ARROW-15006: [Python][CI][Doc] Enable numpydoc check PR03

[amoeba]

Adds an additional numypdoc check to CI (PR03) and fixes all corresponding violations.

This includes one minor breaking change to parameter order in pyarrow.parquet.read_table which makes order more uniform across ParquetDataset.__new__, read_table, and read_pandas which have related signatures.

Note this does not fully resolve ARROW-15006.

[github-actions]

https://issues.apache.org/jira/browse/ARROW-15006

[github-actions]

⚠️ Ticket has not been started in JIRA, please click 'Start Progress'.

[jorisvandenbossche]

Maybe we could rather change the order in the actual signature, since it seems useful to have skip_rows and skip_rows_after_names together?

[amoeba]

That makes sense. And other places in pyarrow and libarrow that reference both have them grouped together. Changed in 44acfb3. Thanks.

[amoeba]

Thanks for reviewing this @jorisvandenbossche. I took your suggestion on keeping skip_rows and skip_rows_after_names together. I re-ran archery numpydoc --allow-rule PR03 which still passes. This is ready for another look.

[jorisvandenbossche]

Thanks for the updates! There is a conflict with dataset.pyx now that needs to be resolved.

[jorisvandenbossche]

For this one, I would maybe also change the order of the signature instead, and make all keywords (except the first) keyword-only, using , *,

[amoeba]

I pushed a change to address this in d04b228 that:

• Re-orders the signatures and docstrings inside the Scanner class to be consistent and follow what I think is a sensible order
• Makes Scanner.from_dataset, Scanner.from_fragment, and Scanner.from_batches use the above pattern to make all args except the first keyword only

Waiting on CI.

[amoeba]

Thanks @jorisvandenbossche, the conflict should now be resolved. I made an attempt at addressing your comment so let me know if that's what you were thinking. This is ready for another look when you have time.

[amoeba]

Ping @jorisvandenbossche: Any chance you have a spare moment to take a look at this again? Thanks!

[amoeba]

Sorry for the noise, I was cleaning up my fork and wiped out the commits. They've since been restored and this is ready for review.

[jorisvandenbossche]

A very tiny last comment

[amoeba]

Thanks for the review @jorisvandenbossche. I didn't mean to remove that bit and I'm not sure how I did. Good catch! All three instances of that param docstring now have the expected text.

Ready for review/merge.

[jorisvandenbossche]

Thanks!

[amoeba]

Adds an additional numypdoc check to CI (PR03) and fixes all corresponding violations.

This includes one minor breaking change to parameter order in pyarrow.parquet.read_table which makes order more uniform across ParquetDataset.__new__, read_table, and read_pandas which have related signatures.

Note this does not fully resolve ARROW-15006.

[github-actions]

https://issues.apache.org/jira/browse/ARROW-15006

[github-actions]

⚠️ Ticket has not been started in JIRA, please click 'Start Progress'.

[jorisvandenbossche]

Maybe we could rather change the order in the actual signature, since it seems useful to have skip_rows and skip_rows_after_names together?

[amoeba]

That makes sense. And other places in pyarrow and libarrow that reference both have them grouped together. Changed in 44acfb3. Thanks.

[amoeba]

Thanks for reviewing this @jorisvandenbossche. I took your suggestion on keeping skip_rows and skip_rows_after_names together. I re-ran archery numpydoc --allow-rule PR03 which still passes. This is ready for another look.

[jorisvandenbossche]

Thanks for the updates! There is a conflict with dataset.pyx now that needs to be resolved.

[jorisvandenbossche]

For this one, I would maybe also change the order of the signature instead, and make all keywords (except the first) keyword-only, using , *,

[amoeba]

I pushed a change to address this in d04b228 that:

• Re-orders the signatures and docstrings inside the Scanner class to be consistent and follow what I think is a sensible order
• Makes Scanner.from_dataset, Scanner.from_fragment, and Scanner.from_batches use the above pattern to make all args except the first keyword only

Waiting on CI.

[amoeba]

Thanks @jorisvandenbossche, the conflict should now be resolved. I made an attempt at addressing your comment so let me know if that's what you were thinking. This is ready for another look when you have time.

[amoeba]

Ping @jorisvandenbossche: Any chance you have a spare moment to take a look at this again? Thanks!

[amoeba]

Sorry for the noise, I was cleaning up my fork and wiped out the commits. They've since been restored and this is ready for review.

[jorisvandenbossche]

A very tiny last comment

[amoeba]

Thanks for the review @jorisvandenbossche. I didn't mean to remove that bit and I'm not sure how I did. Good catch! All three instances of that param docstring now have the expected text.

