Dumping files with disallowed file names

[robkam]

This was being discussed in now closed #58, brought here as the topic itself.

The following characters are not allowed in NTFS file names:

< (less than)
> (greater than)
: (colon)
" (double quote)
/ (forward slash)
\ (backslash)
| (vertical bar or pipe)
? (question mark)
* (asterisk)

When dumping the images any of the above characters will be problematic.

Maybe use percent-encoding to get around this?

[yzqzss]

please DO NOT USE percent-encoding, which will increace the filename and make it cannot created by OS when the filename is very long.

#58 (comment):

MediaWiki determines which folder the file is stored in based on the md5 of the file name (this is the biggest problem, i.e. if we change the file name, the dumped image will be difficult to import losslessly by the MediaWiki Image Import CLI tool)

Also, most file system has max filename length limit, ext*/btrfs/zfs are 255 bytes, NTFS is 255 characters (Unicode).
(But ~240 is a safer number for Windows API(maybe))

[yzqzss]

Perhaps we should remind users that it is better not to use NTFS when doing mass archiving.
After all, MediaWiki has the best compatibility with Linux.

[yzqzss]

And most Linux servers are using ext4.

[robkam]

It would be better to have "Linux, Mac and Windows compatible".

[elsiehupp]

I don't know if it's possible to have Python determine the filesystem, but if so perhaps there could be special rules in place for filesystem-specific workarounds?

[elsiehupp]

Or, better, if the workarounds could be abstracted away by using different Python commands?

[robkam]

However the file is renamed also add a message in errors.log e.g. File:--*"?---------boom.png has been renamed so that it doesn't get lost.