On 18 September 2011 17:38, Seb35 seb35wikipedia@gmail.com wrote:

Hi,

While reading a bug pointed by the i18n triage https://bugzilla.wikimedia.org/show_bug.cgi?id=16026, Happy-melon says "A more general bug (which might well exist, I didn't check thoroughly) would be "ALL MediaWiki messages should be either plaintext or accept wikimarkup."

So I had the idea of censusing the wfMsg calls in the MW code and try to detect which ones accept wikimarkup or other dialects (it depends of the wfMsg function and of the parameters of that function). So I wrote yesterday a Python script to census the wfMsg calls in one side and the messages in another side. The latter is not very useful in fact but the former gives some interesting results.

I’ve done some statistics about the wfMsg calls on http://www.mediawiki.org/wiki/WfMsg_statistics. There are also the raw data and the link to the script (output to CSV).

You should also include ->addWikiMsg ->wrapWikiMsg and perhaps ->addWikiMsgArray since they are the third major group of message functions after wfMsg* and wfMessage.

In the most used messages is 'nbbytes', is that typo from nbytes?

We can see that wfMsgWikiHtml is the first of wfMsg* functions that is nicely getting phased out, just as we want to.

-Niklas

On 18/09/11 16:38, Seb35 wrote: <snip>

So I had the idea of censusing the wfMsg calls in the MW code and try to detect which ones accept wikimarkup or other dialects (it depends of the wfMsg function and of the parameters of that function). So I wrote yesterday a Python script to census the wfMsg calls in one side and the messages in another side.

We probably want to have that python script in /trunk/tools/code-utils/

http://svn.wikimedia.org/viewvc/mediawiki/trunk/tools/code-utils/