consolidated the duplicate definitions of NA values (in parsers & IO)

[OlegShteynbuk]

xref #16534 (comment)

xref #16606

[codecov]

Codecov Report

Merging #16589 into master will decrease coverage by 0.02%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #16589      +/-   ##
==========================================
- Coverage   90.95%   90.92%   -0.03%     
==========================================
  Files         161      161              
  Lines       49276    49276              
==========================================
- Hits        44817    44805      -12     
- Misses       4459     4471      +12

Flag	Coverage Δ
#multiple	88.68% <ø> (-0.03%)	⬇️
#single	40.18% <ø> (ø)	⬆️

Impacted Files	Coverage Δ
pandas/plotting/_converter.py	63.23% <0%> (-1.82%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d298414...10cba2c. Read the comment docs.

[jreback]

this needs to stay it asserts that nothing has changed

[OlegShteynbuk]

reverted the change.

[jreback]

i think we have a routine already like this
look around

[OlegShteynbuk]

thanks, i think have found the function
_NA_VALUES = _ensure_encoded(parsers._NA_VALUES)
then even list comprehension is not needed; need to double check as haven't used Cython before

[OlegShteynbuk]

done

[OlegShteynbuk]

another function that can be used is asbytes

[jreback]

@gfyoung look ok to you?

[gfyoung]

Awesome that this works! My only question is whether we should be importing from io.common or from parsers (does it really matter, or does the former not work?)

[OlegShteynbuk]

i think it should not matter from the execution point of view as parsers import _NA_VALUES from common, "Explicit is better than implicit." common makes it explicit, on the other hand not sure what is the role of parsers? is it a gatekeeper?

[jreback]

importing from io.common is best

[OlegShteynbuk]

will change to io.common

[gfyoung]

@OlegShteynbuk : nitpicking here: I think we generally use import pandas.io.common as com. However, LGTM otherwise.

[OlegShteynbuk]

thanks, changed, was not aware of this

[OlegShteynbuk]

got the message - 'No test commands were found", local tests were ok.
not sure what happened, is there a way to repeat the tests without pushing changes?

[jreback]

ok restarted circle. ping on green.

[OlegShteynbuk]

repetition of the same default NA values in io.rst - removed;

in io.rst just after the defaults are listed there is a sentence

Although a 0-length string
'' is not included in the default NaN values list, it is still treated
as a missing value.

but '' is listed as a default and it is also in common.py is part of the defaults

should i remove this sentence?

[jreback]

can you build the docs and show a rendering of this page. I think this might generate a build warning (and may not render correctly)

@jorisvandenbossche

[OlegShteynbuk]

I have build the docs locally before commit, there were warnings, some of them might be related to python 3, i have 2.7.13 on linux ; file doc/source/style.ipynb also was a problem, but the generated html looks ok, can not attach html file to this replay

[OlegShteynbuk]

the only difference that i can see is a blank line because of the new label, but it might be not at bad thing at all, but i can remove the label and reuse the existing one that is for the heading NA Values, that will be several lines above

[jreback]

can u show a screenshot of the built page & also of the doc-string (in ipython)

[OlegShteynbuk]

[OlegShteynbuk]

have added the label and reference to the label, not the doc-string (in ipython) probably will be the next step

[jreback]

lgtm. @TomAugspurger @jorisvandenbossche

[TomAugspurger]

The sentence "By default the following values are interpreted as NaN:" can be removed. I think

See :ref:`na values const <io.navaluesconst>` below for a list of the values interpreted as NaN by default.

[jreback]

thanks @OlegShteynbuk
have a look at the built docs (prob take a bit to generate) http://pandas-docs.github.io/pandas-docs-travis/

and if anything looks amiss, pls open a PR

[OlegShteynbuk]

@jreback built docs are OK, it really took some time to generate.