Skip to content

Simplify units.Registry.get_converter. #9314

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Feb 28, 2019

Conversation

anntzer
Copy link
Contributor

@anntzer anntzer commented Oct 8, 2017

The caching of the converter was disabled 10 years ago
(be73ef4) and is unlikely to ever work
as the correct converter for ndarrays depends on the type of the
contents, not on the container, so just drop the cache-related code.

When a masked array is passed, it's OK to just get the first underlying
data, even if it is masked (as it should have the same type as the
unmasked entries), for the purpose of getting the converter.

PR Summary

PR Checklist

  • Has Pytest style unit tests
  • Code is PEP 8 compliant
  • New features are documented, with examples if plot related
  • Documentation is sphinx and numpydoc compliant
  • Added an entry to doc/users/next_whats_new/ if major new feature (follow instructions in README.rst there)
  • Documented in doc/api/api_changes.rst if API changed in a backward-incompatible way

@dstansby dstansby added this to the 2.2 (next feature release) milestone Oct 8, 2017
Copy link
Member

@dstansby dstansby left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

A couple of things that I think need leaving in.


def get_converter(self, x):
'get the converter interface instance for x, or None'

if not len(self):
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think it would be good to keep this in, to avoid stepping through the following logic if there aren't any convertors registered in the first place.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think the cost of an instance check, a dict lookup and attempting to get the first element is quite small. For code like this I would like to see at least a microbenchmark showing at least some minor improvement (not necessarily a big one) before making it more complex than necessary.

converter = None
classx = getattr(x, '__class__', None)

if classx is not None:
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Shouldn't this also stay in, as there's no guarantee that unit data is a subclass of numpy array or an iterable?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should be handled by return self[type(x)], no?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah yes

@dstansby
Copy link
Member

(but I think simplifying the numpy logic is a good thing to do!)

@tacaswell tacaswell modified the milestones: v2.2, v3.0 Feb 10, 2018
@tacaswell
Copy link
Member

Please hold of on merging this for 2.2.

@jklymak
Copy link
Member

jklymak commented Jul 9, 2018

I think this should wait for the units MEP, if such is forthcoming...

@tacaswell tacaswell added this to the v3.1 milestone Jul 10, 2018
@anntzer
Copy link
Contributor Author

anntzer commented Jul 10, 2018

AFAICT this should not change any semantics.

@jklymak
Copy link
Member

jklymak commented Feb 27, 2019

Can this get a rebase?

I'm still nervous about any changes to units without more thorough testing, i.e. w/ pandas, datetime, pint, etc. OTOH, this seems a lot simpler, so I'm inclined to approve given that tghe existing tests do pass...

@anntzer
Copy link
Contributor Author

anntzer commented Feb 27, 2019

rebased

@jklymak
Copy link
Member

jklymak commented Feb 27, 2019

... good thing we added some tests 😉

@anntzer
Copy link
Contributor Author

anntzer commented Feb 27, 2019

good catches, should be fixed now.

Copy link
Member

@jklymak jklymak left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks right to me. Could have an extra comment or two, for instance after the except KeyError, but I supposes its clear enough


# If x is an array, look inside the array for data with units
"""Get the converter interface instance for *x*, or None."""
if hasattr(x, "values"):
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@ImportanceOfBeingErnest had some concerns about this way to check for pandas:
#11664 (comment)

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

One could do something like type(x).__module__.startswith("pandas.") and hasattr(x, "values") but that should be a separate PR; this PR doesn't change the way pandas testing is done.

The caching of the converter was disabled 10 years ago
(be73ef4) and is unlikely to ever work
as the correct converter for ndarrays depends on the type of the
contents, not on the container, so just drop the cache-related code.

When a masked array is passed, it's OK to just get the first underlying
data, even if it is masked (as it should have the same type as the
unmasked entries), for the purpose of getting the converter.
@anntzer
Copy link
Contributor Author

anntzer commented Feb 28, 2019

added comments

Copy link
Member

@timhoffm timhoffm left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks right, but I would be more comfortable if there were some tests.

@anntzer
Copy link
Contributor Author

anntzer commented Feb 28, 2019

There are already tests (as noted by Jody anbove, an earlier version of this failed them).

@timhoffm timhoffm merged commit cd3ed7a into matplotlib:master Feb 28, 2019
meeseeksmachine pushed a commit to meeseeksmachine/matplotlib that referenced this pull request Feb 28, 2019
@anntzer anntzer deleted the simpler-units branch February 28, 2019 17:00
@jklymak
Copy link
Member

jklymak commented Feb 28, 2019

There are some tests. I'd still argue there are not enough, but I guess we will find out what breaks...

dstansby added a commit that referenced this pull request Feb 28, 2019
…14-on-v3.1.x

Backport PR #9314 on branch v3.1.x (Simplify units.Registry.get_converter.)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants