Summary

Advanced classification features of the MediaWiki software, and ways to leverage it to understand the content in this site.

Discussion

I've installed the Semantic MediaWiki extension. Play around with it. Figure out what can be accomplished. Get stuff better-organized around here. Things I expect will be possible

Turn Task pages into reports using SMW
SMW should be able to handle the kind of queries that power the "Unimplemented in X" pages
Be able to create pages for "examples of X in language Y"
Be able to treat things like PureBasic and GAMBAS as supersets of BASIC, etc. --Michael Mol 21:34, 26 July 2010 (UTC)

See Hough transform and Talk:Hough transform; the content on that page does a decent job of demonstrating a lot of the existing semantic structure on the wiki, so it's our guinea pig.

I also strongly recommend reading these pages:

--Michael Mol 11:20, 27 July 2010 (UTC)

What's up with this page?: Special:SemanticStatistics. Also, it looks like it's getting some things confused with library names that contain "::". --Mwn3d 19:22, 28 July 2010 (UTC)

Ah, yes. It'll pick up an any link that looks like [[blah1::blah2]]; the double-colon is the separator between property and value. Prefix a : to clear it up, same way you have to prefix a colon to make a category link. (i.e. [[:Category:Programming Tasks]]) --Michael Mol 19:28, 28 July 2010 (UTC)

Ok, here's the general scoop. As far as Rosetta Code is concerned, Semantic MediaWiki doesn't do a whole lot for site structure that we weren't already massively applying categories for already. The biggest thing it will allow us to do is break code examples into individual pages, and be able to build page listings (like Task pages) based on any category intersection we choose. That means we can get a listing of all examples using a particular library/language cross-section, or a listing of all examples from languages A and B from tasks where solutions from languages A and B are both present. --Michael Mol 14:17, 29 July 2010 (UTC)

So far, SMW has already done one useful thing: It's let us count how many language/task intersections we have. See Special:Properties, "Implemented in language", and the number between the parentheses. That's how many times Template:Header is being used, which is currently 13571. That's a lot of examples. --Michael Mol 14:17, 29 July 2010 (UTC)

One other thing that would be really handy is if you could query for all unimplemented examples in all languages in the mylang section of your user page. This would allow you to quickly locate tasks that you could implement. If it's possible, it would be fun to be able to sort by the number of unimplemented examples a particular language has, so the unimplemented tasks in an infrequently used language are sorted to the top. --Tyrok1 14:23, 31 July 2010 (UTC)

Issues

Remaining

These issues are new

Resolved

These issues have been resolved.

Look at Category:AutoIt. I think some SMW error messages are leaking through. I think it's got something to do with Template:stub. --Mwn3d 17:30, 30 July 2010 (UTC)

Looks like you got it. I'll be on the lookout for more problems like that. I can try to fix them when I see them. I must have added the extra colon myself when I tried to fix it. --Mwn3d 18:22, 30 July 2010 (UTC)

Fixed. The normal SMW syntax for adding properties to things is in the form of [[property::value]], but "value" shows up as an intrawiki hyperlink (to the 'property' page, I believe). Since that's not usually what I've wanted when I'm adding properties without changing the actual structure of the wiki (categories and properties are somewhat analogous, but SMW supports a few more features when dealing with properties), I try to remember to use this syntax, instead: [[property::value| ]]. I can hide a regular MW link the same way: ... --Michael Mol 18:25, 30 July 2010 (UTC)

A similar fix should be done to Libheader template, since libheader with "arg" like DBD::Mysql or alike is again interpreted by SMW. --ShinTakezou 08:35, 6 August 2010 (UTC)

Yeah, I know. I need (anyone could) to talk to the folks in #semantic-mediawiki on Freenode to find out exactly how, but I haven't had the time. --Michael Mol 13:24, 6 August 2010 (UTC)

I think SMW ate the "Example" namespace because it added a "Property" namespace, and was not expecting there to be any other namespaces. It seems to be treating the "Example" namespace as it should be treating the "Property" namespace. It may be calling out for the namespace by the index value it expected it to have. --Coderjoe 05:59, 11 August 2010 (UTC)

it appears that the following setting may be responsible: smw:Help:Configuration#smwgNamespaceIndex.

Yeah, I set that correctly (At least, I think I did), but I think a bug in SMW didn't account for it properly. I haven't had time to poke the SMW folks for proper extraction and recovery. (I know a little bit about how the DB schema in MW is set up, but I don't know what happens if I remove a namespace. The "Reports" namespace would need to get munged around, too.) --Michael Mol 10:32, 11 August 2010 (UTC)

Goals

In no particular order:

Turn Task pages into reports using SMW

Would this mean a return to the example pages of old? Is that namespace still around? If we did it that way we might end up with complications when a task is renamed. --Mwn3d 19:08, 29 August 2010 (UTC)

SMW should be able to handle the kind of queries that power the "Unimplemented in X" pages
Be able to create pages for "examples of X in language Y"
Be able to treat things like PureBasic and GAMBAS as supersets of BASIC, etc.