You can call checkCurrentAuthor(false) and checkCurrentUser(false) on the query to disable it. On Mon, Mar 5, 2018 at 9:45 AM, Thomas Mortagne <[email protected]> wrote:
Also the SOLR API you used is checking the right for each result which is not the case for the xwql one.
On Mon, Mar 5, 2018 at 9:31 AM, Vincent Massol <[email protected]> wrote:
On 5 Mar 2018, at 09:28, Thomas Mortagne <[email protected]> wrote:
Well "XWQL" does not really mean much since that's not what is executed. Better indicate the database you are comparing with SOLR, I guess it's HSQLDB here. You would most probably don't get the same result with a different database and it also depends a lot on the database configuration (how much is cached, etc.).
Good point.
It’s still interesting to know that with our standalone distribution it would be faster with XWQL on HSQLDB than SOLR queries.
Next step: test this on mysql.
Thanks -Vincent
On Sun, Mar 4, 2018 at 12:46 PM, Vincent Massol <[email protected]> wrote:
On 4 Mar 2018, at 12:36, Vincent Massol <[email protected]> wrote:
On 4 Mar 2018, at 12:20, Ludovic Dubost <[email protected]> wrote:
Hi
This is interesting.
Now I believe these results might change under volume and also depending on the type of queries. A like on xwql/hql would be slow while a solr text search would give fast results.
Yes I agree. Obviously if we want to check this further we need to do more tests. Doing a “LIKE” one is interesting. I’ll try to do one.
On a small data set, the advantage is still for XWQL, see http://snippets.xwiki.org/xwiki/bin/view/Extension/Performance%20of%20SOLRQL...
Note that the majority of queries done by XWiki during a page rendering are done on small result set.
But indeed, it would be interesting to load XWiki with, say 2M pages and do the test again.
Thanks -Vincent
I’m lacking the data to do volume testing both in term of quantity of data and in term of load on XWiki.
Another point to consider: this test is done with SOLR being embedded (which should provide the best perf on low volumes IMO). On larger volumes you’d need to use an external SOLR which would perform better but you’d suffer from the marshalling/unmarshalling of requests/responses.
Thanks -Vincent
Ludovic
Le 4 mars 2018 11:46, "Vincent Massol" <[email protected]> a écrit :
Hi devs,
I was curious to know how SOLRQL compared vs XWQL (for example to progress on the idea of moving more to using SOLRQL for doing queries and thus being able for ex to use a store based on, say, git).
I put my quick result here: http://snippets.xwiki.org/ xwiki/bin/view/Extension/Performance%20of%20SOLRQL%20vs%20XWQL/
In short, it seems that XWQL wins by a factor of at least x2.
Is that your experience too?
Let me know if I’ve made a mistake somewhere.
Thanks -Vincent
-- Thomas Mortagne
-- Thomas Mortagne
-- Thomas Mortagne