Ready for review/merge.

[jorisvandenbossche]

Thanks!

[amoeba]

Adds an additional numypdoc check to CI (PR03) and fixes all corresponding violations.

This includes one minor breaking change to parameter order in pyarrow.parquet.read_table which makes order more uniform across ParquetDataset.__new__, read_table, and read_pandas which have related signatures.

Note this does not fully resolve ARROW-15006.

[github-actions]

https://issues.apache.org/jira/browse/ARROW-15006

[github-actions]

⚠️ Ticket has not been started in JIRA, please click 'Start Progress'.

[jorisvandenbossche]

Maybe we could rather change the order in the actual signature, since it seems useful to have skip_rows and skip_rows_after_names together?

[amoeba]

That makes sense. And other places in pyarrow and libarrow that reference both have them grouped together. Changed in 44acfb3. Thanks.

[amoeba]

Thanks for reviewing this @jorisvandenbossche. I took your suggestion on keeping skip_rows and skip_rows_after_names together. I re-ran archery numpydoc --allow-rule PR03 which still passes. This is ready for another look.

[jorisvandenbossche]

Thanks for the updates! There is a conflict with dataset.pyx now that needs to be resolved.

[jorisvandenbossche]

For this one, I would maybe also change the order of the signature instead, and make all keywords (except the first) keyword-only, using , *,

[amoeba]

I pushed a change to address this in d04b228 that:

• Re-orders the signatures and docstrings inside the Scanner class to be consistent and follow what I think is a sensible order
• Makes Scanner.from_dataset, Scanner.from_fragment, and Scanner.from_batches use the above pattern to make all args except the first keyword only

Waiting on CI.

[amoeba]

Thanks @jorisvandenbossche, the conflict should now be resolved. I made an attempt at addressing your comment so let me know if that's what you were thinking. This is ready for another look when you have time.

[amoeba]

Ping @jorisvandenbossche: Any chance you have a spare moment to take a look at this again? Thanks!

[amoeba]

Sorry for the noise, I was cleaning up my fork and wiped out the commits. They've since been restored and this is ready for review.

[jorisvandenbossche]

A very tiny last comment

[amoeba]

Thanks for the review @jorisvandenbossche. I didn't mean to remove that bit and I'm not sure how I did. Good catch! All three instances of that param docstring now have the expected text.

Ready for review/merge.

[jorisvandenbossche]

Thanks!

[amoeba]

Adds an additional numypdoc check to CI (PR03) and fixes all corresponding violations.

This includes one minor breaking change to parameter order in pyarrow.parquet.read_table which makes order more uniform across ParquetDataset.__new__, read_table, and read_pandas which have related signatures.

Note this does not fully resolve ARROW-15006.

[github-actions]

https://issues.apache.org/jira/browse/ARROW-15006

[github-actions]

⚠️ Ticket has not been started in JIRA, please click 'Start Progress'.

[jorisvandenbossche]

Maybe we could rather change the order in the actual signature, since it seems useful to have skip_rows and skip_rows_after_names together?

[amoeba]

That makes sense. And other places in pyarrow and libarrow that reference both have them grouped together. Changed in 44acfb3. Thanks.

[amoeba]

Thanks for reviewing this @jorisvandenbossche. I took your suggestion on keeping skip_rows and skip_rows_after_names together. I re-ran archery numpydoc --allow-rule PR03 which still passes. This is ready for another look.

[jorisvandenbossche]

Thanks for the updates! There is a conflict with dataset.pyx now that needs to be resolved.

[jorisvandenbossche]

For this one, I would maybe also change the order of the signature instead, and make all keywords (except the first) keyword-only, using , *,

[amoeba]

I pushed a change to address this in d04b228 that:

• Re-orders the signatures and docstrings inside the Scanner class to be consistent and follow what I think is a sensible order
• Makes Scanner.from_dataset, Scanner.from_fragment, and Scanner.from_batches use the above pattern to make all args except the first keyword only

Waiting on CI.

[amoeba]

Thanks @jorisvandenbossche, the conflict should now be resolved. I made an attempt at addressing your comment so let me know if that's what you were thinking. This is ready for another look when you have time.

[amoeba]

Ping @jorisvandenbossche: Any chance you have a spare moment to take a look at this again? Thanks!

[amoeba]

Sorry for the noise, I was cleaning up my fork and wiped out the commits. They've since been restored and this is ready for review.

[jorisvandenbossche]

A very tiny last comment

[amoeba]

Thanks for the review @jorisvandenbossche. I didn't mean to remove that bit and I'm not sure how I did. Good catch! All three instances of that param docstring now have the expected text.

Ready for review/merge.

