hist(..., histtype='step') does not set ylim properly.

[tkf]

Checked with the latest version from master (d5071f4).

import numpy
import pylab

x = numpy.array(
      [ 94,  95,  96,  98,  84, 104,  99,  88,  80,  90,  99,  95,  89,
        88,  94,  99,  99,  91, 100,  86, 102,  85,  84,  95,  87,  87,
       106, 109,  97,  97, 106,  93,  90,  91,  95,  93,  81,  93, 103,
        82,  94,  89, 113,  93,  95, 103,  99,  99,  93, 103,  87,  94,
        96, 104,  88,  98,  89, 117,  88,  87,  97,  91,  84,  90,  98,
        90, 100,  92, 102,  89, 104, 104,  92,  96,  88,  97, 105, 106,
       104, 105,  88,  93, 112,  82, 102,  81,  80,  86, 100, 101,  93,
        93,  91,  99,  90, 100,  93,  96,  99,  99])
pylab.hist(x, histtype='step')
pylab.savefig('fig.png')

It failed to show the lower y values. If you set ymin=0, you can see it:

pylab.ylim(ymin=0)
pylab.savefig('fig2.png')

[mdboom]

Marking as 1.3.x blocker -- hopefully the fix is simple an obvious, otherwise it might not make the cutoff, of course.

[dhyams]

I can confirm that if you stick in a relim() call right at the end of Axes.hist(), it fixes this bug:

    self.set_autoscalex_on(_saved_autoscalex)
    self.set_autoscaley_on(_saved_autoscaley)
    self.relim()    #<------------------HERE
    self.autoscale_view()

    if nx == 1:
        return n[0], bins, cbook.silent_list('Patch', patches[0])
    else:
        return n, bins, cbook.silent_list('Lists of Patches', patches)

What I cannot confirm is that this is the best way to fix this bug. I'm not entirely clear why there is funny business going on with flipping autoscaling on and off in the hist() routine. I guess there is some updating of self.dataLim that gets missed out on, because of autoscaling being turned on and off during hist()?

[mdboom]

Fixed by #2082.