Pandas: Pandas Series should provide to_html method

[scls19fr]

Hello,

Pandas Series provide to_json method but they doesn't provide to_html method.
Adding to_html will avoid user to convert a pandas.core.series.Series to a pandas.core.frame.DataFrame in order to output HTML file using s.to_frame().to_html(...) but directly use s.to_html(...)

see also https://github.com/pydata/pandas/issues/8825

Kind regards

[jreback]

I'll add to the list. But would be really helpful if you would submit a pull-request to fix.

[cpcloud]

@scls19fr Can you give us an example of why this is necessary? Are you converting Series to html on the command line so much that it's inconvenient? Are you calling to_html on things which may be coming in as either DataFrames or Series? The former case isn't convincing to me. The latter is more convincing, but often times when one is getting mixed Series and DataFrame output there's a way to make sure you're always getting the same type of object upstream. IMO that would be the way to solve this problem, since it takes advantage of existing APIs and doesn't require an API addition to fix the problem.

A Series isn't a Table-like thing, it's a vector like thing. As such, making users be explicit about making it a single column table is more readable IMO.

[scls19fr]

@cpcloud I'm displaying sensor values to a Python Flask app. Values of a sensor are stored in a Series, index is time.
I'm using Datatables JS library http://www.datatables.net/ to provide sorting, filter... functionalities

[cpcloud]

oh ok sounds interesting. what's wrong with calling .to_frame().to_html() then? if you only ever have Series then what's the issue? just too verbose?

[scls19fr]

@cpcloud yes quite verbose... that was just my point of view!

But anyway I was also having trouble with to_html method because id parameter is not yet supported

see https://github.com/pydata/pandas/issues/8496

I've defined this uggly function

def df_to_html_with_id(df, id, *args, **kwargs):
    s = df.to_html(*args, **kwargs)
    return s[:7] + 'id="%s" ' % id + s[7:]

so I can also do

def ts_to_html_with_id(ts, id, *args, **kwargs):
    s = ts.to_frame().to_html(*args, **kwargs)
    return s[:7] + 'id="%s" ' % id + s[7:]

or mix both using isintance and write something like

def to_html_with_id(param, id, *args, **kwargs):
    if isinstance(param, pd.core.frame.DataFrame):
        s = param.to_html(*args, **kwargs)
    elif isinstance(param, pd.core.series.Series):
        s = param.to_frame().to_html(*args, **kwargs)
    else:
        raise(NotImplementedError)
    return s[:7] + 'id="%s" ' % id + s[7:]

[bsweger]

Adding a note here because I'm working on #16180, which requests a to_latex method on Series, which will enable some desired code cleanup noted in PR #16171

If I'm interpreting correctly, implementing to_html() for a Series is the other thing that needs to happen before said cleanup. Am happy to work on this once to_latex is all squared away (PR #16465)

[dhouck]

Currently, a Series provides _repr_latex_ (if you enable the display.latex.repr option), but not _repr_html_. The Jupyter notebook will attempt to render it using MathJax, which doesn’t support the LaTeX used in the representation. This means that in order to get notebooks which export to LaTeX documents with nice tables, you need to either juggle options constantly or sacrifice the ability to read Series in HTML view.

Therefore, even if to_html is kept for DataFrames only as was argued three years ago, it would be nice to have _repr_html_.

[mrocklin]

Checking in here. Would a Series._repr_html_ implementation be welcome?

[WillAyd]

I don’t have an objection to to_html for Series since we have that same signature for many other formats

[mrocklin]

I suspect that _repr_html_ will be slightly more contentious than to_html because it will change the rendering in Jupyter notebooks, which has a greater impact on user experience.

[quasiben]

Can Series._repr_html_ call to Series.to_frame()._repr_html_ ?

[TomAugspurger]

Not quite, as we need the Series repr to be visually distinct from the
equivalent 1-column DataFrame.

Even just placing the Series and Index names at the bottom, rather than the
top, of the output table may be sufficient.

On Fri, Mar 8, 2019 at 12:49 PM Benjamin Zaitlen [e-mail]
wrote:

Can Series._repr_html_ call to Series.to_frame()._repr_html_ ?

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
https://github.com/pandas-dev/pandas/issues/8829#issuecomment-471034935,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ABQHIpWOwCu3ZYInpdvTAvG4hLwpCSmaks5vUrDWgaJpZM4C7-kN
.

[mrocklin]

There is some small amount of work here: https://github.com/pandas-dev/pandas/pull/27228