Import

From a spreadsheet to a typed table - and updating it later.

Upload → analyze → preview → import. The target is either a new entity (admins only) or an existing one. When importing into an existing entity, you can choose a key field: matching records are then updated instead of created. Invalid rows are reported with their line number and skipped, and all other rows are imported.

All values go through one converter, so the import, the admin app and the API accept the same formats. Examples: 12,50 and 1.234,56 for decimals, 29.10.2025 for dates, ja/nein for yes/no.

References: a column can point to records of another entity, for example Country to countries:

  • The analysis offers the entities whose records the values name: for each entity, how many sample values were found (Countries (Name): 48 of 50). With 90 % or more, the reference is preset. One click links the column.
  • Values are found by a field of the target (match, default id), in every language of the field and regardless of upper and lower case: deutschland finds the country whose German name is Deutschland.
  • Add missing ones (create_missing: true): values that are not found are added to the target, so every row is imported. This needs a match field (not the id), permission to create records there, and no other required fields in the target.
  • New entity from the values (new_reference: {"slug": "genres", "name": "Genres"}, administrators): the column becomes an entity of its own with the field Name and one record per value (Crime and crime are one). Importing books creates the genres right away. With several values per cell (repeatable: true), Crime | Fantasy gives two genres and a list reference.
  • The preview shows what will be created in addition, and nothing is written yet. If the import fails, a new entity is removed again.

Yes/no columns: the analysis lists the different values of the column, and the import lets you choose which of them mean yes and which no (several each, e.g. 1, true and x for yes) and whether an empty cell means no. Other values are reported as errors. Without a choice, the usual words count (ja/nein, yes/no, 1/0, true/false …). In the plan: true_values, false_values and empty_false per column.