getpersonas stores all it's images in a similar method as we do on AMO: x/y/z/*.jpg where x y and z are numbers based on some ID. We can write a complicated migration script (boo) or just rsync the filesystem to AMO somewhere. This bug is to determine: 1) Can we just rsync somewhere? Can we recreate the x/y/z algorithm in AMO or would it be better to migrate them using AMO add-on IDs and mixing with our add-ons? 2) I think there are 9 files for each persona in each directory. Pretty sure I only care about the persona header and footer and none of the json or preview* files. (A quick look says these are named arbitrarily based on whatever the author uploaded including jpg vs JPG...omg) 3) Where should we put these on AMO? A new directory/setting I suppose. The personas FS can be seen on khan:/mnt/personas_static/
If it helps... The x/y/z algorithm I pulled from one of cvan's old patches and is pretty simple... I have it in a patch for the persona update pings but it would make sense to have this in a central place, but here it is: def url_prefix(self, id): """Uses last 2 digits to add extra directories. For example: url_prefix(1234) => '4/3/1234' """ return '%s/%s/%s' % (id % 10, (id // 10) % 10, id)
This seems like a big blocker to me.