How to use render.py

[TITC]

Here is my dataset demo(200 equations), I call the file use python render.py -i 200.lst -o 200.

The shell hasn't any error output, but nothing rendered at folder 200.

---

I wrote a small demo for render and it works fine with these data, but I think your code can make more variability for equations, such as font. so I wonder how to use it.

import os
from tqdm.std import tqdm
from handytools import op
current_dir = op.parent_dir(__file__)
os.chdir(current_dir)
os.makedirs(os.path.join(current_dir, 'rendered'), exist_ok=True)
formulas = op.readlines("/mnt/d/data/NLP/wikipedia/latex_1024_200.lst")
formulas = [_ for _ in formulas if len(_) != 0]
for i, formula in enumerate(tqdm(formulas, total=len(formulas))):
    latex_code = """\\documentclass[preview]{standalone}\n\\begin{document}\n\\[ %s \\]\n\\end{document}""" % formula
    op.savelist(
        [latex_code], "%s/temp_simple.tex" % current_dir)
    os.system('latex2png -d 1080 %s/temp_simple.tex' % current_dir)
    os.system("cp -r %s/temp_simple.png %s/rendered/%s.png" %
              (current_dir, current_dir, str(i)))

[lukas-blecher]

The command worked fine for me. My guess is that your there are missing some packages for the latex to compile?
This is my template

\documentclass[varwidth]{standalone}
\usepackage{fontspec,unicode-math}
\usepackage[active,tightpage,displaymath,textmath]{preview}
\setmathfont{%s}
\begin{document}
\thispagestyle{empty}
%s
\end{document}

Try to compile it and see if it throws errors. You also need to install some of the fonts used 'GFSNeohellenicMath.otf', 'Asana Math', 'XITS Math', 'Cambria Math'. These are all I could find online. I'd love to have even more though. Because I think this is one of the main reasons why this repo works much better with images found in the wild in comparison to similar projects. This and rendering the formulas with different DPIs.

Also what is the latex2png command you are using?

[lukas-blecher]

I feel like that should be a \mathbf instead of \textbf.
Interesting. Could easily be replaced in every instance I belive. No need for textbf in math code
Edit: apparently spaces are not ignored in \textbf. So not so trivial after all

[TITC]

yeah, you are right. maybe it can be added to preprocessing.

[lukas-blecher]

While that's true, it should not even happen in the first place. As you noticed yourself the code wouldn't even compile.
I think fixing latex code is too much for preprocessing. The sample in question should just be dropped.

[bruno-who-likes]

I tested both 'Latin Modern Math' and 'GFS Neohellenic Math' on some of your examples in 200.lst. They both worked.

[TITC]

sorry, it's my fault. This problem is terminated at #136 (comment). And the other part from #136 (reply in thread) is extend content, maybe you noticed this equation I mentioned at #136 (reply in thread) can't be found at 200.lst file.

$$ G ( \textbf { r } , \textbf { r ^ { \prime } } ) $$

We are not tested at the same file. And you can add this equation in the middle of 200.lst then test again. I think you can see the interrupt both 'Latin Modern Math' and 'GFS Neohellenic Math'.