add a cumsum transform to cumulatively sum a single column

[bryevdv]

I ran across the SO question Create pie-chart using bokeh with dictionary
and thought I woudl see if it could be done with Stack but soon realized that it could not, since Stack stacks multiple entire columns element-wise. It occured to me that a cummulative sum was a very useful but missing piece that would also be trivial to implement, so I threw together this PR.

I will add docs and tests in subsequent commits.

[bryevdv]

With this PR the SO question can be achieved much more cleanly with:

from collections import Counter
from math import pi

import pandas as pd

from bokeh.palettes import Category20c
from bokeh.plotting import figure, show
from bokeh.transform import cumsum

x = Counter({
    'United States': 157,
    'United Kingdom': 93,
    'Japan': 89,
    'China': 63,
    'Germany': 44,
    'India': 42,
    'Italy': 40,
    'Australia': 35,
    'Brazil': 32,
    'France': 31,
    'Taiwan': 31,
    'Spain': 29
})

data = pd.DataFrame.from_dict(dict(x), orient='index').reset_index().rename(index=str, columns={0:'value', 'index':'country'})
data['angle'] = data['value']/sum(x.values()) * 2*pi
data['color'] = Category20c[len(x)]

p = figure(plot_height=350, title="Pie Chart", toolbar_location=None,
           tools="hover", tooltips=[("Country", "@country"),("Value", "@value")])

p.wedge(x=0, y=1, radius=0.4, 
        start_angle=cumsum('angle', include_zero=True), end_angle=cumsum('angle'),
        line_color="white", fill_color='color', legend='country', source=data)

p.axis.axis_label=None
p.axis.visible=False
p.grid.grid_line_color = None

show(p)

[mattpap]

^^^

[bryevdv]

We should really just make a code quality check that looks for this

[mattpap]

Sure, there's a built-in rule for this (https://palantir.github.io/tslint/rules/no-debugger/).

[bryevdv]

fixed in c591751

[mattpap]

If your are at fixing things, you can add "no-debugger": true to bokehjs/tslint.json.

[mattpap]

It's a part of codebase job on travis ci. Otherwise you have to run node make tslint. Perhaps we could make it part of quality tests at some point.

[mattpap]

but it doesn't currently actually fail when there is a violation:

It was supposed to get --emit-error, the same as scripts:compile and test:compile have. Must have forgotten about this or maybe there is a regression since then new build is in place.

[bryevdv]

OK I added it, tho as I said it's not clear to me that a linter failure results in a test failure since the exit always seems to be 0

[bryevdv]

I tried adding --emit-error locally, that still did not result in a nonzero exit code when a lint violation was present

[mattpap]

In fact this was done on purpose, because initially tslint reported a lot of errors and it was tedious dealing with this. Now all were resolved, so failing tslint build task on error should be implemented. I will do it with the next round of build improvements.

[philippjfr]

Do Expressions actually allow positional arguments? Might be the case but maybe you confused it with the cumsum helper function signature?

[mattpap]

No, this code is incorrect.

[mattpap]

Also, it would be good to use actual Python shell syntax (>>>).

[bryevdv]

fixed in 7f2a0dc

[mattpap]

This could have been reduced to one line as on the line before.

[bryevdv]

fixed in 0c3eaa9