Pure-Go DataFrames, polars-style
golars

Eager + lazy execution. Optimiser, streaming engine, SIMD kernels.
Matches or beats polars 1.39 on most polars-compare workloads. No cgo.

Get Started →View on GitHub

Eager + Lazy

Build pipelines as logical plans. Optimiser fuses projections, pushes filters, collapses scans. Then Collect.

Streaming engine

Morsel-driven execution for datasets that don't fit in memory. stream.New(cfg, source, stages, sink).

polars-grade performance

SIMD kernels on amd64 via GOEXPERIMENT=simd. AVX2 VPBLENDVB blend, SIMD compare + arithmetic.

.glr scripting

One command per line. Drive the REPL, run from disk, embed from your own tools.

Rich CLI

golars run, fmt, lint, sql, schema, head, diff, browse, explain, completion.

LLM-native

MCP server exposes schema, head, describe, sql, diff to Claude / Cursor / Windsurf.

Quick Start

bash
# Add to your module
go get github.com/Gaurav-Gosain/golars@latest
bash
# Query any file
golars sql 'SELECT * FROM trades WHERE volume > 100' trades.csv

Essential commands:

golars schema→ column names + dtypes
golars stats→ describe() summary
golars browse→ TUI viewer
golars repl→ interactive .glr
MIT LicensePure Go (no cgo)Arrow-nativePolars-compatible