Deprecate is_string_like, is_sequence_of_strings by anntzer · Pull Request #8011 · matplotlib/matplotlib

anntzer · 2017-02-03T00:01:52Z

While implementing #7996 (see also #7835), I had missed that numpy's str_ (and unicode_ on Py2) derive from the corresponding builtin type(s), so isinstance(x, six.text_type) is just as good. So I just inlined that and restored the old definition of is_string_like, immediately deprecating it.

Would also (upon removal) close #7725.

QuLogic · 2017-02-03T00:03:19Z

Has this been true since NumPy 1.7.1?

anntzer · 2017-02-03T00:11:36Z

Yes.

$ conda create -n tmpenv --yes python=3.3 numpy=1.7.1 >/dev/null && source activate tmpenv &&
python -c 'from numpy import str_, unicode_; print(str in str_.__mro__, str in unicode_.__mro__)'
True True

and likewise on Py2 (but with unicode in unicode_.__mro__)

QuLogic · 2017-02-03T01:07:56Z

Looks like an example might have broken.

anntzer · 2017-02-03T02:12:18Z

Fixed.
Axes3D.scatter now doesn't accept 2D inputs anymore (which were silently flattened before). If we want to document the change it should probably go with #7786.

I had missed that numpy's `str_` (and `unicode_` on Py2) derive from the corresponding builtin type(s), so `isinstance(x, six.string_types)` is just as good.

`Axes3D.scatter` no longer flattens a 2D color input.

anntzer · 2017-04-15T07:13:21Z

rebased. bumping :-)

QuLogic · 2017-04-15T07:37:17Z

doc/api/api_changes/2017-01-30-AL_is_numlike_stringlike.rst

-means that arrays are now never num-like or string-like regardless of their
-dtype.
+`cbook.is_numlike` now only checks that its argument is an instance of
+``(numbers.Number, np.Number)`` and.  In particular, this means that arrays are


typo, fixed

QuLogic · 2017-04-15T07:39:57Z

examples/mplot3d/2dcollections3d.py

 c_list = []
 for c in colors:
-    c_list.append([c]*20)
+    c_list.extend([c] * 20)


Is this actually related?

not really, moved out.

oh actually i remember now, it is.
this is because 3d scatter no longer implicitly flattens its input (see the last hunk of the diff).

Probably should add that to the API what's new, then.

Done. Also documented the similar change in #7786 at the same time.

QuLogic · 2017-04-15T07:41:52Z

doc/api/api_changes/2017-01-30-AL_is_numlike_stringlike.rst

+
+`cbook.is_string_like` and `cbook.is_sequence_of_strings` have been
+deprecated.  Use ``isinstance(obj, six.string_types)`` and ``all(isinstance(o,
+six.string_types) for o in obj) and not iterable(obj)`` instead.


That doesn't seem right:

In [9]: from matplotlib.cbook import iterable, is_sequence_of_strings In [10]: obj = ['a', 'b', 'c'] In [11]: is_sequence_of_strings(obj) Out[11]: True In [12]: import six In [13]: all(isinstance(o, six.string_types) for o in obj) and not iterable(obj) Out[13]: False

(using 2.0.0, though)

the "not" needs to be removed, fixed

QuLogic · 2017-04-15T07:49:42Z

lib/matplotlib/cbook/__init__.py

    seqlist = [False] * len(args)
    for i, x in enumerate(args):
-        if (not is_string_like(x)) and iterable(x) and len(x) == nrecs:
+        if ((not isinstance(x, six.string_types)) and iterable(x)


Can drop extra parentheses.

QuLogic · 2017-04-15T07:50:52Z

lib/matplotlib/cbook/__init__.py

    """
    if case:
-        names = [(name, name) for name in dir(o) if is_string_like(name)]
+        names = [(name, name) for name in dir(o) if isinstance(name, six.string_types)]


Isn't this too long?

QuLogic · 2017-04-15T07:52:17Z

lib/matplotlib/colors.py

        else:
-            if (cbook.is_string_like(self.colors) and
+            if (isinstance(self.colors, six.string_types) and
                    cbook.is_hashable(self.colors)):


Hmm, what strings are not hashable?

probably was the case of masked string arrays. removed the now redundant check.

I think there might be a few more of these that I didn't comment on.

Caught two more, should be all fixed now.

I don't see it...

There's only one line that changes is_hashable there (this one.)

Ah sorry, thought you meant overly long lines. Now I have removed the hashability checks too.

QuLogic · 2017-04-15T07:53:59Z

lib/matplotlib/lines.py

    def _get_rgba_face(self, alt=False):
        facecolor = self._get_markerfacecolor(alt=alt)
-        if is_string_like(facecolor) and facecolor.lower() == 'none':
+        if isinstance(facecolor, six.string_types) and facecolor.lower() == 'none':


Probably too long, also.

