[Matplotlib-users] Axes array for subplots

[Benjamin Root]

Virgil, I will be very specific: The following all references the copy of
Plotting_Test03.py you sent me.

At line 263, you have a comment: "# The Axes array (axarr) will be a list
of NSub objects  (an axis for each subplot)". This is incorrect. `axarr` is
not a list, but a numpy array *or scalar*. When plt.subplots(1) is called,
you get a numpy scalar, which is the problem that you originally approached
us about. When plt.subplots(2) is called, you get a 1D numpy array, which
looks a lot like a list.

However, if plt.subplots(2, squeeze=False) is called, then you get a *2D*
numpy array of axes, of shape (2, 1). That is because plt.subplots(), in
the general case, is used for specifying the number of rows and columns of
subplots, so without squeezing, the axes array is 2D.

Therefore, if you have squeeze=False turned on in plt.subplots(), then line
306 needs to change to `ax = axarr[j, 0]`. Similar to line 318, and the
area around line 331.

Ben Root



On Tue, Apr 17, 2018 at 2:26 PM, Virgil Stokes <vs at it.uu.se> wrote:

> ax is used in a function that references axarr. My problem is not about 2D
> arrays.
>
> --V
>
> On 2018-04-17 17:22, Benjamin Root wrote:
>
> Virgil,
>
> How did you get from `axarr` to `ax`? The error message suggests that you
> haven't fully indexed the result. Remember, with squeeze=False, `axarr`
> will be a 2-D array, requiring two indices.
>
> Ben
>
> On Tue, Apr 17, 2018 at 11:14 AM, Virgil Stokes <vs at it.uu.se> wrote:
>
>> Ok Ben,
>>
>> I tried the following per your suggestion:
>>
>>   fig, axarr = plt.subplots(NSub, figsize=(width,height), squeeze=False,
>> sharex=True)
>>
>> but when the following is executed:
>>
>>   ax.grid(True)
>>
>> I get the following error message:
>>   builtins.AttributeError: 'numpy.ndarray' object has no attribute 'grid'
>>
>> But, thanks for your help :-).
>>
>> On 2018-04-17 16:30, Benjamin Root wrote:
>>
>> By default, `squeeze` is called on the array prior to returning it. This
>> way, users don't need to deal with 2D arrays when most of the time, they
>> are dealing with 1D setups. You can specify squeeze=False to subplots to
>> turn this behavior off and always have a 2D array.
>>
>> I hope that helps!
>> Ben Root
>>
>>
>> On Tue, Apr 17, 2018 at 10:12 AM, Virgil Stokes <vs at it.uu.se> wrote:
>>
>>> The following line is part of a much larger python (3.6) with Matplotlib
>>> (2.2.2) program in which the number of subplots is determined from input
>>> data:
>>>
>>>   fig, axarr = plt.subplots(NSub, figsize=(width,height), sharex=True)
>>>
>>> This works fine when the number of subplots (NSub) is greater than 1.
>>> For example when NSub=3, axarr is an array of length 3 and contains:
>>>
>>>   array([<matplotlib.axes._subplots.AxesSubplot object at
>>> 0x0000025900E9DDD8>,
>>>          <matplotlib.axes._subplots.AxesSubplot object at
>>> 0x00000259011897F0>,
>>>          <matplotlib.axes._subplots.AxesSubplot object at
>>> 0x00000259011C7128>],
>>>         dtype=object)
>>>
>>> However, when NSub=1, axarr contains:
>>>
>>>   <matplotlib.axes._subplots.AxesSubplot object at 0x00000161AB26AE80>
>>>
>>> and of course, will give an error if axarr is an array; i.e.
>>>
>>>   builtins.TypeError: 'AxesSubplot' object does not support indexing
>>>
>>> For my code this requires special handling because axarr is no longer an
>>> array. Why not have axarr contain:
>>>
>>>   array([<matplotlib.axes._subplots.AxesSubplot object at
>>> 0x00000161AB26AE80>],dtype=object)
>>>
>>> when NSub=1. IMHO this is consistent; i.e. it is an array with a length
>>> that is equal to the number of subplots.
>>>
>>> _______________________________________________
>>> Matplotlib-users mailing list
>>> Matplotlib-users at python.org
>>> https://mail.python.org/mailman/listinfo/matplotlib-users
>>>
>>>
>>
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/matplotlib-users/attachments/20180417/ec472126/attachment-0001.html>