DataFrame could have a to_markdown method.

[petebachant]

Similar to to_latex and to_html.

[TomAugspurger]

Is there a widely agreed upon format for markdown tables? It's not in Gruber's original version, and IIRC CommonMark even punted on them.

[petebachant]

Good point. I'm not sure. Maybe the method could specify a flavor, e.g., GitHub or Pandoc?

[TomAugspurger]

I suspect that anything too complicated won't find much support here (maintaining stuff is no fun). Especially now that we have pipe. df.pipe(to_markdown) isn't much worse than df.to_markdown.

Is your usecase here to present / read the mardown, or convert it to something else? Since markdown is a superset of HTML, you may be able to get away with to_html before converting.

[petebachant]

My specific use case would be to create a DataFrame, copy/paste to GitHub flavored Markdown document (or GitHub issues, PRs, wikis, etc.), but still be able to read/edit it later without the HTML mess.

As an example, below is a table I copied from a Jupyter Qt console. First I printed the DataFrame to the terminal, copied/pasted here, then entered the dashes and pipes manually for the GH Markdown. Then I generated HTML with the to_html method, which doesn't render quite right here.

GitHub Markdown

Source

    | speed | Re_tip | Re_root | Re_ave | Re_D
---|--------|-------|--------|--------|--------
0 | 4.0e-01 | 5.0e+04 | 8.3e+04 | 6.6e+04 | 4.3e+05
1 | 6.0e-01 | 7.4e+04 |  1.2e+05 | 9.9e+04 | 6.4e+05
2 | 8.0e-01 | 9.9e+04 | 1.7e+05 | 1.3e+05 | 8.6e+05
3 | 1.0e+00 | 1.2e+05 |  2.1e+05 | 1.7e+05 | 1.1e+06
4 | 1.2e+00 | 1.5e+05 |  2.5e+05 | 2.0e+05 | 1.3e+0

Results

	speed	Re_tip	Re_root	Re_ave	Re_D
0	4.0e-01	5.0e+04	8.3e+04	6.6e+04	4.3e+05
1	6.0e-01	7.4e+04	1.2e+05	9.9e+04	6.4e+05
2	8.0e-01	9.9e+04	1.7e+05	1.3e+05	8.6e+05
3	1.0e+00	1.2e+05	2.1e+05	1.7e+05	1.1e+06
4	1.2e+00	1.5e+05	2.5e+05	2.0e+05	1.3e+0

Pandas HTML

Source

<table border="1" class="dataframe">\n  <thead>\n    <tr style="text-align: right;">\n      <th></th>\n      <th>speed</th>\n      <th>Re_tip</th>\n      <th>Re_root</th>\n      <th>Re_ave</th>\n      <th>Re_D</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>4.0e-01</td>\n      <td>5.0e+04</td>\n      <td>8.3e+04</td>\n      <td>6.6e+04</td>\n      <td>4.3e+05</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>6.0e-01</td>\n      <td>7.4e+04</td>\n      <td>1.2e+05</td>\n      <td>9.9e+04</td>\n      <td>6.4e+05</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>8.0e-01</td>\n      <td>9.9e+04</td>\n      <td>1.7e+05</td>\n      <td>1.3e+05</td>\n      <td>8.6e+05</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>1.0e+00</td>\n      <td>1.2e+05</td>\n      <td>2.1e+05</td>\n      <td>1.7e+05</td>\n      <td>1.1e+06</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>1.2e+00</td>\n      <td>1.5e+05</td>\n      <td>2.5e+05</td>\n      <td>2.0e+05</td>\n      <td>1.3e+06</td>\n    </tr>\n  </tbody>\n</table

Results

\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n

	speed	Re_tip	Re_root	Re_ave	Re_D
0	4.0e-01	5.0e+04	8.3e+04	6.6e+04	4.3e+05
1	6.0e-01	7.4e+04	1.2e+05	9.9e+04	6.4e+05
2	8.0e-01	9.9e+04	1.7e+05	1.3e+05	8.6e+05
3	1.0e+00	1.2e+05	2.1e+05	1.7e+05	1.1e+06
4	1.2e+00	1.5e+05	2.5e+05	2.0e+05	1.3e+06

[hayd]

One issue is what should the row headers be for MI columns/index. It seems that GH registers only the first (unless there is some syntactical trick).

For most flavors you can include html.

[jankatins]

As the new styler uses jinja (#10250), this shouldn't be too hard? I would also love this feature for knitpy, which is a markdown based format which is converted into all kind of formats (docx...pdf...html).

Up to now (and as workarounds...), I recommended tabulate to convert a DataFrame to markdown. There is also pandoc (e.g. via pypandoc), which can take the output of df.to_html() and convert that to markdown.

[jreback]

@JanSchulz yes, I could see .to_markdown() method in Styler (as better API than to put it directly in DataFrame (though could have that as well).

[jankatins]

IMO, on .styler it doesn't make sense, markdown unfortunately do not provide styles :-( I only mentioned .styler as it already (soft) requires jinja and that should make it easy to build a markdown representation...

IMO it should be DataFrame.to_markdown() and DataFrame._repr_markdown_()

I'd love DataFrame.to_markdown() too. I was surprised when it didn't work already.