Incidentally, I just noticed that the current case massaging (Lyon
metro -> Lyon Metro) is more complex than I thought. Anyone know the
exact rules? It seems like these are possible:
SOME ARTICLE -> some article
some article-> SOME ARTICLE (eg, sdi silently -> SDI)
some article -> Some Article
Some Article -> some article
some ARTICLE -> some article (or Some Article)
sOme artIclE -> some article (or Some Article)
*but* this isn't:
some other article -/-> Some other Article
(example: "departements of france" -/-> "Departements of France")
So it seems like it tries this:
match name verbatim
try lowercasing the entire search string
try uppercasing the first letter of each word
try uppercasing the entire search string
But that's it.
With a bit of testing (not at Wikipedia) I determined that the order
is as above. That is, the search string "soMe artICLE" will match
"Some Article" before it matches "SOME ARTICLE". This strikes me as
another limitation: when the capitalisation of an article's title is
ambiguous, no warning is displayed to the user. I guess we usually
handle that situation with dablinks at the top of the page, but
still...
Anyway, to improve the matching it would presumably be necessary to
store some sort of canonical hash for each title, perhaps lowercased,
and with all accents replaced by non-accented letters, or whatever.
But there are probably better places to expend that effort. (like
fixing search :))
Steve
On 6/2/06, Steve Bennett <stevage(a)gmail.com> wrote:
On 6/2/06, Timwi <timwi(a)gmx.net> wrote:
Surely the most useful solution would be to make
the Go button redirect
non-silently, i.e. in the same way as a real redirect. A wikilink [[Lyon
metro]] would still be a redlink, but typing "Lyon metro" in the Go box
or the actual URL should bring up [[Lyon Metro]] with the note,
"(redirected from _Lyon metro_)", where "Lyon metro" is, of course,
a
redlink. You can then use that to create the page if you wanted to.
Yes...why didn't I think of that?
Steve