Back to Blog

Engineering

File importsValidationReliability

Record imports: from a file to trustworthy data

Sample parsing, full validation, confirmation, and the boundaries of retries and duplicate detection.

ISEA Team

Moving old records into a new app can look successful while dates are wrong, amounts are missing, or a retry creates a second copy of everything.

ISEA reads a sample, parses and validates the full file, then asks you to confirm the import. Each step has its own result, so you can see how far the process has reached.

Understand the format, then prepare the full result

Exports differ in column names, date formats, and units. Import first reads relevant content and parses a sample to check the proposed mapping. It then parses and validates the entire file.

The first few rows may look right while later rows use a different format. Full parsing checks each record's structure and the processing limits. If a limit is reached, the import should report it rather than silently omit data and claim success.

Confirmation belongs to a particular result

After full parsing, the system keeps the data ready to import, along with the file identity and parser version. The confirmation button refers to that checked result.

If you change the file or parse it again, the result needs to be prepared and confirmed again. The import writes the prepared data, keeping what you approved consistent with what gets saved.

Check dates, amounts, units, record types, and totals. If they are wrong, correct the mapping or source and prepare it again.

Writes and receipts belong together

Local imports handle duplicate detection, writes, and batch receipts in a transaction. The result reports actual inserted or skipped counts. Retrying with the same batch identity reuses its corresponding outcome.

Cloud imports also retain operation identities and receipts and wait for the relevant data confirmation. A network retry and a deliberately started new import are different events.

Cancellation at confirmation leaves that import unwritten. Failures identify whether reading, parsing, validation, or persistence failed, so the next step is clearer.

Duplicate detection has boundaries

Stable external record IDs can identify the same source record across imports. File identity and batch identity also help recognize retries along a supported path.

Without an external ID, switching the same file between local and cloud import does not currently guarantee file-fingerprint-level deduplication across those paths. Two identical-looking expenses may also be two real purchases.

If the result is uncertain, inspect import history and record counts before starting again. Repeatedly sending the same file through different entry points is not a reliable recovery method.

Check the records afterward

Pick a few familiar dates and compare counts, important fields, and totals. For large historical imports, retain the source file and existing backups.

Local and cloud imports can differ in file handling and capacity, so check the limits shown in the app. After completion, open the records and check their fields and counts against the source file.