ENH: Add support for dataclasses in the DataFrame constructor

[asosnovsky]

Added support for data-classes when used in the construction of a new dataframe.

• closes Dataclass support #21910
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff

Simply put, added support to use dataclasses in the following way:

from dataclasses import dataclass

@dataclass 
class Person:
    name: str
    age: int

df = DataFrame([Person("me", 25), Person("you", 35)])

[pep8speaks]

Hello @asosnovsky! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2020-03-14 23:03:57 UTC

[TomAugspurger]

Needs a release note (1.0.rst).

A few edge cases we should have tests for

1. A mixture of dataclass types (probably takes the union of all the columns?)
2. A mixture of dataclass and non-dataclass (this probably raises. Can we ensure that the error message is reasonable)

[jorisvandenbossche]

Added some comments on the doc addition!

[jreback]

looks pretty good, some changes, ping on green.

[jreback]

can you try to type things (you can add in a TYPE_CHECKING block if need be)

[asosnovsky]

ok

[MarcoGorelli]

Hi @asosnovsky - sorry to chase you up, just wanted to ask whether you're still working on this :)

[jreback]

looks good. some minor versionadded docs need updating. pls merge master and ping on green.

[MarcoGorelli]

Hi @asosnovsky , is this active? I'm hesitant to close this as stale as it looks like it's almost done, but we haven't heard from you in over 2 months

[jreback]

this looks fine, but needs a rebase if someone wants to merge master. ping on green.

[WillAyd]

@jreback green

[jreback]

thanks @asosnovsky

[simonjayhawkins]

Thanks @asosnovsky Very nice. a few comments to follow-up.

1. I think we should have some mention of this enhancement a whatsnew.
2. Dataclass support #21910 also suggested a to_dataclasses method so maybe should not be closed, or open another issue to discuss this further.
3. Would it make sense to also add dataclass support to the Series constructor for consistency? although the field names would now be the index and dtype would be either object to maintain the field types or maybe coerced?

[asosnovsky]

Thanks @asosnovsky Very nice. a few comments to follow-up.

1. I think we should have some mention of this enhancement a whatsnew.

2. Dataclass support #21910 also suggested a to_dataclasses method so maybe should not be closed, or open another issue to discuss this further.

3. Would it make sense to also add dataclass support to the Series constructor for consistency? although the field names would now be the index and dtype would be either object to maintain the field types or maybe coerced?

Not sure if it makes sense to add a "to_dataclasses" method. Would you have to create a dataclass on the fly or feed in one?

Not sure about series either. But if we were to do that, don't series need to have a single dtype? What happens to a dataclass with a mixed bag of types ?

[mathbunnyru]

I think this feature is quite cool and worth mentioning here:
https://github.com/pandas-dev/pandas/blob/master/doc/source/whatsnew/v1.1.0.rst

[jreback]

hmm this originally did have a whats new note ; must have gotten lost

would take a patch to add one

[mathbunnyru]

@asosnovsky please do that. I'm not a native speaker and I think it'll be easier for you to describe the change :)