Rln is a free, open-source, offline-capable data analysis tool for researchers.
A compact, command-driven language for cleaning, exploring, describing, visualizing, and modeling data — on your laptop or your phone. No per-seat licenses, no cloud, no account. Your data never leaves your device.
A free, modern alternative to costly proprietary statistics packages: descriptive statistics, cross-tabulation and correlation, the full regression family (OLS, logit/probit, Poisson, GLM), panel and difference-in-differences estimators, robust and clustered standard errors, post-estimation diagnostics, and publication-style charts — with a larger-than-RAM mode for datasets of millions of rows that don't fit in memory. Drive it from a text REPL, a desktop GUI, or the phone app — the same scripts run identically on all three.
To the scientists and researchers in our field, may powerful tools always be free and within reach. And to Marlene, the one who inspired me to build and help others 🌱
First public release — 14 June 2026.
- 📂 Read anything — CSV,
.dta, Excel (.xlsx/.xls), Parquet, JSON, and more - 🔎 Explore & describe —
describe,summarize,tabulate,correlate - 📈 Charts —
histogram,scatter,line,twoway, rendered right in the app - 🎨 Colour-coded explorer — type-aware data grid (numbers, text, missing) in every
version;
browse "file.parquet"previews any file without touching your dataset - 🚀 Larger-than-RAM (LRTM) — stream millions of rows with polars: 2,000,000 rows summarized in 0.04 s — on a phone (see the demo above)
- 📝 Reproducible scripts — multi-line scripts with syntax highlighting
- 🧮 Econometrics —
regress,logit,probit,poisson, GLM, panel (xtreg),didregress, with robust/clustered SEs and diagnostics — now on Android too, via a built-in NumPy/SciPy backend (ivregressand panel random-effects remain desktop-only) - 📱 Phone-first Android app — the full engine, native and offline, dark mode, pinch-free zoom, and an always-visible script editor
Built for social scientists who work with survey, panel, and administrative data but shouldn't have to pay per-seat for commercial software, stand up cloud infrastructure, or learn the entire Python ecosystem before running a regression. One small, memorable command language; the same scripts run identically on desktop and mobile.
- 🧮 Econometrics on mobile.
regress/logit/probit/poisson/ GLM,xtreg, fe,didregress(TWFE), VIF and the heteroskedasticity tests now run on Android through a built-in, statsmodels-equivalent NumPy/SciPy backend (validated to ~1e-11). If a native statsmodels is present it's still used; otherwise the backend takes over automatically, so estimation no longer hard-fails on a phone. - 🎨 Colour-coded data explorer everywhere. Numbers, text and missing values are
type-coloured across the terminal, desktop and Android grids.
browse "file.parquet"previews any file (csv/dta/xlsx/parquet/…) without disturbing the in-memory dataset. - 👀 Live Parquet preview (parquet-explorer style). After
lrtm use, the data browser immediately streams a colour-coded preview of the file — see what the data looks like before loading it — then explore the whole dataset once youlrtm collect. Works on desktop and Android, no pyarrow required. - 📁 Examples ready out of the box. Bundled sample datasets are located automatically:
use "demographics.csv"anddo "sample.do"just work, and the file pickers open in the right folder. Point Rln at your own data withset workdir "<path>"(remembered between sessions). - 🔗 Fuzzy merge in every build.
fuzzmergenow falls back polyfuzz → rapidfuzz → stdlib, so approximate string joins work on lite and Android, not just the full build. - 🩹 Robustness fixes. Larger-than-RAM
lrtm collect/sampleno longer require pyarrow; charts work in the lite desktop build (Pillow is now bundled); and a missing optional library reports the real reason instead of a misleading message.
use "examples/demographics.csv", clear
summarize age income n_children
tabulate gender
histogram age
* larger-than-RAM mode — millions of rows, streamed
lrtm use "eu_firms_panel.parquet"
lrtm summarize
| Explore & describe | Larger-than-RAM (2M rows) | Script editor |
|---|---|---|
 |
 |
 |
| Variables browser | Data browser | Plots |
 |
 |
 |
media/rln_demo.mp4
Desktop GUI — the same engine with a point-and-click front end:
- 🌐 Interactive docs & download page — editions,
install steps, demo video, capability matrix, and a searchable command reference. Served
from
index.html(enable GitHub Pages on themainbranch to publish it). - 📘 Lite Desktop & Mobile guide — a PDF walkthrough focused on the two everyday editions.
- 📚 Full reference manual — every command, the expression language, and appendices (PDF).
Download rln-<version>-arm64-v8a-debug.apk from the Releases page, copy it
to your phone, and install (you'll need to allow installs from unknown sources). On first
Open, grant All files access so Rln can read your data files.
Download the rln-lite build from Releases. Double-click Rln-GUI.bat
for the graphical app, or run rln-lite.exe for the command-line REPL.
git clone https://github.com/akirawisnu/Rln.git
cd Rln
pip install -r requirements.txt
python main.py --gui # GUI
python main.py # REPLThe Android APK is built with buildozer /
python-for-android — see android/ for the recipe set (numpy, scipy, pandas,
matplotlib, kivy, polars for LRTM, and a from-source statsmodels recipe).
MIT © 2026 Akirawisnu. Made with care, and meant to stay free.