PEP8 should allow whitespace for tables of data

[endolith]

The purpose of PEP8 is to promote readability, but it's currently interpreted to prohibit tables like this:

                 [[          0,           0,           0,    0,          0],
                  [ 1518491136,  2147432704,  1518419200,    0,  821805056],
                  [ 2147474176,           0, -2147398656, -768, 1518498560],
                  [ 1518499328, -2147479296,  1518492928,    0, 1984015872],
                  [       -256,           0,        -512,  512, 2147481600],
                  [-1518491648,  2147433728, -1518420992, -512, 1984013056],
                  [-2147472128,           0,  2147377920, -512, 1518498048],
                  [-1518496256, -2147460096, -1518462464,  512,  821805824],
                  [          0,           0,           0,  256,       -256]]

in favor of tables without extra whitespace like this:

                 [[0, 0, 0, 0, 0],
                  [1518491136, 2147432704, 1518419200, 0, 821805056],
                  [2147474176, 0, -2147398656, -768, 1518498560],
                  [1518499328, -2147479296, 1518492928, 0, 1984015872],
                  [-256, 0, -512, 512, 2147481600],
                  [-1518491648, 2147433728, -1518420992, -512, 1984013056],
                  [-2147472128, 0, 2147377920, -512, 1518498048],
                  [-1518496256, -2147460096, -1518462464, 512, 821805824],
                  [0, 0, 0, 256, -256]])

which I think is not nearly as readable.

I think this is a misinterpretation of intent, but pycodestyle will not change this behavior, because they are trying to follow PEP8 to the letter: PyCQA/pycodestyle#289

So can PEP8 be changed to explicitly allow formatting tables of data like this? I've listed a bunch of other examples in that issue: PyCQA/pycodestyle#289 (comment)

See also spyder-ide/spyder#9955 and this PR failure.

[Rosuav]

This is an administrative repository. If you want the substance of PEP 8 to be changed, start appealing elsewhere.

And good luck.

[endolith]

sigh

Start appealing where, exactly?

[gvanrossum]

With whoever forces you to follow PEP 8 literally without allowing common sense to prevail, obviously. The PEP even says that.

[endolith]

@gvanrossum I already tried that and they closed and locked the issue:

This project also does its best to implement the rules as enumerated in the official document. We do our best to not try to interpret the intent because that's far too open to wholly subjective interpretation which would be far harder to maintain over the course of years and different maintainers with different interpretations and opinions.

I don't know where else to go. Do you agree that aligning tables of numbers with extra whitespace is better?

[gvanrossum]

I think you should follow their advice.

[endolith]

I think you should follow their advice.

Do you mean this comment?

If you were to use a different tool, like Flake8, you could disable E201, E202, and E222 and implement your own plugin that allows tables (which pycodestyle doesn't even today know how to detect and I suspect would be complex and error-prone in practice) to do whatever they want within reason while preventing everything else from violating what would otherwise be caught by E201, E202, and E222

So NumPy, SciPy, etc. should switch from pycodestyle to Flake8 for linting, and implement a plugin to allow whitespace for aligning arrays?

[Rosuav]

I don't know. Maybe ask the NumPy or SciPy people?

(a) The Python Steering Council does not decide this question.
(b) Even if it did, this git repository and associated tracker is not the place to request a change.

Please take this matter elsewhere. We cannot help you here.