// the take
VisiData is a terminal TUI for exploring and manipulating tabular data: CSV, TSV, JSON, SQLite, Excel, HDF5, Parquet, shapefiles, pcap captures, and dozens more. It's for anyone who wants to slice, pivot, or clean a dataset from a shell without firing up a notebook or spreadsheet app.
The loader ecosystem is genuinely broad — sample_data alone covers parquet, arrow, hdf5, sqlite, shapefiles, SPSS .sav, Stata .dta, pcap, and more, each with its own loader rather than a generic adapter. It has a real plugin API (loaders.rst, plugins.rst, extensible.rst) so you can add a format or command instead of forking. The test suite uses golden-file comparisons plus a replay format (.vdx/.vd) that doubles as a macro/scripting mechanism, so repeatable terminal workflows are a first-class feature, not a hack.
It's a dense TUI with hundreds of keybindings — the README itself points you to a cheat sheet rather than claiming it's learnable by exploration, so the learning curve is real. It's pure Python loading data into memory, and the project maintains a dedicated PERFORMANCE.md and a benchmark design doc, which tells you large datasets need explicit tuning rather than just working. Format support is split across optional extras (openpyxl, lxml, etc.), so a plain pip/pipx install silently lacks Excel or certain XML support until you remember to inject the right package — the README even has a canned troubleshooting answer for 'package X not installed' because this trips people up. GPLv3 is also worth knowing before embedding it in anything commercial.
We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →