Why does my `df.to_html()` output look unstyled?

Because it isn't a page. `to_html()` returns a bare ` ` — no ` `, no ` `, no stylesheet. Browsers render it anyway (they're lenient), which is why it looks like a raw table with 1990s borders. Use `df.style.to_html()` for a version that carries its own ` ` block.

How do I export a DataFrame to a proper HTML file?

`df.style.to_html('table.html', doctype_html=True)` emits a complete document with the generated CSS included. If you're using plain `to_html()`, wrap the returned string in your own ` …` before writing it, or you're shipping a fragment.

How do I keep the number formatting I see in the notebook?

Use the Styler, not `to_html()`. `df.style.format('{:,.2f}').background_gradient()` carries formatting and conditional colours through to the HTML; plain `to_html()` renders the raw repr and drops all of it. `float_format=` on `to_html()` covers only the simplest case.

My table has 200,000 rows. What should I send?

Not all of them. `to_html(max_rows=…, max_cols=…)` truncates with an ellipsis row, which is usually the honest thing for a shared table. If people genuinely need to explore it, send the aggregate as the report and attach the CSV — a browser rendering a 200k-row table is a hung tab.

Can someone comment on one cell of the table?

Yes, once it's published. Threads anchor to the rendered content, so 'this total is double-counting refunds' sits on that cell instead of arriving as a screenshot with a red circle on it.

Share a pandas DataFrame as HTML

The table is right there in the notebook and someone wants it. The export step has one trap in it.

`to_html()` gives you a fragment

df.to_html("table.html")

That writes exactly this:

<table border="1" class="dataframe">
  <thead>
    …
  </thead>
  <tbody>
    …
  </tbody>
</table>

No <html>, no <head>, no CSS. Browsers are forgiving enough to render it, which is why the result looks like an unstyled table rather than an error — and why it looks nothing like what you saw in the notebook.

`Styler` gives you a page

styled = (
    df.style
      .format({"revenue": "${:,.0f}", "margin": "{:.1%}"})
      .background_gradient(subset=["margin"], cmap="RdYlGn")
      .hide(axis="index")
)
styled.to_html("table.html", doctype_html=True)

Styler.to_html() emits the generated <style> block with the table, and doctype_html=True wraps it in a real document. Everything you set in the notebook — number formats, conditional colouring, hidden index — comes through. This is the one to send.

Useful arguments either way:

escape=False — render HTML you built in a cell (links, badges) instead of escaping it.
max_rows= / max_cols= — truncate with an ellipsis row rather than emitting a table nobody can scroll.
index=False on to_html(), .hide(axis="index") on the Styler.

Publish it instead of sending the file

Drag table.html into Comma and you get a URL:

Renders faithfully, including your Styler CSS.
Opens on a phone, which an .html attachment does not — see someone sent me an HTML file.
Access per report: private, team, anyone signed in, or link holders with view, comment or edit — sharing & access control.

If you built a fuller page — a couple of tables, a chart, some prose — publish that page instead. Plotly figures, folium maps and profiling reports all publish the same way: Plotly HTML, folium maps, ydata-profiling.

From a script or a schedule

import requests

html = df.style.format("{:,.2f}").to_html(doctype_html=True)
requests.patch(
    f"https://commareports.com/api/v1/reports/{report_id}",
    headers={"Authorization": f"Bearer {token}"},
    json={"title": "Daily revenue by region", "html": html},
    timeout=30,
).raise_for_status()

Same id, same URL, one revision per run — so "revenue by region" is a living link rather than a new attachment every morning. Put it on a timer with scheduled HTML reports, or publish from the job that already builds it — publishing from CI.

Then: comment on the cell

Anchored threads mean a correction lands on the number it's about and survives the next run — commenting on HTML and stop screenshotting reports.

Limits

Entry HTML 5 MB — which a wide table hits at roughly tens of thousands of rows. Truncate with max_rows, or publish the aggregate. Assets: 25 MB per file, 250 MB and 500 files per report.

Try it

Comma is free — unlimited reports, unlimited commenters, unlimited revision history.

Publish a table →