Enable decimal support in parquet writer

[devavret]

Resolves #7669

[kkraus14]

@devavret are there any issues with round tripping decimal64 vs decimal32 in generating Arrow / Pandas compatible metadata, or is that handled by the logical / physical types in Parquet anyway?

[codecov]

Codecov Report

Merging #7673 (44a2457) into branch-0.19 (7871e7a) will increase coverage by 0.22%.
The diff coverage is n/a.

❗ Current head 44a2457 differs from pull request most recent head 38c0c49. Consider uploading reports for the commit 38c0c49 to get more accurate results

@@               Coverage Diff               @@
##           branch-0.19    #7673      +/-   ##
===============================================
+ Coverage        81.86%   82.08%   +0.22%     
===============================================
  Files              101      101              
  Lines            16884    17036     +152     
===============================================
+ Hits             13822    13984     +162     
+ Misses            3062     3052      -10     

Impacted Files	Coverage Δ
python/cudf/cudf/core/column/categorical.py	91.62% <ø> (+0.23%)	⬆️
python/cudf/cudf/core/column/column.py	87.77% <ø> (+0.01%)	⬆️
python/cudf/cudf/core/column/datetime.py	89.09% <ø> (ø)
python/cudf/cudf/core/column/decimal.py	92.75% <ø> (-2.12%)	⬇️
python/cudf/cudf/core/column/lists.py	91.81% <ø> (+0.42%)	⬆️
python/cudf/cudf/core/column/numerical.py	94.83% <ø> (-0.20%)	⬇️
python/cudf/cudf/core/column/string.py	86.58% <ø> (+0.08%)	⬆️
python/cudf/cudf/core/column/timedelta.py	88.23% <ø> (ø)
python/cudf/cudf/core/column_accessor.py	95.28% <ø> (-0.03%)	⬇️
python/cudf/cudf/core/dataframe.py	90.79% <ø> (+0.32%)	⬆️
... and 30 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c21bd0e...38c0c49. Read the comment docs.

[devavret]

@devavret are there any issues with round tripping decimal64 vs decimal32 in generating Arrow / Pandas compatible metadata, or is that handled by the logical / physical types in Parquet anyway?

Correct me if I'm wrong but I thought cudf only had decimal64

cudf/python/cudf/cudf/core/dtypes.py

Line 234 in 04f9021

class Decimal64Dtype(ExtensionDtype):

[kkraus14]

@devavret are there any issues with round tripping decimal64 vs decimal32 in generating Arrow / Pandas compatible metadata, or is that handled by the logical / physical types in Parquet anyway?

Correct me if I'm wrong but I thought cudf only had decimal64

cudf/python/cudf/cudf/core/dtypes.py

Line 234 in 04f9021

class Decimal64Dtype(ExtensionDtype):

For now yes, but Decimal32 is planned to be supported in the future and we want to make sure that we don't introduce changes into the Parquet metadata that aren't forward compatible if possible.

[devavret]

@devavret are there any issues with round tripping decimal64 vs decimal32 in generating Arrow / Pandas compatible metadata, or is that handled by the logical / physical types in Parquet anyway?

Correct me if I'm wrong but I thought cudf only had decimal64

cudf/python/cudf/cudf/core/dtypes.py

Line 234 in 04f9021

class Decimal64Dtype(ExtensionDtype):

For now yes, but Decimal32 is planned to be supported in the future and we want to make sure that we don't introduce changes into the Parquet metadata that aren't forward compatible if possible.

Alright then yes, for that, it will be taken care of during the conversion of Column to cudf::column_view. And as long as the dtype class for Decimal32 has a precision member then no changes would be required to the bindings either.

[kkraus14]

@gpucibot merge