More than one blank line at the end of CSV file: automate cleanup?

[aborruso]

Hi @johnkerl,
I sometimes run into errors like this::

mlr: mlr: CSV header/data length mismatch 2 != 1 at filename tmp.csv row 4.

This error occurs in many cases. Even when in CSV there are two blank lines at the end and not one.

a,b
1,2
4,6

I prefer to attach also a screenshot

For very large files, it's something I often miss, in the sense that I can't notice it and look for errors of other kinds.

Do you think it makes sense to introduce a more consistent error message, and/or automatically clean up the blank end lines and if there are more than one reduce them to one?

Thank you

[aborruso]

@johnkerl If you think it's not worth pursuing, I'll close it.

Thank you always

[johnkerl]

@aborruso thanks -- let's keep this open -- it's worth pursuing

[johnkerl]

@aborruso there are a few issues:

• Enabling non-compliant CSV by default is a slippery slope
• Even as opt-in behavior:
  • If the CSV has 2 or more columns then clearly the entirely empty lines can be ignored on an opt-in basis
  • But if the CSV has 1 column then it's not clear whether the blank line is
    • A row to be ignored
    • A completely legitimate row which has the empty-string value

My thought is to have an opt-in flag but with the caveat to the user that it will "eat" legitimate empty-string final rows in the one-column case ...

[aborruso]

But if the CSV has 1 column

This is the problem :(

The only thing I can think of is this: if the CSV has more than one blank line at the end, and we have for these end lines data length mismatch 2 (or more) != 1, keep only a blank line at the end.

It may be risky, though, and it's better to bump with the error and fix it.