inccsv 0.1.0

Python implementation of the INC file format (INI metadata + CSV data)

IncCSV.py

A Python implementation of the INC file format, interoperable with the original Julia library IncCSV.jl.

What is INC?

INC (INi-Csv) is a lightweight file format that embeds INI-style metadata directly inside a CSV file, separated by a --- delimiter:

---
title = Sensor readings
version = 1
[columns]
time = seconds
temperature = Celsius
---
time,temperature
0,21.4
1,21.8
2,22.1

The goal is to make tabular data self-describing — units, provenance, and other context travel with the file rather than being stored separately. Plain CSV files are read without modification, so the format is fully backward compatible.

See the Julia reference implementation for the full format specification and design rationale.

Installation

pip install inccsv

With pandas support:

pip install inccsv[pandas]

Usage

Reading

import inccsv

# Read an INC file
f = inccsv.read_inc("data.inc")
print(f.metadata)          # {'title': 'Sensor readings', 'version': 1, 'columns': {...}}
print(f.rows)              # [{'time': '0', 'temperature': '21.4'}, ...]

# Convert to a pandas DataFrame
df = f.to_dataframe()

# Plain CSV files work too
f = inccsv.read_inc("data.csv")

Writing

rows = [
    {"time": "0", "temperature": "21.4"},
    {"time": "1", "temperature": "21.8"},
]
metadata = {
    "title": "Sensor readings",
    "version": 1,
    "columns": {"time": "seconds", "temperature": "Celsius"},
}

inccsv.write_inc("data.inc", rows, metadata=metadata)

Schema validation

A schema is itself an INC file that declares required and optional metadata fields:

---
[schema]
allow_extra = false

[MUST]
title = String
version = Int

[OPTIONAL]
author = String
---

schema = inccsv.read_schema("schema.inc")
result = inccsv.validate_schema(f, schema)
print(result.valid)    # True / False
print(result.missing)  # required fields absent from the file
print(result.extra)    # fields not declared in the schema

Summary

s = inccsv.summarise(f)
print(s.n_rows, s.n_cols)
inccsv.print_summary(f)

Format notes

• Metadata values are int (unquoted integers) or str (everything else). Quote values that look like integers to preserve them as strings: id = "007".
• The [structure] section passes CSV options to the reader: delimiter, quotechar, comment.
• The delimiter line accepts any sequence of 3+ Unicode dash characters (-, –, —, …).
• Files are UTF-8 encoded.

Interoperability

This library aims to be fully interoperable with IncCSV.jl. Files written by either library should be readable by the other.

Development

git clone https://github.com/lewismath/IncCSV.py
cd IncCSV.py
pip install -e ".[dev]"
pytest