[jorisvandenbossche]

Thanks!

[amoeba]

Adds an additional numypdoc check to CI (PR03) and fixes all corresponding violations.

This includes one minor breaking change to parameter order in pyarrow.parquet.read_table which makes order more uniform across ParquetDataset.__new__, read_table, and read_pandas which have related signatures.

Note this does not fully resolve ARROW-15006.

[github-actions]

https://issues.apache.org/jira/browse/ARROW-15006

[github-actions]

⚠️ Ticket has not been started in JIRA, please click 'Start Progress'.

[jorisvandenbossche]

Maybe we could rather change the order in the actual signature, since it seems useful to have skip_rows and skip_rows_after_names together?

[amoeba]

That makes sense. And other places in pyarrow and libarrow that reference both have them grouped together. Changed in 44acfb3. Thanks.

[amoeba]

Thanks for reviewing this @jorisvandenbossche. I took your suggestion on keeping skip_rows and skip_rows_after_names together. I re-ran archery numpydoc --allow-rule PR03 which still passes. This is ready for another look.

[jorisvandenbossche]

Thanks for the updates! There is a conflict with dataset.pyx now that needs to be resolved.

[jorisvandenbossche]

For this one, I would maybe also change the order of the signature instead, and make all keywords (except the first) keyword-only, using , *,

[amoeba]

I pushed a change to address this in d04b228 that:

• Re-orders the signatures and docstrings inside the Scanner class to be consistent and follow what I think is a sensible order
• Makes Scanner.from_dataset, Scanner.from_fragment, and Scanner.from_batches use the above pattern to make all args except the first keyword only

Waiting on CI.

[amoeba]

Thanks @jorisvandenbossche, the conflict should now be resolved. I made an attempt at addressing your comment so let me know if that's what you were thinking. This is ready for another look when you have time.

[amoeba]

Ping @jorisvandenbossche: Any chance you have a spare moment to take a look at this again? Thanks!

[amoeba]

Sorry for the noise, I was cleaning up my fork and wiped out the commits. They've since been restored and this is ready for review.

[jorisvandenbossche]

A very tiny last comment

[amoeba]

Thanks for the review @jorisvandenbossche. I didn't mean to remove that bit and I'm not sure how I did. Good catch! All three instances of that param docstring now have the expected text.

Ready for review/merge.

[jorisvandenbossche]

Thanks!

[amoeba]

Adds an additional numypdoc check to CI (PR03) and fixes all corresponding violations.

This includes one minor breaking change to parameter order in pyarrow.parquet.read_table which makes order more uniform across ParquetDataset.__new__, read_table, and read_pandas which have related signatures.

Note this does not fully resolve ARROW-15006.

[github-actions]

https://issues.apache.org/jira/browse/ARROW-15006

[github-actions]

⚠️ Ticket has not been started in JIRA, please click 'Start Progress'.

[jorisvandenbossche]

Maybe we could rather change the order in the actual signature, since it seems useful to have skip_rows and skip_rows_after_names together?

[amoeba]

That makes sense. And other places in pyarrow and libarrow that reference both have them grouped together. Changed in 44acfb3. Thanks.

[amoeba]

Thanks for reviewing this @jorisvandenbossche. I took your suggestion on keeping skip_rows and skip_rows_after_names together. I re-ran archery numpydoc --allow-rule PR03 which still passes. This is ready for another look.

[jorisvandenbossche]

Thanks for the updates! There is a conflict with dataset.pyx now that needs to be resolved.

[jorisvandenbossche]

For this one, I would maybe also change the order of the signature instead, and make all keywords (except the first) keyword-only, using , *,

[amoeba]

I pushed a change to address this in d04b228 that:

• Re-orders the signatures and docstrings inside the Scanner class to be consistent and follow what I think is a sensible order
• Makes Scanner.from_dataset, Scanner.from_fragment, and Scanner.from_batches use the above pattern to make all args except the first keyword only

Waiting on CI.

[amoeba]

Thanks @jorisvandenbossche, the conflict should now be resolved. I made an attempt at addressing your comment so let me know if that's what you were thinking. This is ready for another look when you have time.

[amoeba]

Ping @jorisvandenbossche: Any chance you have a spare moment to take a look at this again? Thanks!

[amoeba]

Sorry for the noise, I was cleaning up my fork and wiped out the commits. They've since been restored and this is ready for review.

[jorisvandenbossche]

A very tiny last comment

[amoeba]

Thanks for the review @jorisvandenbossche. I didn't mean to remove that bit and I'm not sure how I did. Good catch! All three instances of that param docstring now have the expected text.

Ready for review/merge.

[jorisvandenbossche]

Thanks!

