How do I rename a file after is has been uploaded and saved? My problem is that I need to parse information about the files automatically in order to come up with the file
I'd like to donate my "safe move" solution that doesn't rely on any private API and protects against data loss due to network failure:
First, we get the old and new paths for every style:
styles = file.styles.keys+[:original]
old_style2key = Hash[ styles.collect{|s| [s,file.path(s).sub(%r{\A/},'')]} ]
self.file_file_name = new_filename
new_style2key = Hash[ styles.collect{|s| [s,file.path(s).sub(%r{\A/},'')]} ]
Then, we copy every file to it's new path. Since the default path includes both object ID and filename, this can never collide with the path for a different file. But this will fail if we try to rename without changing the name:
styles.each do |style|
raise "same key" if old_style2key[style] == new_style2key[style]
file.s3_bucket.objects[old_style2key[style]].copy_to(new_style2key[style])
end
Now we apply the updated model to the DB:
save!
It is important to do this after we create the new S3 objects but before we delete the old S3 objects. Most of the other solutions in this thread can lead to a loss of data if the database update fails (e.g. network split with bad timing), because then the file would be at a new S3 location but the DB still points to the old location. That's why my solution doesn't delete the old S3 objects until after the DB update succeeded:
styles.each do |style|
file.s3_bucket.objects[old_style2key[style]].delete
end
Just like with the copy, there's no chance that we accidentally delete another database object's data, because the object ID is included in the path. So unless you rename the same database object A->B and B->A at the same time (e.g. 2 threads), this delete will always be safe.