Not exactly sure the reasoning to use only one index only, other than it was a performance choice at the time.
Are you sure you are seeing both indexes being used? In Geode, with the following query,select * from /region p where p.ID > 0 AND p.status = 'on' ORDER BY ID, I only see one of the indexes being used. I haven't seen a second stage query in GEODE but maybe my query is not correct. On Sun, Sep 3, 2017 at 9:17 AM Roi Apelker <roi.apel...@amdocs.com> wrote: > Thank you, > > Can you explain why " The query engine was modified to try to only use one > index if it can."? > > I also noticed that even if I query on A and B, and ORDER BY B - it seemed > to perform the query on B only, at least in a separate stage. Why is that? > > > > > -Roi > > -----Original Message----- > From: Jason Huynh [mailto:jhu...@pivotal.io] > Sent: Wednesday, August 30, 2017 10:15 PM > To: dev@geode.apache.org > Subject: Re: Indxes and hints > > You will probably have to step through debugger for this one.. it really > depends on the query. For this query, I expect the query engine to pick > one index and run the rest of the criteria on the results of the first > index used. My guess is you have created a CompactRangeIndex, and if so, > you can see in CompactRangeIndex.java around line 811: > > if (ok && runtimeItr != null && iterOps != null) { > > ok = QueryUtils.applyCondition(iterOps, context); > > } > This is where it would apply the older conditions (B and C or A and C > depending on which index was selected) The query engine was modified to try > to only use one index if it can. > > The load from disk (again assuming CompactRangeIndex) is probably > occurring in MemoryIndexStore.getTargetObject. > > On Wed, Aug 30, 2017 at 9:21 AM Roi Apelker <roi.apel...@amdocs.com> > wrote: > > > One more question: > > > > As I am trying to create a situation where the disk is accessed as > > least as possible (with a select distinct from X where a=1 and b>10 > > and c=true; In which a and b are indexes and c is not, and c is in the > > value which is evicted to disk) > > > > Did I get it right - that if I use a hint on a, or a hint on b, or a > > hint on both, it will first do a select on the hinted, and ONLY THEN the > others? > > > > Can anyone refer me to the code (where the 2 phase search occurs)? > > > > Where is the value finally loaded from disk? > > > > Thank you > > > > Roi > > -----Original Message----- > > From: Roi Apelker > > Sent: Tuesday, August 29, 2017 4:02 PM > > To: dev@geode.apache.org > > Subject: RE: Indxes and hints > > > > Thank you Jason :-) > > > > -----Original Message----- > > From: Jason Huynh [mailto:jhu...@pivotal.io] > > Sent: Monday, August 28, 2017 7:24 PM > > To: dev@geode.apache.org > > Subject: Re: Indxes and hints > > > > Hi Roi, > > > > Answers are below the questions... > > > > Question 1. Is it true to say, that the query as it is will load all > > the data values from the file, since the field C is part of the value, > > which is already persisted to file? > > > > Depending on if an index is used or not, if an index is used, the > > values that are part of the results will need to be loaded to actually > > return a result. If an index is not used, then the all the values > > would need to be loaded to actually have something to evaluate the > filter criteria on. > > > > Question 2. If I add a hint on A and B, will it mean that there will > > be a > > "2 phase search", first the select on A and B, and then, only on the > > results, on the field C? (this way, not all records will be loaded > > from file, only those that suit the A and B condition) > > > > Depending on the query, it could use one, or more. If it's a query > > with only AND clauses, it should just choose one and then evaluate the > > other filters on the subset that is returned from the index. > > > > Question 3. Is it possible to define an index on a value field? (i.e. > > not from the key) - will it work exactly like defining one form the > > key or are three any limitations? (again, I am looking to overcome the > > situation, where as it seems, the records are loaded unnecessarily > > from disk) > > > > Yes, indexes can be defined on fields in the value. It will work the > same. > > > > > > If you are sure you are already using an index in the query and still > > loading every value for every execution of that query, there may be > > something weird going on... > > > > On Sun, Aug 27, 2017 at 2:55 AM Roi Apelker <roi.apel...@amdocs.com> > > wrote: > > This message and the information contained herein is proprietary and > > confidential and subject to the Amdocs policy statement, > > > > you may review at https://www.amdocs.com/about/email-disclaimer < > > https://www.amdocs.com/about/email-disclaimer> > > This message and the information contained herein is proprietary and > > confidential and subject to the Amdocs policy statement, > > > > you may review at https://www.amdocs.com/about/email-disclaimer < > > https://www.amdocs.com/about/email-disclaimer> > > > This message and the information contained herein is proprietary and > confidential and subject to the Amdocs policy statement, > > you may review at https://www.amdocs.com/about/email-disclaimer < > https://www.amdocs.com/about/email-disclaimer> >
