Community
Participate
Working Groups
Using Eclipse 2.0, Build F1 (buildid: 20020521) on Windows 2000. From Eclipse, go to menu 'Help'->'Help Contents' and do a search for "Eclipse". Instead of finding all instances of "Eclipse" it will search for (and highlight) "Eclips". But this is not an off-by-one error, if you search for "something", it instead looks for "someth". By searching for "used" you can see that it is indeed searching incorrectly, and not just highlighting incorrectly. It will search for "us" and will match all the "Using ..." pages.
Original Summary: Help window does incorrect searching and highlighting The behavior you are seeing is caused by stemming being enabled in help search. It is not fuzzy search that would allow any arbitrary letter (or more) to differ. For searching documentation it is desired for search engine to perform stemming. It causes search for using eclipse to find using eclipse how eclipse can be used how to use eclipse which is an improvement over finding just using eclipse. There is no perfect stemming algorithms, and one we use will extract root of words from Eclipse and something as eclips and someth, but this does not negatively affect effectiveness of search in most cases. I believe what you are looking for is exact string match. Help system defines extension point for plugging in language analyzer. By default, on English locale, help uses analyzer that performs stemming, but this can be easily changed by plugging in a simpler analyzer, that breaks text into words without stemming or stop words removal. What help does not currently support is both search with stemming enabled and search for exact match at the same time. With stemming being enabled, the quoted query means a search for a phrase consisting of consecutive words but not an exact string match. I will change this bug to a request for this enhancement, and defer it for later. Implementing this requirement will probably require second index to be created in parallel, and there are performance issues associated with this approach.
*** Bug 20802 has been marked as a duplicate of this bug. ***
Re-opening the bug...
Targeted for 2.1.
We will add another fields in the index to keep not stemmed version of ducument words. This fields will going to be searched when keywords are double quoted in the search query.
Imlemented and released in head.
I tested this and it works great! Any way to get it into 2.0.2? I think it would be a great improvement for any users of the help system.
We have received good feedback regarding this feature, and no regression has been reported. Since it is a great improvement, I have merged the code into 2.0.2 builds as well. 20021009 2.0.2 build will pick up the changes.
*** Bug 24801 has been marked as a duplicate of this bug. ***
