datapeek

Paste JSON or CSV, get a sortable table, per-column statistics and a chart. Everything runs in the browser; nothing is uploaded. Share a link and the data travels inside the URL fragment.

Try it: https://peroksai2-arch.github.io/datapeek/

What it does

• Accepts messy input. CSV, TSV, semicolon or pipe delimited (sniffed automatically), JSON arrays of objects, {"data": [...]} wrappers, objects of parallel arrays, arrays of arrays with a header row, and NDJSON. Nested objects are flattened to dotted keys. Quoted fields with embedded commas and newlines parse correctly.
• Infers column types. Numbers (including 1,234.5), ISO dates, booleans (yes/no, true/false) and strings. Blanks, NA, null and - become nulls instead of poisoning a numeric column.
• Table. Click a header to sort, type to filter across all columns, numeric columns right-aligned, 200 rows at a time with "show more".
• Stats. Per column: fill rate, distinct count, min/max, mean/median for numbers, top-5 values for everything else.
• Chart. Bar or line with group-by and sum/avg/count/min/max, or a scatter of two numeric columns. Long tails fold into "(other)".
• Share. The pasted text is deflated and base64url-encoded into the URL fragment. Fragments are never sent to servers, so the link is the only copy.

Why it exists

Looking at a JSON response or a CSV export usually means either squinting at raw text or firing up a spreadsheet and fighting its import dialog. This is the middle ground: one paste, immediate feedback, and a link you can drop in a chat to show a colleague the same view.

Development

npm install
npm run dev        # http://localhost:5173
npm test           # vitest, parsing/inference/stats/aggregation/share round-trips
npm run lint
npm run build

Parsing, type inference, statistics, aggregation and share-link encoding live in src/lib/ as plain TypeScript with no React dependency, and that is where the tests are. Components in src/components/ are thin.

Deploys to GitHub Pages on every push to main via .github/workflows/deploy.yml.

Limits

• The whole dataset lives in memory and the table renders rows on demand; a few hundred thousand rows is fine, tens of millions is not the goal.
• Share links grow with the data. Browsers handle very long URLs, but chat apps and some proxies do not; the UI warns past roughly 8 kB.
• Dates are parsed only in ISO 8601 form. 12/31/2024 stays a string on purpose, because 01/02/2024 is ambiguous.

License

MIT