I am finding that some patterns cannot be searched for.
I want to find all posts in a certain newsgroup (which has a lot of german posts) which don't include the word german, but
^"german" search failed
Also, I get too many results when searching for some words, so I tried
"searchword"&".par2" but that yields the same results, so then I tried, by themselves:
".par2" search failed
".rar" search failed
".zip" works
Is this kind of failure due to there being too many results for a given search term?
pure negation / stop words search limitations
Yes, those kinds of searches are not supported and they are very problematic in principle at the extent what is currently possible. It is expensive to use indexing to download unspecifically.
Try also to imagine how it is possible to implement "^german" search pattern. The only practical way is to exhaust all possibilities, thus the search server would need at worst to compare say 100 million of records against the search pattern.
With too frequent words as to stop words (when there are too many matches), it also in a way may need excessive CPU usage, since it may lead to a kind of linear search as well.
If linear search worked for indexing, you would see a lot of search engines, which just add all headers into a database then perform search by applying the search pattern to all records, but in practice with current CPU speeds versus the number of records it will be too slow.
Best for such matches you should use a usenet provider which supports compressed headers, then you can download headers in the newsgroups and then use quick filter to look for what you need.
Try also to imagine how it is possible to implement "^german" search pattern. The only practical way is to exhaust all possibilities, thus the search server would need at worst to compare say 100 million of records against the search pattern.
With too frequent words as to stop words (when there are too many matches), it also in a way may need excessive CPU usage, since it may lead to a kind of linear search as well.
If linear search worked for indexing, you would see a lot of search engines, which just add all headers into a database then perform search by applying the search pattern to all records, but in practice with current CPU speeds versus the number of records it will be too slow.
Best for such matches you should use a usenet provider which supports compressed headers, then you can download headers in the newsgroups and then use quick filter to look for what you need.
Re: pure negation / stop words search limitations
i've "improved" the behaviour a bit.
now if the subject pattern is well defined - the author pattern can be anything and vice versa.
it is equivalent to making a search without the stop word or negation and then applying the quick filter in UE, now one can do it on the server side.
now if the subject pattern is well defined - the author pattern can be anything and vice versa.
it is equivalent to making a search without the stop word or negation and then applying the quick filter in UE, now one can do it on the server side.