feature: custom axis labeling from kernel side (tick_labels).

[maartenbreddels]

This allows any custom labeling from the kernel side, and code reuse of labeling code for bqplot and matplotlib.
While I think it's always good to have more scales, there is always a use case of not being able to let the scales+axis do what you want.
Using tick_values was already useful, but in combination with the new tick_labels (dict that maps values to labels) users have ultimate freedom to position and format the ticks on that axis.

Listening to min and max on the scales, we can update the ticks when we pan (no code example for that yet).

e.g.:

# given a locator and formatter
import numpy as np

def locator(xmin, xmax, ticks=5):
    # Given view interval, returns the ticks to show in data coordinates
    return (xmin + (xmax - xmin) * np.linspace(0, 1, ticks) ** 0.5).astype(int)

def formatter(x, pos=None):
    # Given x values in data coordinates, return the label to show
    return f'{100 * int(x):.1f}nm'

We can use that for matplotlib (code by @astrofrog ):

import numpy as np

def locator(xmin, xmax, ticks=5):
    # Given view interval, returns the ticks to show in data coordinates
    return (xmin + (xmax - xmin) * np.linspace(0, 1, ticks) ** 0.5).astype(int)

def formatter(x, pos=None):
    # Given x values in data coordinates, return the label to show
    return f'{100 * int(x):.1f}nm'

And also use that in bqplot (yes, react-ipywidget code 😛 )

import react_ipywidgets as react
import react_ipywidgets.bqplot as bq
import react_ipywidgets.ipywidgets as w
import bqplot


@react.component
def Plot():
    ticks, set_ticks = react.use_state(4)

    y = data = np.random.random((64, ))
    x = np.arange(len(y))

    x_min = 0
    x_max = len(data) - 1
    # x is in 'data' coordinates
    x_scale = bq.LinearScale(allow_padding=False, min=x_min, max=x_max)

    # use the same locator as matplotlib code
    tick_values = locator(x_min, x_max, ticks)
    ymax = None
    y_scale = bq.LinearScale(min=0, max=ymax)
    display_legend = True
    color = 'red'
    label = 'test'

    lines = bq.Lines(x=x, y=y, scales={"x": x_scale, "y": y_scale}, stroke_width=3, colors=[color], display_legend=display_legend, labels=[label])

    # same formatted as matplotlib code
    tick_labels = {k:formatter(k) for k in tick_values}
    # using .element because the wrapper does not know tick_labels yet
    x_axis = bqplot.Axis.element(scale=x_scale, tick_labels=tick_labels, tick_values=tick_values)
    y_axis = bq.Axis(scale=y_scale, orientation="vertical")
    with w.VBox() as main:
        w.IntSlider(value=ticks, on_value=set_ticks, description="Ticks")
        bq.Figure(axes=[x_axis, y_axis],
                  marks=[lines],
                  scale_x=x_scale,
                  scale_y=y_scale,
                  layout={"min_width": "800px"})
    return main
        
Plot()

Giving:

See also the notebook example, where I added roman numbering:

[martinRenou]

Thanks!

[martinRenou]

meeseeksdev please backport to master