search for "locale" finds bugs containing only "local"

[Martin Pool]

I did a search for "locale" (no quotes) in bzr. It finds lots of bugs that do not mention locales at all. Maybe some incorrect word-stemming or typo-correction is matching the word "local"?

There are some other bugs like bug 29227 saying that search is too tight, in this case it seems to be too lose. See also bug 1022.

[Craig Huffstetler (xq)]

Confirmed. It may also be from a launchpad's search engine's database of common "synonym" words (if it has one). Just a venture at a guess. I am not familiar with the interworking of launchpad's system or if it has a synonym engine.

This is confirmed, though, in the meantime.

[Francis J. Lacoste (flacoste)]

Assigning to stub for consideration when we tweak our full text indexing (probably after upgrading to postgresql 8.3)

[Robert Collins (lifeless)]

Yes, almost certainly a stemming problem. It should still weight locale hits higher, but we don't use fti rank as part of the sort order in bug searches.

[Abel Deuring (adeuring)]

using FTI rank wouldn't help: the FTI stores the stemmed words, so there is no way to tell from the FTI data if the indexed word is "locale" or "local".