ENH: Html table export: Add of other attributes

[Gjacquenot]

The to_html method allows to specify a set of classes for table rendering

However, I would like to specify other attribute to the table markup

Typically I want to use the DataTables (table plug-in for jQuery) library, that requires another attribute for rendering table
Here is an example

https://datatables.net/release-datatables/examples/basic_init/multi_col_sort.html

In method write_result of class HTMLFormatter in pandas/core/format.py
there is this command that I would like to update with other attributes

    self.write('<table border="1" class="%s">' % ' '.join(_classes),
               indent)

I tweak it for my personal needs, and it looks like this.

    self.write('<table border="1" class="%s" id="example">' % ' '.join(_classes),
               indent)

But I would like to be more general.
I would like to give a dictionary to the to_html method that will expand and produce the line below.

Is it possible?

[jreback]

related is #3190 which proposes to do this with templates

however, I think a possible solution is as follows.

modify core/frame/to_html.py to allow it to take a custom HTMLFormatter (as a keyword argument), defaulting to the current HTMLFormatter, which is then instantiated in core/format.py/DataFrameFormatter/to_html.

Then modify HTMLFormatter to be a bit more flexible (so in stead of 'hard coding' things, would just call methods that have defaults); these can be overriden by the user via a custom sub-class of HTMLFormatter, or by passing in different defaults at construction time

e.g.

from pandas.core.format import HTMLFormatter
df.to_html(........, formatter = HTMLFormatter(.....))

and need to modify HTMLFormatter a bit to make it a bit more sub-class friendly, as well as to deal with passes in kwargs better. (as it will be instantiated before the format
is available)

class HTMLFormatter(TableFormatter):

    def __init__(self, fmt=None,.......):
          self.fmt = fmt

    def set_formatter(self, fmt):
          self.fmt = fmt

    def write_result(self, buf):
            ........
            self.write_classes(self):
            self.write(....)

lmk if you are up for this....It seems a much bigger change that it actually is, but it is necessary
in order to avoid having the user pass a gazillion number of parameters in that pandas then has to handle. This makes the most useful of parameters/attributes available for modification by simply overriding the HTMLFormatter class.

[hmeine]

Currently, the HTMLFormatter methods take a 'tags' kwarg, IMHO that should be named 'attributes' instead (in order to match expected HTML/XML terminology).

[CRP]

How about adding the scope attribute to the <td>
tag in order to allow for separate formatting of columns and index?
Currently, I can add a style such as <style>th {text-align: left}</style>, but this will affect both axes. Or does anyone know a different way to achieve the same result?

Thanks

[hmeine]

This could be done similar to #7338 , but I think in this case it may even be possible using non-trivial CSS selectors like

table.dataframe tr:firstchild { color: red }

Of course, that would not deal with hierarchical indices yet, so I am +1 on additional classes.

[CRP]

Yeah, I tried that but results are bad with MultiIndex, so this is a no go.
I understand that the pandas team is trying to keep the "output formatting" features at the minimum level, as this can be handled perfectly well by dedicated packages or style sheets, but for the latter to be really useful classes should be added wherever possible.