Setting dataframe width doesn't have effect

[alexscott-ff]

Summary

Providing st.dataframe a width doesn't seem to have the desired effect.

Steps to reproduce

Code snippet:

st.dataframe(df, width=1200)

That's the output of the table, regardless of whether I give the width argument or not.

Expected behavior:

The table width is set to the integer number of pixels and the column widths are dispersed evenly within that width.

Actual behavior:

The table doesn't change size.

Is this a regression?

Don't think so.

Debug info

• Streamlit version: 1.10
• Python version: 3.9.13
• OS version: OS X
• Browser version: Chrome latest

Additional information

I believe this is due to the columns themselves not having width and just taking enough width to fit the values. But the header values (column names) seem to get ignored when making that calculation.

I tried to play around with the CSS but due to Canvas was unable to do things easily in inspection mode in the browser.

As an aside, would really like a full-width, responsive dataframe option (potentially as default).

[LukasMasuch]

@alexscott-ff Thanks for reporting this issue.

I believe this is due to the columns themselves not having a width and just taking enough width to fit the values.

That's indeed the reason why it currently fails to set the width. Unfortunately, there is no good workaround at the moment to fix this with st.dataframe. We will soon investigate some time to improve the automatic column sizing, which I assume will also fix this issue.

[alexscott-ff]

Sounds good, thanks for the quick response @LukasMasuch !

[TheRealJoci]

Has there been any updates on this? This makes Dataframes unusable right?

[gigantic65]

Has there been any updates on this? I had same problems.

[katiaviene]

any updates? this df behavior makes my eye tick!

[astrowonk]

I have the same issue. I can use st.table which looks fine but of course I lose the interactivity.

I think if it could just take the header width into account when sizing the columns that might help, though I'd still eventually want the width parameter to have an effect.

[scott-trinkle]

+1 for this, the current implementation is unusable, I have to use st._legacy_dataframe for all dataframes instead.