ARROW-5714: [JS] Inconsistent behavior in Int64Builder with/without BigNum#4691
ARROW-5714: [JS] Inconsistent behavior in Int64Builder with/without BigNum#4691TheNeuralBit wants to merge 5 commits intoapache:masterfrom
Conversation
|
@TheNeuralBit I was thinking about this more last night, and while this works for the JS-number case, it won't work if we try to set in a pair of |
|
@TheNeuralBit TheNeuralBit#3 here's a pull that should do what we want in FF |
Codecov Report
@@ Coverage Diff @@
## master #4691 +/- ##
=========================================
+ Coverage 88.94% 90.2% +1.25%
=========================================
Files 705 101 -604
Lines 81275 6399 -74876
Branches 1420 1420
=========================================
- Hits 72288 5772 -66516
+ Misses 8977 617 -8360
Partials 10 10Continue to review full report at Codecov.
|
3688820 to
2e4ce7a
Compare
Use WideBufferBuilder for Int64 and Uint64 builders
* Also modifies logic slightly, now the various input types are evenly distributed.
|
@trxcllnt I merged your PR and also clarified the int64 generation for the builder tests. Does this look ok to you? |
|
@TheNeuralBit yep, that looks great. thanks! |
DataBufferBuilder.setto ensure that Int64Builder is consistent with and without BigNum.WideBufferBuilderforInt64andUint64builders, even whenBigIntis not available. Moves check forBigIntavailability intoWideBufferBuilder. (Thanks to Paul Taylor)