[Feature] siunitx mplstyle

[fnemina]

Hello, hope you are doing great.

There is this package siunitx which allows for a better usage of units in latex. This allows for different usage of units inside latex and to be more consistent in their use inside and outside of equations.

I believe it would be easy to implement in SciencePlots as an additional mplstyle.

Proposal

For this it could be easy to create a latex-siunitx.mplstyle inside the misc folder with the line

text.latex.preamble :\usepackage{amsmath} \usepackage{amssymb} \usepackage{siunitx}

Example

import numpy as np
import matplotlib.pyplot as plt
import scienceplots

x = np.linspace(0.75, 1.25, 201)

def model(x, p):
    return x ** (2 * p + 1) / (1 + x ** (2 * p))

With the current implementation units would be like

pparam = dict(xlabel='Voltage (mV)', ylabel=r'Current ($\mu$A)')

with plt.style.context(['science']):
    fig, ax = plt.subplots()
    for p in [10, 15, 20, 30, 50, 100]:
        ax.plot(x, model(x, p), label=p)
    ax.legend(title='Order')
    ax.autoscale(tight=True)
    ax.set(**pparam)
    fig.savefig('fig01.png', dpi=300)
    plt.close()

with the new implementation it would change to

pparam = dict(xlabel=r'Voltage (\si{\milli\volt})', ylabel=r'Current (\si{\micro\ampere})')

with plt.style.context(['science','latex-siunitx']):
    fig, ax = plt.subplots()
    for p in [10, 15, 20, 30, 50, 100]:
        ax.plot(x, model(x, p), label=p)
    ax.legend(title='Order')
    ax.autoscale(tight=True)
    ax.set(**pparam)
    fig.savefig('fig01-siunitx.png', dpi=300)
    plt.close()

Caveats

The only problem I seem to find with this is that if another style changes the text.latex.preamble line it would clash. It would be a problem for nature, russian-font, turkish-font, latex-sans, pgf and sans. Also it wouldn't work with styles that don't use latex like notebook.

Maybe this could be worked around in some way.

Best, Fran

[echedey-ls]

I feel the problem with the preambles is impossible to tackle down. It's more likely something that could be resolved on the matplotlib interface side.

BTW, matplotlib provides a way to use math text without latex that is much faster, so pursuing this change may not have a broad impact in the long term. Anyways, I don't want to discourage you. Feel free to open a PR. Don't forget to add text.usetex : True so it can be used as a stand-alone style.

[WhiteSymmetry]

Turkish font is working (latex-siunitx style has been removed).

sudo apt-get install texlive-science (for siunitx – A comprehensive (SI) units package)
apt-cache show texlive-science | grep siunitx
siunitx -- A comprehensive (SI) units package.

import numpy as np
import matplotlib.pyplot as plt
import scienceplots

#from mplfonts import use_font
#use_font('Noto Serif CJK SC')

x = np.linspace(0.75, 1.25, 201)

def model(x, p):
return x ** (2 * p + 1) / (1 + x ** (2 * p))

pparam = dict(xlabel=r'Voltaj/Gerilim (\si{\milli\volt})', ylabel=r'Akım (\si{\micro\ampere})')

with plt.style.context(['science','turkish-font']):
fig, ax = plt.subplots()
for p in [5, 7, 10, 15, 20, 30, 38, 50, 100]:
ax.plot(x, model(x, p), label=p)
ax.legend(title='Düzen: öçüğşıÖÇÜĞŞİ', fontsize=7)
ax.set(xlabel=r'Gerilim/Volt (mV)')
ax.set(ylabel=r'Mevcut Güç/Akım ($\mu$A)')
ax.autoscale(tight=True)
fig.savefig('figures/fig17.jpg', dpi=300)
plt.close()

[WhiteSymmetry]

I feel the problem with the preambles is impossible to tackle down. It's more likely something that could be resolved on the matplotlib interface side.

BTW, matplotlib provides a way to use math text without latex that is much faster, so pursuing this change may not have a broad impact in the long term. Anyways, I don't want to discourage you. Feel free to open a PR. Don't forget to add text.usetex : True so it can be used as a stand-alone style.

Average execution time of the code:

1. only Matplotlib
   0.403068066-0.67 sn & 7.97-8.08 µs ± 50.7 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)
2. no-latex, no-grid
   0.454268694-0.672160387 sn & 13.6 µs ± 108 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)
3. Science-Latex no-grid
   0.639259100-8.895558834 sn & 78.4 µs ± 404 ns per loop (mean ± std. dev. of 7 runs, 10,000 loops each)