Map CSV column names

As comma separated old=new pairs, for example "First Name=first_name,E-mail=email". Whitespace around each part is trimmed. Anything without an equals sign is refused with a message naming the offending pair, and a name that does not appear in the header gets "column(s) not found in the header" along with a printout of the real header.

By default they are kept, in place, under their existing names. Turn keepUnmapped off and every column not named in the mapping is dropped, which turns this tool into a way to select a subset and rename it in one pass. The result text says how many columns were kept out of the original count when you do that.

No. Only the header names change, and only the ones you named. Column order stays exactly as it was and every cell is copied across untouched. If you also need to reorder or insert a column, use Reorder / insert CSV columns.

The tool does not stop you, so you would end up with a duplicate header. That is worth avoiding, because tools further down the line reject it: converting such a file with CSV to Parquet is refused by name, since the Arrow based readers cannot open a file with duplicate field names.

Yes. The file is parsed with quote state tracked across the whole text, so a quoted cell holding a comma or a literal newline stays a single cell in a single record. The output is re-quoted the same way, so nothing is corrupted on the way out.

One CSV named after your file with -mapped.csv on the end, plus a line saying how many columns were renamed. Everything happens in your browser, in memory, with no upload. That memory is also the practical size limit, since the whole file and its rewritten copy are held at once.