PEP 485 (math.isclose) implementation

[ncoghlan]

BPO	24270
Nosy	@rhettinger, @pfmoore, @mdickinson, @ncoghlan, @scoder, @taleinat, @skrah, @1st1
Files	math_isclose.patch: initial patch math_isclose_v2.patch: revised patch with minor documentation improvements math_isclose_v3.patch: revised patch with keyword-only parameters and refactored tests math_isclose_v4.patch: revised patch with fixed tests and reworded doc-string isclose.patch: complete patch including complex version of isclose()

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2015-06-03.08:39:22.417>
created_at = <Date 2015-05-23.13:18:03.154>
labels = ['type-feature', 'library']
title = 'PEP 485 (math.isclose) implementation'
updated_at = <Date 2015-06-03.08:39:22.416>
user = 'https://github.com/ncoghlan'

bugs.python.org fields:

activity = <Date 2015-06-03.08:39:22.416>
actor = 'berker.peksag'
assignee = 'none'
closed = True
closed_date = <Date 2015-06-03.08:39:22.417>
closer = 'berker.peksag'
components = ['Library (Lib)']
creation = <Date 2015-05-23.13:18:03.154>
creator = 'ncoghlan'
dependencies = []
files = ['39481', '39495', '39509', '39518', '39532']
hgrepos = []
issue_num = 24270
keywords = ['patch']
message_count = 29.0
messages = ['243914', '243972', '243976', '243979', '243984', '243985', '244049', '244103', '244104', '244133', '244173', '244291', '244292', '244293', '244294', '244295', '244296', '244297', '244308', '244311', '244338', '244375', '244416', '244427', '244556', '244557', '244571', '244712', '244728']
nosy_count = 11.0
nosy_names = ['rhettinger', 'paul.moore', 'mark.dickinson', 'ncoghlan', 'scoder', 'taleinat', 'stutzbach', 'Chris.Barker', 'skrah', 'python-dev', 'yselivanov']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue24270'
versions = ['Python 3.5']

[ncoghlan]

Tracking issue for the PEP-485 math.isclose() implementation: https://www.python.org/dev/peps/pep-0485/

Chris's implementation review request to python-dev: https://mail.python.org/pipermail/python-dev/2015-May/140031.html

Working repo: https://github.com/PythonCHB/close_pep

[taleinat]

I'm now working this into a patch against current default.

[taleinat]

Attached is a patch based on Chris Barker's implementation on github[1]. This includes only the C implementation, as well as tests, documentation and entries in NEWS and whatsnew.

I had to keep the (PyCFunction) cast in the module method list in Modules/mathmodule.c. That part should certainly be reviewed.

As for documentation etc., I took the best wording I could find from the PEP and the docs in the github repo, and made very slight changes only where I though they made things significantly clearer.

.. [1]: https://github.com/PythonCHB/close_pep

[scoder]

The cast is correct and required (the actual signature is determined by the METH_* flags).

Patch LGTM, FWIW.

[taleinat]

I have a question regarding complex values. The code (from Chris Barker) doesn't support complex values (only things that can be converted into doubles). However, the PEP states the following under "Non-float types":

"complex : for complex, the absolute value of the complex values will be used for scaling and comparison. If a complex tolerance is passed in, the absolute value will be used as the tolerance."

Should math.isclose() support complex values? Should an equivalent function be added to cmath? Should we just leave things as they are and remove mention of complex values from the PEP (it isn't mentioned in the docs)?

[scoder]

Eventually, I think a corresponding function should be added to cmath. math.isclose() shouldn't deal with complex values by itself (other than rejecting them as non-floatish input).

[taleinat]

Attached is a slightly revised patch.

This mostly fixes minor documentation wording and formatting issues, including those pointed out by Chris Barker on the python-dev mailing list.

Also, since it has been decided to support complex values only in a separate cmath.isclose() function, the previously included commented-out complex test cases have been removed.

[taleinat]

Significant questions brought up by Berker Peksağ in his review of the latest patch (thanks for the review!):

1. Should the tolerance parameters be keyword-only? Berker suggests that they should be. I agree.

2. Should the math.isclose() tests be split into a separate TestCase class with many separate methods? It is currently a single method which does all of the testing for math.isclose(). (Chris's original code has it separated into several TestCase classes; I consolidated it into a single method to keep in line with the current structure of the math module's tests.)

[taleinat]

Regarding the tests, I now realize that most of them should be reused for testing cmath.isclose(), which means they'll have to be defined outside of test_math.MathTests.

[taleinat]

Attached is a revised patch including changed due to the reviews by Berker and Yuri.

The significant changes from the previous patch are:

1. The "rel_tol" and "abs_tol" parameters have been made keyword-only.

2. The tests have been extracted into a separate TestCase sub-class. They are now better organized and will be easy to re-use for testing cmath.isclose when it is added (hopefully soon).

[taleinat]

Attached yet another revised version of the math.isclose() patch.

This patch fixes a problem with the tests in the previous patch which causes them to fail when the full test suite is run.

I've also slightly reworded the doc-string.

Hopefully this is ready to go in!

[taleinat]

Hopefully final patch attached. This adds cmath.isclose() along with relevant tests and documentation.

Note that cmath.isclose() rejects complex tolerances -- only the values may be complex.

[skrah]

I think users may be surprised that any two large Decimals like
"1e400000" and "1e999" are "close". In the Decimal world these
aren't infinite.