[amoeba]

Adds an additional numypdoc check to CI (PR03) and fixes all corresponding violations.

This includes one minor breaking change to parameter order in pyarrow.parquet.read_table which makes order more uniform across ParquetDataset.__new__, read_table, and read_pandas which have related signatures.

Note this does not fully resolve ARROW-15006.

[github-actions]

https://issues.apache.org/jira/browse/ARROW-15006

[github-actions]

⚠️ Ticket has not been started in JIRA, please click 'Start Progress'.

[jorisvandenbossche]

Maybe we could rather change the order in the actual signature, since it seems useful to have skip_rows and skip_rows_after_names together?

[amoeba]

That makes sense. And other places in pyarrow and libarrow that reference both have them grouped together. Changed in 44acfb3. Thanks.

[amoeba]

Thanks for reviewing this @jorisvandenbossche. I took your suggestion on keeping skip_rows and skip_rows_after_names together. I re-ran archery numpydoc --allow-rule PR03 which still passes. This is ready for another look.

[jorisvandenbossche]

Thanks for the updates! There is a conflict with dataset.pyx now that needs to be resolved.

[jorisvandenbossche]

For this one, I would maybe also change the order of the signature instead, and make all keywords (except the first) keyword-only, using , *,

[amoeba]

I pushed a change to address this in d04b228 that:

• Re-orders the signatures and docstrings inside the Scanner class to be consistent and follow what I think is a sensible order
• Makes Scanner.from_dataset, Scanner.from_fragment, and Scanner.from_batches use the above pattern to make all args except the first keyword only

Waiting on CI.

[amoeba]

Thanks @jorisvandenbossche, the conflict should now be resolved. I made an attempt at addressing your comment so let me know if that's what you were thinking. This is ready for another look when you have time.

[amoeba]

Ping @jorisvandenbossche: Any chance you have a spare moment to take a look at this again? Thanks!

[amoeba]

Sorry for the noise, I was cleaning up my fork and wiped out the commits. They've since been restored and this is ready for review.

[jorisvandenbossche]

A very tiny last comment

[amoeba]

Thanks for the review @jorisvandenbossche. I didn't mean to remove that bit and I'm not sure how I did. Good catch! All three instances of that param docstring now have the expected text.

Ready for review/merge.

[jorisvandenbossche]

Thanks!

[amoeba]

Adds an additional numypdoc check to CI (PR03) and fixes all corresponding violations.

This includes one minor breaking change to parameter order in pyarrow.parquet.read_table which makes order more uniform across ParquetDataset.__new__, read_table, and read_pandas which have related signatures.

Note this does not fully resolve ARROW-15006.

[github-actions]

https://issues.apache.org/jira/browse/ARROW-15006

[github-actions]

⚠️ Ticket has not been started in JIRA, please click 'Start Progress'.

[jorisvandenbossche]

Maybe we could rather change the order in the actual signature, since it seems useful to have skip_rows and skip_rows_after_names together?

[amoeba]

That makes sense. And other places in pyarrow and libarrow that reference both have them grouped together. Changed in 44acfb3. Thanks.

[amoeba]

Thanks for reviewing this @jorisvandenbossche. I took your suggestion on keeping skip_rows and skip_rows_after_names together. I re-ran archery numpydoc --allow-rule PR03 which still passes. This is ready for another look.

[jorisvandenbossche]

Thanks for the updates! There is a conflict with dataset.pyx now that needs to be resolved.

[jorisvandenbossche]

For this one, I would maybe also change the order of the signature instead, and make all keywords (except the first) keyword-only, using , *,

[amoeba]

I pushed a change to address this in d04b228 that:

• Re-orders the signatures and docstrings inside the Scanner class to be consistent and follow what I think is a sensible order
• Makes Scanner.from_dataset, Scanner.from_fragment, and Scanner.from_batches use the above pattern to make all args except the first keyword only

Waiting on CI.

[amoeba]

Thanks @jorisvandenbossche, the conflict should now be resolved. I made an attempt at addressing your comment so let me know if that's what you were thinking. This is ready for another look when you have time.

[amoeba]

Ping @jorisvandenbossche: Any chance you have a spare moment to take a look at this again? Thanks!

[amoeba]

Sorry for the noise, I was cleaning up my fork and wiped out the commits. They've since been restored and this is ready for review.

[jorisvandenbossche]

A very tiny last comment

[amoeba]

Thanks for the review @jorisvandenbossche. I didn't mean to remove that bit and I'm not sure how I did. Good catch! All three instances of that param docstring now have the expected text.

Ready for review/merge.

[jorisvandenbossche]

Thanks!