matplotlib table

[arsenovic]

it would be cool if a DataFrame could be displayed on a matplotlib plot, perhaps by calling the table() function.
( i am thinking about displaying the output of DataFrame.describe() or similar (small) amount of data)

related use of the table:
http://stackoverflow.com/questions/10194482/custom-matplotlib-plot-chess-board-like-table-with-colored-cells

[cpcloud]

Can you give an example plot?

[arsenovic]

here is an example of the table feature of matplotlib, if this is thats what you mean.

http://matplotlib.org/examples/pylab_examples/table_demo.html

so i am using ipython notebook, which automatically displays DataFrames, which is great. but it would be nice to incorporate the table into a plot.

here is an example

[jreback]

you can easily do this directly with the axis

just format the 'text' with in-line matplotlib: http://stackoverflow.com/questions/8482588/putting-text-in-top-left-corner-of-matplotlib-plot

t = '$coef = %12.2f$\n$tstat = %12.2f$\n$r2 = %12.2f$\n'%(coeff,r.t_stat[pl],r.r2_adj)
ax.text(0.05, 0.95, t, transform=ax.transAxes, fontsize=7, family='monospace', multialignment = 'left', verticalalignment='top', bbox=props)

[jtratner]

btw - that is a really nice-looking plot. Would you mind posting the code you used to generate it?

We could consider consider this as a feature (or think of it like another display backend for a DataFrame), so you could convert any DataFrame into a format suitable for inserting into a matplotlib chart as text....

[arsenovic]

http://nbviewer.ipython.org/6528228

[jreback]

@arsenovic pls submit a PR for this..nice!

[arsenovic]

@jreback submit it to what exactly?

[cpcloud]

submit a pull request. see the wiki page on contributing for more details

[arsenovic]

yeah, i am familiar with contributing, i suppose i was saying its not obvious to me where in pandas this would fit

[jreback]

well...this could be an additional set of options to .plot I suppose

e.g

df.plot(....., table = { 'color' : 'blue', loc : 'bottom' })

maybe need some formatting info for the actual data, (e.g. a float_format)

should be straightforward to do

[arsenovic]

i never got around to adding this, but for anyone interested this yields nice results.

import pandas as pd
from pylab import *
df= pd.DataFrame(rand(100))
df.hist()
def draw_describe_table(df, cellFmt ='%.1e', colWidths= [.1,.2], 
                        loc = 'center right', **kwargs):

    kwargs.update({'colWidths':colWidths,'loc':loc})
    kwargs['cellText'] = array([cellFmt%k for k in df.describe().values]).reshape(-1,1)
    kwargs['rowLabels'] = df.describe().index.values
    kwargs['zorder']=100 # prevents grid from drawing ontop of table

    tab = table(**kwargs )
    return tab

draw_describe_table(df)

[jreback]

pls make a pull request for that!

and API maybe would supply a table keyword to plot/hist?