, store the result against its token, and apply that result to every repeated occurrence during rehydration. The CLI reports this potential repeated-operation reduction among mapped cells.

For example, 937 mapped cell occurrences backed by 82 unique map values imply 855 avoidable repeated operations, or about 91.2%. This is a deduplication ratio for mapped work, not a file-compression ratio or a guarantee of total pipeline savings.

Warning

Concentration does not guarantee a smaller combined output. Replacing short values such as True, M, or NY with a 79-character token will increase the CSV size. Select columns containing sufficiently long, repeated values and consider both the concentrated CSV and its JSON map when evaluating storage savings.

Find matches in a CSV

find_matches_in_csv searches a CSV file for a target value and returns match records containing coordinates and row context information.

Python API:

from csvsmith import find_matches_in_csv

results = find_matches_in_csv("input.csv", "target")

CLI:

csvsmith find-matches input.csv target

Options:

• --ignore-case: ignore case while matching
• --ignore-whitespace: ignore whitespace while matching
• --no-nfkc: disable NFKC normalization

If matches are found, the CLI prints formatted JSON. If no matches are found, it prints a simple message.

Other Python APIs

The package also exposes a few other helper functions and classes from its top-level API.

Numeric and row tools:

from csvsmith import (
    clean_numeric,
    count_duplicates_sorted,
    add_row_digest,
    find_duplicate_rows,
    dedupe_with_report,
    read_csv_rows,
    write_csv_rows,
)

CSV classification and filtering:

from csvsmith import CSVClassifier, DropRowsBySubstring

File and conversion helpers:

from csvsmith import (
    concentrate_csv,
    excel_to_csv,
    move_by_suffix,
    rehydrate_csv,
    save_csv,
    strict_concat_rows,
)

String comparison utilities:

from csvsmith import StringDistance, Relation, Result, analyze_pair

Project structure

The code is organized into two main areas:

• csvsmith.tools for higher-level CSV workflows
• csvsmith.utils for reusable utility helpers

Testing

Run the test suite with your preferred Python test runner.

Example:

pytest

License

See the project license for details.