refactor: try to speed up update content hash #3045
Conversation
Codecov Report
@@ Coverage Diff @@
## master #3045 +/- ##
=======================================
Coverage 89.35% 89.35%
=======================================
Files 141 141
Lines 9489 9483 -6
=======================================
- Hits 8479 8474 -5
+ Misses 1010 1009 -1
Flags with carried forward coverage won't be shown. Click here to find out more.
Continue to review full report at Codecov.
|
Latency summaryCurrent PR yields:
Breakdown
Backed by latency-tracking. Further commits will update this comment. |
if this is true and brings more efficiency, then we should replace it globally |
it is true link, checking if it brings more efficiency |
|
on master: update content hash account for 27.71% of execution time on feature branch, update content hash accounts for 16.06% of entire execution time: note line 295 is calling |
|
match on 200 random docs avg time not significant performance increase, speed up 15% percent, similar as what we got in the profiling tool, 10-15 percent speed up. |
the idea:
update_content_hash, we actually never change parameters, soinclude_fieldsnever being used except test. We also get rid of the mutually exclusive check.exclude_fields,but specifyfieldsnecessary for hashing, saferCopyFromandFieldMaskis expensive, we manage to only useFieldMaskwithout creating theempty_protoSerializeToString=SerializePartialToString+ check if message is initialised, we initialise the pb message in the first line of the method, so checking is unnecessary please refer to here.