Default LinearSpace values causes confusion

[nicole-brewer]

This issue is not technically a bug, but this issue caused me a lot of unnecessary confusion as a new user and may be worth changing, addressing, or just documenting for other users who are searching for answers.

The issue is that a the default minimum and maximum values for a LinearScale of a horizontal line are both zero. This means that the Y-axis has no relative scale, which is indicated by the presence of a single tick indicating y-value of the horizontal line, as shown below:

This technically makes sense, but this default behavior was very confusing to me as a new user. The code below is what produced the line/scatterplot shown above. I was attempting to move a point on the scatterplot and observe how far that point had traversed the y-axis.

from bqplot import *
from IPython.display import display

y = np.array([0, 0, 0])
x = np.array([-1, 0, 1])

x_sc = LinearScale()
y_sc = LinearScale(min=-1, max=1)

ax_x = Axis(label='Test X', scale=x_sc, tick_format='0.0f')
ax_y = Axis(label='Test Y', scale=y_sc,
            orientation='vertical', tick_format='0.2f')

line = Lines(x=x,
             y=y,
             scales={'x': x_sc, 'y': y_sc})

scatter = Scatter(x=x,
             y=y,
             scales={'x': x_sc, 'y': y_sc},
             enable_move=True,
             restrict_y=True)

def on_start(s, change):
    print(change)
    
def on_change(s, change):
    print(change)
    
def on_end(s, change):
    print(change)
    
scatter.on_drag_end(on_end)
scatter.on_drag(on_change)
scatter.on_drag_start(on_start)

fig = Figure(axes=[ax_x, ax_y],
             marks=[line , scatter])
display(fig)

But because the y-axis LinearScale had min=0 and max=0, this relative movement has no meaning. Not understanding this, I was very confused by the results I got on moving a scatter point.

When I finally figured out what was wrong, I realized the fix was to set a non-zero minimum and maximum to the y-axis LinearScale. With a small change to line 8 of the code above...

y_sc = LinearScale(min=-1, max=1)

I was able to obtain more sensical results on dragging the scatter plot, as shown below.

This is not a bug, but perhaps an intentional design decision to remain agnostic to infinite possible choices of scale on the y-axis. There may be other reasons also. However given my experience, I think it would be better to avoid confusion and set the default values of LinearScale to min=-1, max=1 when there is a vertical or horizontal line.