Skip to main content

Import & export

Datasets round-trip through CSV and JSON, so you can edit data in bulk elsewhere and bring it back.

Plan availability

Starter and above (datasets need the data store).

The data page toolbar carries import and export

Export​

Export a dataset's records to CSV or JSON — useful for backups, bulk edits in a spreadsheet, or moving data into another tool. The CSV and JSON buttons on the data toolbar export the whole dataset, not the page of records currently on screen.

What you get​

  • Every record, in a stable order that is the same on every export of the same data.
  • Every field in the model, in model order — the CSV's header row is the field ids, and the JSON is an array of objects keyed by the same ids. A field with no value comes out empty rather than being omitted.
  • Values in their portable form, which is the form the importer reads back: dates as ISO-8601, coordinates as latitude, longitude, lists and references comma-joined, and map fields as JSON. That is what makes the round-trip lossless.

The file is named after the dataset. Nothing is filtered out — a search or sort you have applied in the table does not narrow the export.

Large datasets​

The server streams the file rather than assembling it in your browser, so the size of the dataset is not the limit it used to be. Two things follow:

  • The button reads Exporting… while it runs. A large dataset takes a while; leave the page open until the download starts.
  • The download is checked before it is saved. The server states how many records it is sending, and if fewer arrive — a dropped connection mid-transfer — the export is refused with "Export incomplete" and nothing is saved. Run it again. A short file is otherwise perfectly well-formed and gives you no way to notice it is short.

Records added while an export is running may or may not be included; the count is taken when the export starts.

Import​

Import CSV or JSON back into a dataset. Imports are validated against the model on the way in, so malformed rows are caught rather than silently corrupting your data.

Upsert on a key field​

Pick a "Match on field" in the import dialog to de-duplicate: rows whose key value (say, an email) matches an existing record update it in place instead of appending a duplicate, and repeated keys within the same file collapse to the last row. Only genuinely new rows count against your record limit.

Tips​

  • Export first to get the exact column shape, edit that file, then re-import.
  • Keep reference fields pointing at valid record ids so relations survive the round-trip.