2006-03-18

Hi,

first of all, thanks for your work. I'd like to use oddmuse for an internet dictionary for Upper Sorbian and I'd want to keep all the files for one letter in one directory. Oddmuse works this way, but not for unicode characters like ž (they all go under "others/"). I have looked for an easy fix in GetPageDirectory?(), but I couldn't do it.

Also, I want to feed lots of pages (several ten thousand) in an automated way so I'd have to generate the *.pg pages and the index automagically and I want to be sure I won't break anything before I do that. Evidently, the wiki text goes into "text:" but there seems to be almost the same in "blocks:". The rest looks like it could be ignored. The pageidx file looks to me like the names of the pages followed by a number I don't quite understand.

Cheers

Edi

If you use text files and have the time, you could consider using a shellscript and Oddmuse:wikiput to upload the files.

As for the distribution of the files in GetPageDirectory – I'm personally not happy with how it currently works, either. Are you expecting to run into a problem? If you're really planning to run a dictionary, you might have to use a hashing function anyway, since some filesystems will break down above 30'000 files per directory or something like that. I haven't worked on such a system until now because 1. I don't need it, and 2. it would be an incompatible change. If you're going ahead with this, we could write a little module that overwrites GetPageDirectory($id) and returns a hash of the id, with an appropriate hashing algorithm that would generate about different 20'000 values. What do you think?

Maybe the Oddmuse wiki would be a better place to discuss this, however.

– Alex