NowConverter field lesson · Updated 19 October 2026 · Examples use our original test files.
Why a CSV import can fail
CSV represents a table as text, but applications differ in how they handle delimiters, quotes, blank values and text encoding. Splitting every line on a comma is not a valid CSV parser: a quoted field can legitimately contain a comma. A missing column or unexpected extra field can break strict imports.
Reproduce the common mistakes
- Download sample-data.csv and broken-rows.csv.
- Open both in a plain-text editor. Notice that
"Sample, Alpha"in the valid file is one field, even though it contains a comma. - Count columns: the valid file has four header columns and four values for each row (one may be blank). The broken example has three columns in the header, only two fields in the Beta row and four in the Gamma row.
- Try the valid file in CSV/JSON converter. Confirm that accented text, commas inside quotes and missing quantities survive the conversion. Test the broken example separately and inspect whether errors are flagged or data is shifted.
How to repair production data
First obtain a backup copy. Confirm whether the producer used commas, semicolons or tabs. Check the expected header and number of columns; fix mismatches in a text or spreadsheet editor. Ensure literal quote characters inside quoted fields are escaped appropriately, usually as doubled quotes. Save using an encoding accepted by the destination, commonly UTF-8.
What to check after repair
Verify the row count, columns and representative values rather than trusting a success banner. Search for non-ASCII characters, commas in names and empty cells. If the destination requires numeric columns, decide whether an empty quantity should remain null or be handled as an error; do not silently turn it into zero without business justification.
Not all CSV tools are equivalent
Some lightweight converters only support simple rows. Large files and very complex quoting may require a dedicated CSV library or spreadsheet application. Never assume our examples prove support for every dialect or encoding.