NocoDB CSV Import: Preserve Types and Record Identity

Prepare a Small, Awkward Sample

Create a working copy and include leading-zero identifiers, quoted commas, non-English characters, an empty date, a negative amount and a repeated business key. These are proposed validation cases. They expose interpretation problems that an ordinary first row can conceal.

Hypothetical example: equipment number 00127 is an identifier, while quantity 127 is a number. If both look identical after import, checking the row count will not reveal that the equipment key has changed. Decide which columns are text before converting them.

Use the Documented New-Table Import Path

The new-table guide supports CSV, Excel and JSON, with encoding and header choices. It states that imported fields default to single-line text and that the first field becomes the display value. Review the mapping before changing field types. Importing into a new table and updating an existing table are different jobs.

The import page lists a 5 MB maximum, while the environment reference exposes a configurable data-import size setting with a different default. Treat that discrepancy as a reason to verify the actual release and upload path. Do not publish one unconditional maximum for every self-hosted instance.

Documentation: Create a NocoDB table via import · NocoDB environment variables

Keep Technical Identity Separate from Labels

NocoDB distinguishes a primary key from its human-readable display value. Give your source business identifier its own retained column, so you can reconcile records after import. Do not rely on the visible row position or a repeated company name to identify a record.

Use the field-type conversion one column at a time. Compare the raw source value with the converted value before continuing. If a date is ambiguous, resolve its source format rather than guessing from the workstation’s locale.

1. Confirm the expected header and included columns.
2. Compare imported row count with the source data-row count.
3. Inspect all deliberately awkward sample values.
4. Check repeated business keys and missing identifiers.
5. Convert a copy of the relevant column and compare meaning.
6. Export a small verification sample and reconcile it by retained source ID.

Documentation: NocoDB field types, primary keys and display values

Accept the Result with a Written Exception List

Record exclusions, repaired values and rejected rows. Do not rerun an import into a live destination until you know whether the chosen operation appends or updates. A second attempt that doubles the records is not an encoding fix. Preserve the clean destination and source copy until the exception list has an owner and an accepted resolution.

Hypothetical CSV acceptance cases

This is an illustrative worksheet to adapt to your own evaluation. It is not a measured product test or a claim about results in your installation.

Import into a disposable destination first. Retain a source identifier and validate meaning before converting types or rerunning the job.