Add Series.dot method to dataframe module

[Madhu94]

Ref: #1259

• Closes Add missing methods to Series #1259
• Tests added / passed
• Passes black dask / flake8 dask

I had a few questions about how to attempt this so I am marking this WIP.

1. Should I handle the case for dot product of a dask series and a dask array? I didn't do so because a lot of the underlying infrastructure (like the _extract_meta utility) assumed the operands were all either dask series or dask dataframes.
2. Should I implement this for dask dataframe as part of this PR too or can that be done separately?
3. Should the method accept a meta kwarg too, for the case when the other operand could be a dataframe?

Any other feedback on the code would really help, thanks in advance!

[jsignell]

Thanks for opening this! I'll give you my opinionated answers :)

1. Should I handle the case for dot product of a dask series and a dask array? I didn't do so because a lot of the underlying infrastructure (like the _extract_meta utility) assumed the operands were all either dask series or dask dataframes.

I think for this first PR it is ok to not accepst dask.arrays. It is a good idea though, to try to alert the user in a friendly error message if they do pass an array.

2. Should I implement this for dask dataframe as part of this PR too or can that be done separately?

It can be done separately. It'll make this easier to review if it just does series. And then the dataframe PR will be straightforward.

3. Should the method accept a meta kwarg too, for the case when the other operand could be a dataframe?

I think all methods should accept a meta kwarg.

[Madhu94]

I think I've addressed the above issues, thank you!

I had a query about the enforce_metadata kwarg accepted by map_partitions - it doesn't seem to enforce dtype, only the structure (i.e. name attribute in case of series or columns in case of dataframes). Is there a reason dtypes aren't enforced?

[jsignell]

I had a query about the enforce_metadata kwarg accepted by map_partitions - it doesn't seem to enforce dtype

It seems like enforce_metadata is intended to enforce dtype. Can you provide an example of that not working as expected?

[Madhu94]

I had a query about the enforce_metadata kwarg accepted by map_partitions - it doesn't seem to enforce dtype

It seems like enforce_metadata is intended to enforce dtype. Can you provide an example of that not working as expected?

import dask.dataframe as dd
import pandas as pd

s = pd.Series([1, 2, 3])
ds = dd.from_pandas(s, npartitions=1)
ds.map_partitions(lambda x: x.sum(), meta=float, enforce_metadata=True).compute()

I would expect this to coerce the result to float or raise an error. The reverse doesn't seem to work either -

s = pd.Series([1.1, 2.1, 3.1])
ds = dd.from_pandas(s, npartitions=1)
ds.map_partitions(lambda x: x.sum(), meta=int, enforce_metadata=True).compute()

Nothing in the code (_apply_and_enforce utility) seems to be checking dtype - let me know if I'm missing something?

[Madhu94]

The comment inside _dot_series is misleading actually, fixing that.

[jsignell]

Sorry I let this sit for a bit. I think it looks really nice now. I just have a few suggestions of how to improve the tests.

[Madhu94]

Checking CI

[jsignell]

Thanks for sticking with this @Madhu94!

[Madhu94]

Thanks for your help @jsignell 🎉 🎉

[Madhu94]

I believe #1259 should not have been closed as a result of this since there were other methods to be implemented there. I'm sorry my first update wasn't clear. I'll be more careful.