Two-dimensional scatter plot

[nvaytet]

This PR introduces a new 2d scatter plot.
Fixes #264

Example:

import plopp as pp

a = pp.data.scatter()
pp.scatter(a, cbar=True)

See docs notebook for more examples.

In addition, there has been a refactor of the interface to how we update figures:
Old:

fig.update(data_a, key='a')
fig.update(data_b, key='b')

New:

fig.update(a=data_a, b=data_b)

It made sense to do this change here to unify the different views (made a common GraphicalView base class).

Also fixes #315
Also fixes #323

[jokasimr]

What changed here? Before we didn't add a point to the values, but now we do.

[nvaytet]

I think what changed is that before we were using scipp.concat for y values, and numpy for the mask values, but now we switch to numpy earlier (because I had to resort to using np.asarray because in some cases some lists of arrays were supplied)

[jokasimr]

Aha yeah I see what you mean

[jokasimr]

I don't understand what this does, why is the widget wrapped in a VBox?

[nvaytet]

This was an annoying thing where the buttons would end up all the way to the right, and for some reason, wrapping it in a VBox helps.
However, I now see that's it's still a little broken (see #323).

[nvaytet]

I think i fixed it.

[jokasimr]

(minor) The argument name confused me here, I always assume _ is unused, so I was surprised when it was used.
Maybe we should rename it to values?

[nvaytet]

I basically wanted to avoid a name clash from happening. Like when you update a dict in python.
You can do

d.update(a=1, b=3, values=5)
d.update({'a': 1, 'b': 3, 'values': 5})

If my function signature is

def update(self, values=None, **kwargs):

then the first case will fail because it is expecting values to be a dict, not just a single value.

I don't actually know how python does it, because you can also use _ as a key:

d = {'a': 1, 'b': 3}
d.update(_=55)
print(d)
d.update({'a': 33, 'c': 6})
print(d)

{'a': 1, 'b': 3, '_': 55}
{'a': 33, 'b': 3, '_': 55, 'c': 6}

Maybe I should look at how they implemented it...

[nvaytet]

So I think a better way of implementing this is

def update(self, *args, **kwargs):
    new = kwargs
    if args:
        if len(args) > 1:
            raise TypeError('Update expected at most 1 arg.  Got {}.'.format(len(args)))
        else:
            new.update(args[0])
    ...

[jokasimr]

Yeah that makes sense. We can avoid clashes and make the name more clear by using something like: _dict_arg. That's unlikely to cause a clash and it's also a bit more clear. But I'll leave it to you to decide.

[nvaytet]

Even with *args there is no clash, you can still do view.update(args=1, a=2) and the keys will be args and a.

[jokasimr]

I think this needs a short docstring explaining what makes it different from View.

[jokasimr]

Maybe I missed something but it seems we loose this error message here. Is there another check elsewhere using the new _ndim attribute?

[nvaytet]

Yes, at the start of the update function in the GraphicalView ;-)

[jokasimr]

Looks good to me. Tested the scatter plot and played around with it a bit and it seems to work well. The changes to the update method and the to the structure seem to make a lot of sense.

The superplot was not interactive for me, do I need another package to make it work?

[nvaytet]

The superplot was not interactive for me, do I need another package to make it work?

Did you use the notebook from the docs? Did you have %matplotlib widget enabled?

[jokasimr]

The superplot was not interactive for me, do I need another package to make it work?

Did you use the notebook from the docs? Did you have %matplotlib widget enabled?

Yes I did. But when running it the first time it told me to install ipympl, so I had to do that manually before running, then it ran but was not interactive. Maybe that's related to the issue.

[nvaytet]

I think you need to skip the weird cell with

from ipywidgets import HBox

b = p.children[0].children[1].children[0]
sl = p.children[1].controls['y']['slider']
b.click()
sl.value = 20
b.click()
sl.value = 30
f = p.children[0].children[0]
f.children = [
    f.top_bar,
    HBox([f.left_bar, f.canvas.to_image(), f.right_bar]),
    f.bottom_bar,
]

which is hidden in the docs.

[jokasimr]

I think you need to skip the weird cell with

Yes that did the trick. How do you make a cell hidden in the docs?

[nvaytet]

Yes that did the trick. How do you make a cell hidden in the docs?

https://nbsphinx.readthedocs.io/en/0.9.3/hidden-cells.html