Ali Alsuliman has posted comments on this change.

Change subject: [ASTERIXDB-2523][RT][COMP] add support for hashing record fields
......................................................................


Patch Set 2:

(2 comments)

https://asterix-gerrit.ics.uci.edu/#/c/3249/1/asterixdb/asterix-om/src/main/java/org/apache/asterix/dataflow/data/nontagged/hash/AMurmurHash3BinaryHashFunctionFamily.java
File 
asterixdb/asterix-om/src/main/java/org/apache/asterix/dataflow/data/nontagged/hash/AMurmurHash3BinaryHashFunctionFamily.java:

Line 161:             int hash = 0;
> should we use the 'seed' argument as a starting value instead of eliminatin
I was worried about cases like hash join if for example the left branch has a 
different seed than the right branch. Empty array [] would hash to two 
different places. But I don't know if this can ever happen. I checked the usage 
of seed. The majority of them are 0. But it's different for hash group by and 
join where we increase the seed by 1 for each level.

Should I change the starting hash = seed?


Line 200:                     hash ^= 
MurmurHash3BinaryHash.hash(fieldName.getByteArray(), fieldName.getStartOffset(),
> just to confirm. we're not using field names, only their values for hashing
Changed it to include the field names.


-- 
To view, visit https://asterix-gerrit.ics.uci.edu/3249
To unsubscribe, visit https://asterix-gerrit.ics.uci.edu/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I9d18a17089afdc6e28ff01091d890b375868969c
Gerrit-PatchSet: 2
Gerrit-Project: asterixdb
Gerrit-Branch: master
Gerrit-Owner: Ali Alsuliman <[email protected]>
Gerrit-Reviewer: Ali Alsuliman <[email protected]>
Gerrit-Reviewer: Anon. E. Moose #1000171
Gerrit-Reviewer: Dmitry Lychagin <[email protected]>
Gerrit-Reviewer: Jenkins <[email protected]>
Gerrit-Reviewer: Till Westmann <[email protected]>
Gerrit-HasComments: Yes

Reply via email to