David Harrison wrote:
Not directly related to your query, but did you have any problems
building an index of that size with the PHP version? My tests on large
collections (100,000 items +) has the index failing with out-of-memory
errors. I was able to fix it by setting memory_limit explicitly at the
top of the script, but I don't think we have such problems with the
java version.
I don't recall the exact memory limit I needed, but it was over 1GB, so
yes that's another issue.
On the topic - before I start having problems with memory limits and
the framework throwing exceptions, I do seem to be able to search up
to 100,000 items at fast speeds (ie, less than a second to return
results, regardless of how many results are returned).
Does that very with query complexity? Multi-term queries seem to be a
problem for me because my reading of the PHP code says that it does each
term search separately, stores all the results of each in memory and
then combines them at the end.
Kris Jurka
On 7/14/07, Kris Jurka <[EMAIL PROTECTED]> wrote:
I'm having some performance problems with a Lucene index comparing the
PHP version to the Java version. I've got 672213 documents and searches
for small results come back very quickly:
$ ./searchtest.php fred
Search in: 0.0557591915131
Hits: 170
When I search for something that returns more results, the time
increases dramatically:
$ ./searchtest.php 'hey there good looking'
Search in: 7.81737494469
Hits: 60847
The Java version of Lucene doesn't have a problem with more results:
$ java -classpath lucene-core-2.2.0.jar:. SearchTest 'hey there good
looking'
Search in: 0.264
Hits: 38494
Is this a known issue? Can someone point me in the direction of fixing
this? I've got a client with a production website getting killed
because they didn't test carefully before deploying.
Kris Jurka
