Rogue mathtext rendered spaces

[ndevenish]

I'm aware that the built in mathtext rendered is not expected to be perfectly identical to tex, but I'm seeing some egregious behavior with spacing that I cannot seem to explain. With:

import matplotlib.pyplot as plt
plt.figure()
plt.figtext(0.1, 0.9, r"1. ($a^2$)", size=20)
plt.figtext(0.1, 0.8, r"2. $(a^2)$", size=20)
plt.figtext(0.1, 0.7, r"3. $\left(a^2\right)$", size=20)
plt.show()

In both 1.4.2 and a recent development version I get a rogue space before the bracket in all three cases:

Compared with turning text.usetex on:

Using a negative space (\!) fixes the issue rendering with matplotlib, but screws up the latex output.

[tacaswell]

attn @mdboom Could this be related to the recent spacing changes?

[u55]

I get slightly more-reasonable whitespace after the exponent if I enclose the exponent in curly brackets.

import matplotlib.pyplot as plt
s = 0.13
plt.figure(figsize=(3,2.5))
plt.figtext(0.1, 0.9-1*s, r"1a. ($a^2$)", size=20)
plt.figtext(0.1, 0.9-2*s, r"1b. ($a^{2}$)", size=20)
plt.figtext(0.1, 0.9-3*s, r"2a. $(a^2)$", size=20)
plt.figtext(0.1, 0.9-4*s, r"2b. $(a^{2})$", size=20)
plt.figtext(0.1, 0.9-5*s, r"3a. $\left(a^2\right)$", size=20)
plt.figtext(0.1, 0.9-6*s, r"3b. $\left(a^{2}\right)$", size=20)
plt.show()

[jason-sachs]

another example:

plt.text(0.2,0.5,"$d^2 \\leq K^2(q^2 - q_0^2)$",fontsize=30)

vs. MathJax $d^2 \leq K^2(q^2 - q_0^2)$

[WeatherGod]

In that example, I would wonder if the kerning is being turned off for
exponents, somehow.
On Apr 24, 2015 6:26 PM, "Jason Sachs" notifications@github.com wrote:

another example:

plt.text(0.2,0.5,"$d^2 \leq K^2(q^2 - q_0^2)$",fontsize=30)

[image: image]
https://cloud.githubusercontent.com/assets/10506541/7329623/193182b2-ea96-11e4-980c-ffd0a56223c5.png

vs. MathJax $d^2 \leq K^2(q^2 - q_0^2)$

[image: image]
https://cloud.githubusercontent.com/assets/10506541/7329629/2e21b156-ea96-11e4-939e-82f77ab64dd1.png

—
Reply to this email directly or view it on GitHub
#4333 (comment)
.

[ndevenish]

That's an awful-looking example also - spacing screwed up next to the K and the contraction of the spacing in front of the q_0, along with the same spacing before the close bracket.

[zblz]

@ndevenish , @jason-sachs , @u55 : I have been finetuning the super/subscript spacing in mathtext in PR #4873, and applied a fix suggested by @mdboom to the symbol spacing in PR #4872. Applying these two PRs, the examples mention in this thread render like this:

The strings are:

$d^2 \\leq K^2(q^2 - q_0^2)$
($a^2$) $(a^2)$ $\left(a^2\right)$

It would be great if you could take a look at PR #4873 and make any comments you have on how the new script kerning works before it is merged. Thanks!

[jason-s]

It would be great if you could take a look at PR #4873 and make any comments you have on how the new script kerning works before it is merged.

Scanned it briefly. You've gone much more in depth than I have -- thanks, I trust your judgment!