[WIP] Null support for hash-based groupby by jrhemstad · Pull Request #460 · NVIDIA/cudf

Commits on Sep 15, 2018

Commits on Sep 17, 2018

Commits on Sep 19, 2018

Added checking for NULLs in the aggregation column.
jrhemstad
committed
3c6c85b

Commits on Sep 20, 2018

Added function to set a specified bit in a validity bit mask.
jrhemstad
committed
535a777
Updated handling of NULLs in the aggregation column such that NULLs are ignored, unless all of the values for a given key are NULL.
jrhemstad
committed
b7db9eb
Updated setting the validity bits in the aggregation output to use atomicOrs to avoid race conditions.
jrhemstad
committed
c904f47
Merge remote-tracking branch 'origin/master' into fea-ext-null-suport-hash-groupby
jrhemstad
committed
94d1e8c
Added kernel to gather the validity bits of a gdf_column. Used the kernel to rearrange the validity bits of the aggregation output column after sorting the results.
jrhemstad
committed
e4980f6
Fixed bug where the wrong bit was getting set in the aggregation output validity mask during the extraction kernel.
jrhemstad
committed
90f0c0c
Corrected the AVG aggregator implementation so the validity bitmask of the AVG aggregator will get set correctly.
jrhemstad
committed
bd98d2e
Added check to ensure we don't try and sort the result if the output size is zero.
jrhemstad
committed
c5fcaed
Return an error if the output aggregation column's nullmask isn't allocated.
jrhemstad
committed
693895f
No longer return an error if the aggregation column's validity buffer isn't allocated.
jrhemstad
committed
9ba1fc9
Updated the build_aggregation_kernel for Count such that the aggregation output can never be Null. Instead, an aggregation of all NULLs for Count will just return 0.
jrhemstad
committed
19ac0be
Updated scope of the insert location variable in the Count overload of build aggregation table.
jrhemstad
committed
39eaa45

Commits on Sep 24, 2018

Commits on Sep 25, 2018

Merge remote-tracking branch 'public/master' into fea-ext-null-suport-hash-groupby
jrhemstad
committed
4ee4f9d

Navigation Menu

Uh oh!

[WIP] Null support for hash-based groupby - #460

[WIP] Null support for hash-based groupby#460
jrhemstad wants to merge 28 commits into
NVIDIA:branch-0.5from
jrhemstad:fea-ext-null-suport-hash-groupby

Commits on Sep 15, 2018

Merge branch 'fea-ext-null-support-hash-joins' into fea-ext-null-suport-hash-groupby

Implemented null support for hash-based groupbys.

Merge branch 'master' into fea-ext-null-suport-hash-groupby

Commits on Sep 17, 2018

Merge remote-tracking branch 'public/master' into fea-ext-null-suport-hash-groupby

Merge branch 'fea-ext-null-suport-hash-groupby' of github.com:jrhemstad/libgdf into fea-ext-null-suport-hash-groupby

Commits on Sep 19, 2018

Added checking for NULLs in the aggregation column.

Commits on Sep 20, 2018

Added function to set a specified bit in a validity bit mask.

Updated handling of NULLs in the aggregation column such that NULLs are ignored, unless all of the values for a given key are NULL.

Updated setting the validity bits in the aggregation output to use atomicOrs to avoid race conditions.

Merge remote-tracking branch 'origin/master' into fea-ext-null-suport-hash-groupby

Added kernel to gather the validity bits of a gdf_column. Used the kernel to rearrange the validity bits of the aggregation output column after sorting the results.

Fixed bug where the wrong bit was getting set in the aggregation output validity mask during the extraction kernel.

Corrected the AVG aggregator implementation so the validity bitmask of the AVG aggregator will get set correctly.

Added check to ensure we don't try and sort the result if the output size is zero.

Return an error if the output aggregation column's nullmask isn't allocated.

No longer return an error if the aggregation column's validity buffer isn't allocated.

Updated the build_aggregation_kernel for Count such that the aggregation output can never be Null. Instead, an aggregation of all NULLs for Count will just return 0.

Updated scope of the insert location variable in the Count overload of build aggregation table.

Commits on Sep 24, 2018

Removed turn_bit_on function as it's not used and not thread safe.

Updated documentation.

Updated documentation.

Updated documentation.

Updated implementation of AVG aggregator to check the validity of the Count value before doing the division.

Commits on Sep 25, 2018

Merge remote-tracking branch 'public/master' into fea-ext-null-suport-hash-groupby

Commits on Dec 6, 2018

Merge remote-tracking branch 'origin/master' into fea-ext-null-suport-hash-groupby

Updated cudaMallocs to RMM_ALLOCs.

Commits on Dec 14, 2018

Merge remote-tracking branch 'origin/branch-0.5' into fea-ext-null-suport-hash-groupby

Updated outdated reference to Vector to rmm::device_vector.