[Matplotlib-devel] Catching parsing errors in Mathtext

[Antony Lee]

Parsing twice is probably OK, although I'm more puzzled by why jupyter
swallows the error, which seems like something worth fixing regardless.
Antony

On Tue, Oct 12, 2021 at 10:47 AM Jody Klymak <jklymak at uvic.ca> wrote:

> Jupyter also does this
>
> ```python
> import matplotlib.pyplot as plt
> import numpy as np
> import matplotlib as mpl
>
> fig, ax = plt.subplots()
> ax.plot(np.arange(10))
> ax.set_title(r'Boo $\[a\]$')
> plt.show()
> ```
>
> silently fails to produce a plot.
>
> Directly running as a script yields:
>
> ```
> ValueError:
> \[a\]
> ^
> Unknown symbol: \[, found '\'  (at char 0), (line:1, col:1)
> ```
>
> Is it a huge efficiency penalty to simply parse the text twice, once at
> creation, and once at draw?
>
> Cheers,   Jody
>
> On Oct 12, 2021, at  10:03 AM, Antony Lee <antony.lee at institutoptique.fr>
> wrote:
>
> Hi,
> Currently there is no "intermediate format" that would allow generic
> caching: 1) when the text object is created, we don't know yet whether it
> will ultimately be rendered on screen or to a bitmap or a vector file, and
> 2) when rendering to a vector format (pdf, ps, svg), we do cache the
> individual glyphs, but when rendering to a bitmap (screen, png), we
> directly convert the expression to a bitmap, and there is no code to
> instead convert individual glyphs to a bitmap.  In fact, I do have some
> vague project of having a list-of-glyphs intermediate format, but that'll
> require quite a bit of refactoring.
> Cheers,
> Antony
>
> On Tue, Oct 12, 2021 at 2:16 AM Raymond Osborn via Matplotlib-devel <
> matplotlib-devel at python.org> wrote:
>
>> When testing v3.5.0rc1, I had a fright when my plots suddenly disappeared
>> while attempting to change their title with ax.set_title. It turned out
>> that this occurred because of a typo in the title text; I had closed a
>> brace with a parenthesis by mistake. However, I was puzzled why this hadn’t
>> triggered an exception. In the debugger, it looks as if this is because the
>> exception was caught when calling draw_idle, but was never reraised because
>> context managers stopped it. For some reason, errors do get raised when I
>> run Pyplot in IPython with the same typo, but I haven’t worked out yet why
>> embedding Matplotlib stops them appearing. Instead, the plot just
>> disappears when I click on the PyQt canvas.
>>
>> The point of this post is too ask whether it might not be better to call
>> the pyparsing module earlier in this process so that the user is
>> immediately alerted to the invalid text rather than waiting for an
>> attempted redraw. I don’t know enough about the text rendering engine to
>> know whether there are reasons this is not possible, so I thought I would
>> ask the question. Shouldn’t any call to set_title, set_xlabel, etc,
>> immediately warn the user that the text doesn’t parse? If this is an
>> efficiency issue, couldn’t the parsing result be cached in the Text object?
>>
>> With regards,
>> Ray Osborn
>>
>>
>> _______________________________________________
>> Matplotlib-devel mailing list
>> Matplotlib-devel at python.org
>> https://mail.python.org/mailman/listinfo/matplotlib-devel
>>
> _______________________________________________
> Matplotlib-devel mailing list
> Matplotlib-devel at python.org
> https://mail.python.org/mailman/listinfo/matplotlib-devel
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.python.org/pipermail/matplotlib-devel/attachments/20211012/8548279c/attachment.html>