jupyter-book
jupyter-book copied to clipboard
executablebooks
LaTeX in pandas DataFrame does not render
Describe the problem
Pandas DataFrames support LaTeX formatting in column labels, and this is rendered properly in Jupyter Notebooks, but Jupyter Book does not seem to render this in the final HTML properly.
I've tried using various string formats, including $$f$$, $f$, r$f$, but none of those seem to work. Is there perhaps some additional option/argument I am missing?
Link to your repository or website
https://kyleniemeyer.github.io/gas-dynamics-notes/compressible-flows/isentropic.html#isentropic-relations
Steps to reproduce
- Create a Pandas DataFrame with LaTeX formatting in a column label
- Build the JupyterBook
The version of Python you're using
No response
Your operating system
No response
Versions of your packages
Jupyter Book : 0.11.3
External ToC : 0.2.3
MyST-Parser : 0.13.7
MyST-NB : 0.12.3
Sphinx Book Theme : 0.1.5
Jupyter-Cache : 0.4.3
NbClient : 0.5.4
``
### Additional context
_No response_
thanks for reporting this @kyleniemeyer.
@AakashGfude @choldgraf @chrisjsewell might this be a theme issue?
Looking at the html from the above example site should mathjax be parsing anything surrounded by $?
<div class="output text_html"><style type="text/css">
</style>
<table id="T_49793_">
<thead>
<tr>
<th class="col_heading level0 col0">$M$</th>
<th class="col_heading level0 col1">$p/p_t$</th>
<th class="col_heading level0 col2">$T/T_t$</th>
<th class="col_heading level0 col3">$A/A^*$</th>
</tr>
</thead>
<tbody>
<tr>
<td id="T_49793_row0_col0" class="data row0 col0">0.00</td>
<td id="T_49793_row0_col1" class="data row0 col1">1.000000</td>
<td id="T_49793_row0_col2" class="data row0 col2">1.000000</td>
<td id="T_49793_row0_col3" class="data row0 col3">inf</td>
</tr>
...
</table>
</div>
In case someone else runs into this, I did come up with a workaround, although it isn’t especially pretty. The idea is to force MyST-Parser to convert all the dollar-delimited math into pure-HTML-formatted math. The result isn’t as nice as MathJax formatting, but already much better than leaving TeX code in my tables.
It should be possible to tweak this to convert to MathJax math, although that might require more complex configuration of the parser.
Anyway, usage is pretty simple. If your dataframe is named df, then doing
display_dataframe_with_math(df)
in a Jupyter Notebook will create an output cell with a rendered dataframe, which also displays correctly when you build the Jupyter Book.
To use this you will need to stick the following definitions somewhere:
def display_dataframe_with_math(df, raw=False):
import re
from IPython.display import HTML
html = df.to_html()
raw_html = re.sub(r"\$.*?\$", lambda m: convert_tex_to_html(m[0], raw=True), html)
return raw_html if raw else HTML(raw_html)
def convert_tex_to_html(html, raw=False):
import io
import os
import re
import pandas as pd
from textwrap import dedent
from tempfile import NamedTemporaryFile
from contextlib import redirect_stdout
from IPython.display import HTML
from myst_parser.parsers.docutils_ import cli_html
# Manually apply the MyST parser to convert $-$ into MathJax’s HTML code
frontmatter="""
---
myst:
enable_extensions: [dollarmath, amsmath]
---
"""
with NamedTemporaryFile('w', delete=False) as f:
f.write(dedent(frontmatter).strip())
f.write("\n\n")
f.write(html)
with redirect_stdout(io.StringIO()) as sf:
cli_html([f.name])
fullhtml = sf.getvalue() # Returns a large-ish HTML with the full styling header
os.remove(f.name)
# Strip HTML headers to keep only the body with converted math
m = re.search(r'<body>\n<div class="document">([\s\S]*)</div>\n</body>', fullhtml)
raw_html = m[1].strip()
# Special case: if we provided a snippet with no HTML markup at all, don’t wrap the result
# in <p> tags
if "\n" not in html and "<" not in html:
m = re.match(r"<p>(.*)</p>", raw_html)
if m: # Match was a success: there are <p> tags we can remove
raw_html = m[1]
# Manually display the result as HTML
return raw_html if raw else HTML(raw_html)
@mmcky do you know if there has been any progress on this issue? We encountered it recently and were wondering if there was a fix as we would very much like to use pandas dataframes
@mmcky do you know if there has been any progress on this issue? We encountered it recently and were wondering if there was a fix as we would very much like to use pandas dataframes
No fix to this that I know of yet. I've been snowed under with other work. @rowanc1 do you know if this is an issue in the mystmd ecosystem?
This seems to be an upstream problem, in myst-nb or even higher up. One hack is to use DataFrame.to_markdown() and Markdown:
from IPython.display import Markdown
Markdown(df.to_markdown())
You will need to set nb_render_markdown_format though, in _config.yml
sphinx:
config:
nb_render_markdown_format: myst
or in conf.py
myst_enable_extensions = [
"dollarmath",
]
nb_render_markdown_format = "myst"
A minimal test repo can be found here: https://github.com/redeboer/test-sphinx-pandas-dataframe-rendering
