mimetype mismatch (PHP 8.3+) and missing mimetype (PHP 7.4+) for "empty" files (CsvReader)

[DanielRuf]

https://github.com/PHPOffice/PhpSpreadsheet/blob/master/src/PhpSpreadsheet/Reader/Csv.php#L607 only lists inode/x-empty

But newer PHP and libmagic versions use application/x-empty.

So application/x-empty has to be added.

See also:

• Change getimagesize() and friends to report image/bmp php/php-src#2670 (comment)
• https://3v4l.org/l8FBt
• https://3v4l.org/8OBkG

At the last link you can also see, that previously an empty file with a crlf (\r\n) resulted in application/octet-stream and results in text/plain for newer PHP versions. Maybe this can be also be covered in PhpSpreadsheet (e.g. by using trim and checking the content before doing the mimetype check)?

The ideal way would be, to treat "empty" files (which should proably also include files with only whitespace characters) as such (application/x-empty) instead of application/octet-stream by PHP itself - or the mimetype text/plain for files with \r\n (to align the behavior to PHP 8.3+) .

[DanielRuf]

[oleibman]

Thank you for the research. I will definitely fix the application/x-empty omission. I am not so sure I want to deal with the other part - it's a lot of code and tests and overheads for what doesn't seem to be a "real" problem (and which can be avoided by just adding an extension to the filename, or using Php8.3+). I'll leave you some time to convince me that it's worthwhile anyhow.

[DanielRuf]

which can be avoided by just adding an extension to the filename

I'm not sure if I miss something. In our code we have a .xlsx file with just \r\n as content. Does your answer imply, that we forgot to add something somewhere?

After application/x-empty is added, our unit tests for an invalid xlsx file (with just \r\n as content) should at least work again by replacing it with a file with no content at all (by simply using touch() for a new empty file).

I'm not even sure if a xlsx file on our side with \r\n is a good testcase for "invalid" xlsx files - since there are much more possible ways to create invalid xlsx files.

So for me the following also works:

• application/x-empty is added to the list of mimetypes
• in our code we add a check for "empty" (and thus "invalid") files (using trim() on the content and checking the length of it) before using the PhpSpreadsheet functions

[DanielRuf]

I am not so sure I want to deal with the other part - it's a lot of code and tests and overheads for what doesn't seem to be a "real" problem

I agree with you.

[oleibman]

No, you're not missing anything. I was just noting that, if your empty-crlf file name used a csv extension, PhpSpreadsheet would handle it okay.

Excel itself will not open a file with an xlsx extension which is empty or contains cr-lf or variants. It will open such files with a csv extension. PhpSpreadsheet will, and should, emulate Excel in this respect. So, I assume that what you want to work is code like the following (please suggest any changes you feel are necessary):

$spreadsheet = IOFactory::load($filename);

Adding application\x-empty will cause PhpSpreadsheet to treat an empty file as an empty csv, which is what you want. It will not fix your cr-lf 8.1/8.2 problem, but, as discussed, I'm not convinced that's worth fixing.