Importing and exporting data

Alephgrid imports files as tables or sheets and exports them back out. Imports are streamed — a file is never read into memory whole — and reviewed before anything is created.

What you can import

Format	Becomes	Notes
CSV / TSV	Table	Delimiter and types auto-detected
Parquet	Table	Read by DuckDB’s native reader
JSON / JSONL	Table	One table per file
gzip	Table	Compressed CSV/TSV/JSON, decompressed on the fly
XLSX	Editable sheets	One sheet per worksheet, with formulas and formatting

Two ways to get a file in:

• Drag and drop. Drop a file onto an open workbook, or anywhere on the hub page.
• Command line. Pass the file when you launch:

alephgrid flights.parquet   # opens with the file loaded

The import review pane

Before an import creates anything, the file is staged and you review it in a side pane:

• Column types and samples. The detected columns with their types and the first rows, so you can see what the reader made of the file.
• Type overrides. Change a column’s type before importing. For CSV the override is applied inside DuckDB’s CSV reader, so text such as 007 stays exact instead of becoming 7; for Parquet and JSON it is applied as a cast.
• Progress and cancellation. The import runs as a job that reports progress and can be interrupted — only that job, nothing else.
• Errors you can fix. A value the reader cannot parse comes back as a structured problem (line, column, value, suggested type), so you can change the type and retry the same upload instead of starting over.

An XLSX file previews the sheets it would add and lists what will not come across: charts, pivots, macros, links, comments and unsupported functions.

What you can export

Format	Of what
CSV	Any table, or the current filtered view
Parquet	Any table, or the current filtered view
JSON lines	Any table, or the current filtered view
XLSX	Sheets, with a fidelity report

Exporting a filtered view writes the rows you see: hidden columns stay hidden and the raw values are kept, regardless of display formatting. Exports stream straight from a single snapshot of the data to the file — nothing is held in memory — so exporting a hundred-million-row table works the same as a small one.

The XLSX export comes with a report of anything XLSX cannot hold; today charts and some styles are reported rather than exported.

The SQL workspace can also export the complete result of a query — see the SQL workspace.

Where data lives on disk

Everything is plain files in one directory, ./alephgrid-data by default (change it with --data-dir or ALEPHGRID_DATA_DIR):

alephgrid-data/
  <workbook-id>/
    workbook.json   # name, created
    data.duckdb     # tables, sheets, formulas
  .trash/<id>/      # deleted workbooks, restorable

Each workbook is one DuckDB file holding its tables, sheets, formulas and saved queries. Deleting a workbook moves its directory to .trash/, where it stays restorable until purged. Back up the data directory and you have backed up everything.

Details of the on-disk format are in the repository’s architecture guide.

Next

• Analyze without SQL
• Big tables
• Install and first workbook