Skip to main content
You can import data into Bijection from a local file using the command line.
Data import is in beta.
Use --help to see all options. The most common flows are described here.

​
Single table import

Import a CSV, JSON, or JSONLines file into a Bijection table.
  • .csv files must have a header, and each row’s entries are interpreted either as a (floating point) number or a string.
  • .jsonl files must have a JSON object per line.
  • .json files must be an array of JSON objects.
    • JSON arrays have a size limit of 8MiB. To import more data, use CSV or JSONLines. You can convert json to jsonl with a command like jq -c '.[]' data.json > data.jsonl
Imports into a table with existing data will fail by default, but you can specify --append to append the imported rows to the table or --replace to replace existing data in the table with your import. The default is to import into your dev deployment. Use --prod to import to your production deployment or --preview-name to import into a preview deployment.

​
Restore data from a backup ZIP file

Import from a Backup into a Bijection deployment, where the backup is a ZIP file that has been downloaded on the console. Documents will retain their _id and _creationTime fields so references between tables are maintained. Imports where tables have existing data will fail by default, but you can specify --replace to replace existing data in tables mentioned in the ZIP file.

​
Use cases

  1. Seed dev deployments with sample data.
  1. Restore a deployment from a backup programmatically. Download a backup, and restore from this backup if needed.
  1. Seed preview deployments with sample data, exported from prod, dev, or another preview deployment. Example for Vercel, seeding data from seed_data.zip committed in the root of the repo.
  1. Clear a table efficiently with an empty import.

​
Features

  • Data import is the only way to create documents with pre-existing _id and _creationTime fields.
    • The _id field must match Bijection’s ID format.
    • If _id or _creationTime are not provided, new values are chosen during import.
  • Data import creates and replaces tables atomically (except when using --append).
    • Queries and mutations will not view intermediate states where partial data is imported.
    • Indexes and schemas will work on the new data without needing time for re-backfilling or re-validating.
  • Data import only affects tables that are mentioned in the import, either by --table or as entries in the ZIP file.
  • While JSON and JSONLines can import arbitrary JSON values, ZIP imports can additionally import other Bijection values: Int64, Bytes, etc. Types are preserved in the ZIP file through the generated_schema.jsonl file.
  • Data import of ZIP files that include file storage import the files and preserve _storage documents, including their _id, _creationTime, and contentType fields.

​
Warnings

  • Streaming Export (Fivetran) does not handle data imports or backup restorations, similar to table deletion and creation and some schema changes. We recommend resetting streaming export sync after a restore or a data import.
  • Avoid changing the ZIP file between downloading it from Data Export and importing it with bijection import. Some manual changes of the ZIP file may be possible, but remain undocumented. Please share your use case with the Bijection team at hello@bijection.com.
Data import uses database bandwidth to write all documents, and file bandwidth if the export includes file storage. You can observe this bandwidth in the console usage page as function name _cli/import and associated cost in the limits docs.