Alexey Serbin has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/14503 )

Change subject: KUDU-2980: fix ORDERED scans when key columns are misordered in 
projection
......................................................................


Patch Set 4:

(6 comments)

http://gerrit.cloudera.org:8080/#/c/14503/4/src/kudu/client/client-test.cc
File src/kudu/client/client-test.cc:

http://gerrit.cloudera.org:8080/#/c/14503/4/src/kudu/client/client-test.cc@6391
PS4, Line 6391: random
It looks like a pretty deterministic schema based on the circular next_type 
transition, no?


http://gerrit.cloudera.org:8080/#/c/14503/4/src/kudu/client/client-test.cc@6401
PS4, Line 6401: ->NotNull();
Is it important to test the case when nullable columns are present as well?


http://gerrit.cloudera.org:8080/#/c/14503/4/src/kudu/client/client-test.cc@6409
PS4, Line 6409:       case KuduColumnSchema::INT8: data_type = 
KuduColumnSchema::INT16; break;
              :       case KuduColumnSchema::INT16: data_type = 
KuduColumnSchema::INT32; break;
              :       case KuduColumnSchema::INT32: data_type = 
KuduColumnSchema::INT64; break;
              :       case KuduColumnSchema::INT64: data_type = 
KuduColumnSchema::STRING; break;
              :       case KuduColumnSchema::STRING: data_type = 
KuduColumnSchema::INT8; break;
              :       default: LOG(FATAL) << "Unexpected data type " << 
data_type;
Does it make sense to add some randomness here?  For example, choose next type 
based on rand() % num_types?  In that case there might be cases when columns of 
the same type are adjacent in the sequence.


http://gerrit.cloudera.org:8080/#/c/14503/4/src/kudu/client/client-test.cc@6439
PS4, Line 6439: num_cols_to_project
Does it brake if Uniform() returns 0?


http://gerrit.cloudera.org:8080/#/c/14503/4/src/kudu/client/client-test.cc@6491
PS4, Line 6491:   // Use a fault tolerant scan half the time.
              :   if (rng.OneIn(2)) {
              :     ASSERT_OK(scanner->SetFaultTolerant());
              :   }
This is a fuzz test, but maybe it's better to have this as a parameter anyways 
since it's not the part of the targeted fuzzing?


http://gerrit.cloudera.org:8080/#/c/14503/4/src/kudu/client/client-test.cc@6499
PS4, Line 6499: unordered_set
Could you comment on why it's important (or actually not so important) to have 
unrodered_set, and not just set, while performing the comparison?



--
To view, visit http://gerrit.cloudera.org:8080/14503
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Idb213f466c7d01b6953a863d8aa53a34b5ac8893
Gerrit-Change-Number: 14503
Gerrit-PatchSet: 4
Gerrit-Owner: Adar Dembo <[email protected]>
Gerrit-Reviewer: Adar Dembo <[email protected]>
Gerrit-Reviewer: Alexey Serbin <[email protected]>
Gerrit-Reviewer: Andrew Wong <[email protected]>
Gerrit-Reviewer: Attila Bukor <[email protected]>
Gerrit-Reviewer: Grant Henke <[email protected]>
Gerrit-Comment-Date: Wed, 23 Oct 2019 05:59:10 +0000
Gerrit-HasComments: Yes

Reply via email to