Shorten long filenames on import

[dstillman]

e.g. http://forums.zotero.org/discussion/23079

[dstillman]

Still an issue for non-snapshot attachments, and apparently long filenames fail silently.

http://forums.zotero.org/discussion/27122/

[aurimasv]

Here's a quick thought. Instead of using copyTo directly (in a lot of places), could we use Zotero.File.copyToUnique wherever we need to copy files? (similar idea for creating file) copyToUnique uses nsIFile.createUnique, which (A) creates unique file names, so that there's no chance of overwriting and (B) also shortens file names (without truncating the extension). This would certainly be better implemented than the truncateFileName function that I wrote.

We would need to modify copyToUnique to return either newName or newFile.append(newName) so that we can retrieve the new path. This should be safe, since copyToUnique does not appear to be used anywhere ATM.

Thoughts?

[dstillman]

If nsIFile.createUnique really shortens long filenames that would otherwise trigger errors, this sounds good to me.

Returning the new nsIFile sounds good. I don't know why that returns the original nsIFile now.

[dstillman]

@aurimasv: Good find on createUnique(). This seems to work perfectly.

[aurimasv]

I was going to test if the shortening works correctly when the supplied file path is URI encoded. I expect that it should work fine, but if it's not, then there's a small chance that an encoded character gets split. I'll check some time later today.