QuLogic · 2017-04-15T07:57:29Z

lib/matplotlib/text.py

-            if ((is_string_like(s1) and s1.split()[0] == "offset") or
-                  (is_string_like(s2) and s2.split()[0] == "offset")):
+            if ((isinstance(s1, six.string_types) and s1.split()[0] == "offset") or
+                  (isinstance(s2, six.string_types) and s2.split()[0] == "offset")):


Looks a bit over as well.

tacaswell · 2017-04-17T00:50:34Z

lib/matplotlib/blocking_input.py

    def __init__(self, fig, eventslist=()):
        self.fig = fig
-        if not is_sequence_of_strings(eventslist):
-            raise ValueError("Requires a sequence of event name strings")


Why remove this?

Because it's not as if the check was really useful anyways -- one can always pass a sequence of invalid event names and it'll be just as bad as of one passed in a sequence of nonstrings.

See matplotlib/matplotlib#8011

anntzer force-pushed the is_string_like branch from 130dead to 54cbb67 Compare February 3, 2017 00:02

QuLogic added this to the 2.1 (next point release) milestone Feb 3, 2017

anntzer force-pushed the is_string_like branch from 54cbb67 to f649041 Compare February 3, 2017 02:10

afvincent mentioned this pull request Mar 8, 2017

FIX: Remove type checking for strings in '_validate_linestyle' #8165

Merged

anntzer added 2 commits April 15, 2017 00:11

Completely deprecate is_string_like.

5ca32ab

I had missed that numpy's `str_` (and `unicode_` on Py2) derive from the corresponding builtin type(s), so `isinstance(x, six.string_types)` is just as good.

Deprecate is_sequence_of_strings; cleanup.

a4d6768

`Axes3D.scatter` no longer flattens a 2D color input.

anntzer force-pushed the is_string_like branch from f649041 to a4d6768 Compare April 15, 2017 07:13

QuLogic reviewed Apr 15, 2017

View reviewed changes

anntzer added 2 commits April 15, 2017 15:58

Style/typo fixes.

c773d80

Add whatsnew for scatter no longer flattening.

e12f102

anntzer force-pushed the is_string_like branch from 2109a63 to e12f102 Compare April 15, 2017 22:59

No need to check that strings are hashable.

7b05193

QuLogic approved these changes Apr 16, 2017

View reviewed changes

QuLogic changed the title ~~Deprecate is_string_like, is_sequence_of_strings~~ [MRG+1] Deprecate is_string_like, is_sequence_of_strings Apr 16, 2017

tacaswell reviewed Apr 17, 2017

View reviewed changes

tacaswell merged commit 734c09f into matplotlib:master Apr 17, 2017

anntzer deleted the is_string_like branch April 17, 2017 01:52

QuLogic changed the title ~~[MRG+1] Deprecate is_string_like, is_sequence_of_strings~~ Deprecate is_string_like, is_sequence_of_strings Apr 17, 2017

This was referenced Apr 22, 2017

Deprecate is_string_like #7835

Closed

Remove use of (deprecated) is_string_like in mplot3d. #8535

Merged

jarrodmillman added a commit to jarrodmillman/networkx that referenced this pull request Sep 10, 2017

Deprecate is_string_like

ddcd07a

See matplotlib/matplotlib#8011

jarrodmillman mentioned this pull request Sep 10, 2017

Use isinstance in is_string_like networkx/networkx#2660

Merged

jarrodmillman added a commit to jarrodmillman/networkx that referenced this pull request Sep 10, 2017

Use isinstance in is_string_like

fcbc1d2

See matplotlib/matplotlib#8011

jarrodmillman added a commit to networkx/networkx that referenced this pull request Sep 10, 2017

Use isinstance in is_string_like (#2660)

0ab38ce

See matplotlib/matplotlib#8011

tkf mentioned this pull request Dec 13, 2018

Cannot use on Windows/julia 1.0.2 JuliaPy/PyPlot.jl#414

Closed

Uh oh!

Conversation

anntzer commented Feb 3, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

QuLogic commented Feb 3, 2017

Uh oh!

anntzer commented Feb 3, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

QuLogic commented Feb 3, 2017

Uh oh!

anntzer commented Feb 3, 2017

Uh oh!

anntzer commented Apr 15, 2017

Uh oh!

Choose a reason for hiding this comment

Uh oh!

anntzer Apr 15, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

anntzer commented Feb 3, 2017 •

edited

Loading

anntzer commented Feb 3, 2017 •

edited

Loading

anntzer Apr 15, 2017 •

edited

Loading