hists with empty bins and log option

Community matplotlib-users
1

hello,
when histogramming a distribution in log scale, I have some empty bins, which drives the y axis to 1e-100 as a lower limit, completely squashing the histogram.... Bug or feature?
Johann

PS To be specific :
In [22]: entries, bins, patches = pl.hist(np.array(list1),bins=50,normed=1,log=True)

In [23]: entries
Out[23]:
array([ 1.18347753e+00, 4.10926113e-01, 2.53989507e-01,
         1.84039145e-01, 1.27510992e-01, 9.21486321e-02,
         7.45965117e-02, 5.08495253e-02, 4.23315845e-02,
         3.43298825e-02, 2.78768971e-02, 1.83264787e-02,
         1.62615233e-02, 1.23897320e-02, 1.26478515e-02,
         7.74358253e-03, 4.38803010e-03, 5.93674661e-03,
         3.61367185e-03, 4.38803010e-03, 2.58119418e-03,
         1.29059709e-03, 7.74358253e-04, 1.03247767e-03,
         1.54871651e-03, 2.32307476e-03, 5.16238835e-04,
         2.58119418e-04, 1.03247767e-03, 2.58119418e-04,
         5.16238835e-04, 0.00000000e+00, 2.58119418e-04,
         0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
         2.58119418e-04, 2.58119418e-04, 2.58119418e-04,
         0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
         0.00000000e+00, 2.58119418e-04])

In [24]: bins
Out[24]:
array([ 3.36150086e-09, 3.87417583e-01, 7.74835162e-01,
         1.16225274e+00, 1.54967032e+00, 1.93708790e+00,
         2.32450548e+00, 2.71192306e+00, 3.09934064e+00,
         3.48675822e+00, 3.87417580e+00, 4.26159338e+00,
         4.64901096e+00, 5.03642854e+00, 5.42384612e+00,
         5.81126370e+00, 6.19868128e+00, 6.58609885e+00,
         6.97351643e+00, 7.36093401e+00, 7.74835159e+00,
         8.13576917e+00, 8.52318675e+00, 8.91060433e+00,
         9.29802191e+00, 9.68543949e+00, 1.00728571e+01,
         1.04602746e+01, 1.08476922e+01, 1.12351098e+01,
         1.16225274e+01, 1.20099450e+01, 1.23973625e+01,
         1.27847801e+01, 1.31721977e+01, 1.35596153e+01,
         1.39470329e+01, 1.43344504e+01, 1.47218680e+01,
         1.51092856e+01, 1.54967032e+01, 1.58841208e+01,
         1.62715383e+01, 1.66589559e+01, 1.70463735e+01,
         1.74337911e+01, 1.78212087e+01, 1.82086262e+01,
         1.85960438e+01, 1.89834614e+01, 1.93708790e+01])

2

I made changes to svn trunk so that zero area rectangles no longer
influence the autoscaling. So this example scales correctly:

  import numpy as np
  import matplotlib.pyplot as plt

  x = np.random.rand(100)
  fig = plt.figure()
  ax = fig.add_subplot(111)
  n, bins, patches = ax.hist(x, 50, log=True)

  plt.show()

Thanks for the suggestion,
JDH

···

On Tue, May 27, 2008 at 6:29 AM, Johann Cohen-Tanugi <cohen@...1565...> wrote:

hello,
when histogramming a distribution in log scale, I have some empty bins,
which drives the y axis to 1e-100 as a lower limit, completely squashing
the histogram....