# Pivot Shelf

> A drag-and-drop chart and pivot-table explorer that runs entirely in the browser. Drop a CSV, TSV, JSON, Excel (.xlsx) or Parquet file, or paste rows, and drag fields onto Columns, Rows, Colour, Size and Filters shelves. The page writes SQL from the shelves, runs it on an in-page SQL engine, and draws the result. Free: no AI model, no credits, no sign-in needed. Rows are never uploaded.

## What it does
- Reads CSV/TSV (delimiter sniffed), JSON arrays, Excel .xlsx (every visible sheet) and flat Parquet files (SNAPPY, GZIP or uncompressed; data pages v1 and v2; dictionary, plain, RLE and delta encodings; DATE, TIMESTAMP, DECIMAL and STRING types). ZSTD Parquet is refused with advice to re-save as snappy. Up to 300,000 rows.
- Types each column as number, date, timestamp, true/false or text. Numbers are measures; everything else is a dimension. A virtual "Row count" measure is always available.
- Shelves: Columns (X), Rows (Y), Colour, Size, Filters. Aggregations: sum, average, median, min, max, count, distinct count. Date bins: year, quarter, month, week, day, hour, month of year, weekday. Number bins of a chosen width.
- Chart forms: bar (vertical or horizontal; stacked, side by side or 100%), line, area, scatter/bubble, heatmap and pivot table with row, column and grand totals. "Auto" picks the form from the fields and says why; a hand-picked form that does not fit the fields falls back and says so.
- Shows every SQL query it ran (main query plus pivot totals).
- Calculated fields as SQL expressions (e.g. profit / revenue).
- Downloads: chart as PNG or SVG (title and legend included), result as CSV, view as JSON. A share link carries the view (not the data) as URL-safe base64 after #view=.
- Saved views: in the browser, or in the signed-in person's SkillSafe account (collection "views"). A view holds shelf layout only.
- Three made-up sample datasets: shop orders, city weather, home listings.

## Links
- App: https://pivot-shelf.skillsafe.ai/
- Saved-views API: https://pivot-shelf.skillsafe.ai/api.html