columna ← AnKhIT

DATAFRAME LIBRARY

columna

From data to clarity.

columna is a typed DataFrame library for TypeScript. It processes columns on CPU, with narrower WebGPU and WASM paths, in Node.js and in the browser. Statistics are a separate entry point, columna/advanced.

One query. The same library runs in Node.js and in the browser.

import { DataFrame, col } from 'columna'

const out = await DataFrame.fromRows([
  { city: 'Berlin', age: 30, salary: 72000 },
  { city: 'Berlin', age: 22, salary: 48000 },
  { city: 'Paris', age: 41, salary: 91000 },
])
  .select('city', 'age', 'salary')
  .filter(col('age').gt(18))
  .groupBy('city')
  .agg({ salary: 'mean', age: 'count' })
  .sort(col('salary').desc())
  .collect()

npm install columna

DataFrame

A fluent API for select, filter, group, aggregate and sort. Queries can stay lazy until collect, and a run can be cancelled.

Engines

CPU is the main path. WebGPU and WASM cover narrower kernels and report what actually ran. There is no required native binary.

Files

CSV, JSON, Excel and Parquet, plus Apache Arrow IPC for DuckDB, Polars and pyarrow. SQL and Kafka drivers stay optional.

Statistics

columna/advanced is the statistics layer: hypothesis tests, ANOVA, regression, design of experiments, SPC, capability, reliability, time series and multivariate methods. The public docs describe about 250 procedures, checked against SciPy, NumPy and NIST references.

EXAMPLE

A studio for trying the library

Columna Studio is a browser workspace on top of the library: worksheets, statistical procedures, charts and a report you can print. It is here so the library can be tried. It is not a finished product.

The trial studio: a worksheet and statistical charts in the browser.

Trial use