On Fri, 28 Jul 2000, ethan mindlace fremen wrote:
Actually, - is treated as a continuation (ala hyphen) so searching for dtmlin works fine.
_ is a stop word, so searching for "mod pcgi" works.
I'm going to note this on the search page.
I'm looking into changing this, but Splitter.c is a good description of the headache I get :)
If the index treats _ as punctuation or a stop, that would be more or less OK as far as searching goes if the processing in the lookup method for text indexes did the *same* splitting (you just have to make sure it also ANDs (or better yet NEARS) the parts it splits). This appears not to be the case, from this problem description and my own experience. IMO, this is a *serious* bug in ZCatalog, and I put it in the collector a while ago. In my experience cataloging does not always happen correctly, either. For instance, 'T-shirt' as a text index input word does not appear to make the object findable using the keyword 'shirt'. (I have not tested the latter behavior on 2.2 yet; I saw it on 2.1.4.) --RDM