Address #269
The current performance bottleneck is in getting labels by rows.
Typical cases in using Frame.AggregateRowsBy has data frame with way more rows than columns. Switching to getting labels by columns yields a 3x speed improvement in a test case of 100000 rows.
Didn't change the row index to group keys as it will be a breaking change. Will consider it next time as it's more in line with what pandas does.
Needs to fix the breaking performance test to document speed different among different versions in the future.
Address #269 The current performance bottleneck is in getting labels by rows.
Typical cases in using Frame.AggregateRowsBy has data frame with way more rows than columns. Switching to getting labels by columns yields a 3x speed improvement in a test case of 100000 rows.
Didn't change the row index to group keys as it will be a breaking change. Will consider it next time as it's more in line with what
pandas
does.Needs to fix the breaking performance test to document speed different among different versions in the future.