xlsx mime type

[garak]

I'm creating an xlsx file (e.g. using Type::XLSX).
I expect to get a file with mime type application/vnd.openxmlformats-officedocument.spreadsheetml.sheet, instead I'm getting one with application/zip.

[adrilo]

Hmmm this is weird. Are you calling the openToBrowser() function?

[garak]

No, I'm just using openToFile and close to create the file.
I checked the mime type using file -b --mime mygeneratedfile.xlsx

[adrilo]

Oh I see. Then it makes sense, as an XLSX is simply a ZIP file with the "xlsx" extension.
Mime types only make sense from a browser perspective. In a filesystem, the OS tries to guess. Even when you create a file with Excel, the mime type is application/zip, so I guess there is not much I can do.

[garak]

My use case is a generated xlsx file, that I have to serve afterwards.
I'd like to rely on the mime type took from file, instead of forcing it with an HTTP header.
There must be a way, since running the same command (file -b --mime) on other xslx files gives me the correct type.

[adrilo]

How did you create the other xlsx files? Excel? Which version of Excel?
With Excel 2010 on Mac/Windows, file -b --mime returns application/zip.

[garak]

I know that application/zip is also a valid mime type for xlsx, anyway we know that a generic user would expect an excel file to be served/opened as a spreadsheet, not like an archive.
Unfortunately, I don't own any copy of Excel to try, anyway you can find many sample files online (like this one), all with the "correct" mime type.

[adrilo]

Hmmm even with the file you sent me I have:

$ file -b --mime ..../1.xlsx
application/zip; charset=binary

Looks like detecting the mime type is environment specific.

[garak]

Of course it is, it depends on the configuration of your magic.
The point is that library is generating files with a magic number that is not the same as the "official", so even with the correct magic configurated (e.g. on a standard Ubuntu), you get a generic mime instead of a specific one.

[garak]

Here is a different try. Write a script called fileinfo.php:

<?php
$finfo = new \finfo(FILEINFO_MIME_TYPE);
echo $finfo->file($argv[1]).PHP_EOL;

With a such file, php fileinfo.php 1.xlsx is returning application/vnd.openxmlformats-officedocument.spreadsheetml.sheet, while php fileinfo.php file_generated_with_spout.xlsx is returning application/zip

[adrilo]

Cool! I can reproduce the issue :)

$ php file.php spout-test/test-numeric2.xlsx
application/octet-stream

I don't have application/zip but application/octet-stream... Anyway, I'll try to take a look at it to figure out how the mime type is being set

[adrilo]

Hey @garak,

After pulling my hair out for a few hours, I finally figured out why the mime type is wrong. It turns out the order in which the XML files gets added to the final ZIP file (an XLSX file being a ZIP file with the xlsx extension) matters for the heuristics used to detect types.

Currently, files are added in this order:

[Content_Types].xml
_rels/.rels
docProps/app.xml
docProps/core.xml
xl/_rels/workbook.xml.rels
xl/sharedStrings.xml
xl/styles.xml
xl/workbook.xml
xl/worksheets/sheet1.xml

The problem comes from inserting the "docProps" related files. It seems like the heuristic is to look at the first few bytes and check if it finds Content_Types and xl. By having the "docProps" files inserted in between, the first xl occurrence must happen outside of the first bytes the algorithm looks at and therefore concludes it's a simple zip file.

I'll try to fix this nicely

[garak]

Great! Hope to see a new tag released soon.
Thank you.