Colorbar with SymmetricalLogLocator : issue when handling only negative values #7202

egaudry · 2016-09-30T07:49:13Z

Tested with matplotlib 1.4.3 and 1.5.3 built from sources - python 2.7.11 - Windows and Linux (64-bit)

To reproduce this issue, choose a set of values X so that -100<X<-10, then plot a NonUniformImage and attach a colorbar with a SymmetricalLogLocator : no tick or ticklabel will be displayed on the colorbar.

egaudry · 2016-09-30T07:51:55Z

The attached patch allows to draw ticks/ticklabels in case no decade (or only one) was selected based on vmin/vmax detected values.
There is however still an issue in the colorbar height.

Index: lib/matplotlib/colorbar.py
===================================================================
--- lib/matplotlib/colorbar.py  (revision 97805)
+++ lib/matplotlib/colorbar.py  (working copy)
@@ -592,12 +592,14 @@
         formatter.set_data_interval(*intv)

         b = np.array(locator())
-        if isinstance(locator, ticker.LogLocator):
+        if isinstance(locator, ticker.LogLocator) or isinstance(locator, ticker.SymmetricalLogLocator):
             eps = 1e-10
             b = b[(b <= intv[1] * (1 + eps)) & (b >= intv[0] * (1 - eps))]
         else:
             eps = (intv[1] - intv[0]) * 1e-10
             b = b[(b <= intv[1] + eps) & (b >= intv[0] - eps)]
+        if len(b)<2:
+            b = np.array(locator())
         ticks = self._locate(b)
         formatter.set_locs(b)
         ticklabels = [formatter(t, i) for i, t in enumerate(b)]
Index: lib/matplotlib/ticker.py
===================================================================
--- lib/matplotlib/ticker.py    (revision 97805)
+++ lib/matplotlib/ticker.py    (working copy)
@@ -1722,6 +1722,8 @@
                 a_range = get_log_range(t, -vmin + 1)
             else:
                 a_range = get_log_range(-vmax, -vmin + 1)
+                if vmin<0 and vmax<0:
+                  a_range = ( a_range[0], a_range[1]+1 )
         else:
             a_range = (0, 0)

egaudry · 2016-09-30T07:56:08Z

might be linked to #7146

tacaswell · 2016-10-03T12:56:07Z

@egaudry Can you open a pull request with that patch?

attn @efiring

NelleV · 2016-10-24T15:43:06Z

import numpy as np
import matplotlib.pyplot as plt
from matplotlib import colors
from matplotlib import ticker

X = -1 * np.random.randint(10, 90, size=(100, 100))

fig, ax = plt.subplots()
m = ax.imshow(X, norm=colors.SymLogNorm(1))
formatter = ticker.LogFormatter(10)
cb = fig.colorbar(m, format=formatter)
ax.set_title("label only base")

tacaswell · 2016-10-31T19:37:11Z

import numpy as np
import matplotlib.pyplot as plt
from matplotlib import colors
from matplotlib import ticker

plt.ion()

X = np.random.randint(15, 90, size=(100, 100))

fig, ax = plt.subplots()
m = ax.imshow(X, norm=colors.LogNorm())
formatter = ticker.LogFormatter(10)
cb = fig.colorbar(m, format=formatter)
ax.set_title("label only base")

NelleV · 2016-10-31T20:54:35Z

So… The behaviour is not the same for log and symlog Norm… minorticks appear with symlog but not log.
I'll do more tests.

NelleV · 2016-11-05T20:34:56Z

@efiring @tacaswell @dopplershift

Following monday's discussion, I've ran some tests. By default, Log formatter behave extremely poorly with colorbars. The current behaviour is as follows

Disabling labelOnlyBase does not change anything:

Disabling the sublabels only feature (introduced in august) does not change anything:

To have a behaviour that seems reasonable for colorbars we need to modify both the sublabels filtering and the labelOnlyBase:

Now going to using the LogFormatter on a log scaled plot while disabling the sublabels filtering does not change anything on this example: http://matplotlib.org/devdocs/examples/scales/scales.html

With sublabels:

Without sublabels:

From our discussion, I undersand that the labelOnlyBase=True is mostly use for minor ticks. I propose to change the default to have labelOnlyBase = False to remove or disable the sublabel filtering by default. I have yet to find a case where this new filtering improved the readibility of the figure, and I believe this may be due to improvements in the defaults of max number of ticks and placements of those ticks.

I have to tweak matplotlib's code to add an option to disable the sublabels, that you can find in https://github.com/NelleV/matplotlib/tree/test_7202

But here are some tests scripts I wrote:

1. Colorbars

import numpy as np

import matplotlib.pyplot as plt
from matplotlib import ticker, colors

disable_sublabel = True


X_1 = np.random.randint(10000, 90000, size=(100, 100))
X_2 = np.random.randint(10000, 1000000, size=(100, 100))

formatter = ticker.LogFormatter(
    10, disable_sublabel=disable_sublabel,
    labelOnlyBase=False)

X = [X_1, X_2]

fig, axes = plt.subplots(ncols=2, figsize=(10, 4))
for i, x in enumerate(X):
    ax = axes[i]
    m = ax.imshow(x, norm=colors.SymLogNorm(1))
    ax.set_title("Disabling sublabels & labelOnlyBase")
    cb = fig.colorbar(m, format=formatter, ax=ax, shrink=0.9, )
fig.savefig("images/log_colorbar_proposed_new_default.png")

formatter = ticker.LogFormatter(
    10, disable_sublabel=disable_sublabel,
    labelOnlyBase=True)

fig, axes = plt.subplots(ncols=2, figsize=(10, 4))
for i, x in enumerate(X):
    ax = axes[i]
    m = ax.imshow(x, norm=colors.SymLogNorm(1))
    ax.set_title("Disable sublabel only")
    cb = fig.colorbar(m, format=formatter, ax=ax, shrink=0.9)
fig.savefig("images/log_colorbar_disable_sublabels_only.png")

formatter = ticker.LogFormatter(10, labelOnlyBase=False)

fig, axes = plt.subplots(ncols=2, figsize=(10, 4))
for i, x in enumerate(X):
    ax = axes[i]
    m = ax.imshow(x, norm=colors.SymLogNorm(1))
    ax.set_title("Log formatter labelOnlyBase=False")
    cb = fig.colorbar(m, ax=ax, format=formatter, shrink=0.9)
fig.savefig("images/log_colorbar_disable_labelonlybase.png")

formatter = ticker.LogFormatter(10)

fig, axes = plt.subplots(ncols=2, figsize=(10, 4))
for i, x in enumerate(X):
    ax = axes[i]
    m = ax.imshow(x, norm=colors.SymLogNorm(1))
    ax.set_title("Current Default")
    cb = fig.colorbar(m, ax=ax, format=formatter, shrink=0.9)
fig.savefig("images/log_colorbar_default.png")

2. the scales example

import numpy as np
from matplotlib import ticker
import matplotlib.pyplot as plt

disable_sublabel = True

old_formatter = ticker.LogFormatterSciNotation(
    10, labelOnlyBase=False)
formatter = ticker.LogFormatterSciNotation(
    10, disable_sublabel=True,
    labelOnlyBase=False)

np.random.seed(19680801)
# make up some data in the interval ]0, 1[
y = np.random.normal(loc=0.5, scale=0.4, size=1000)
y = y[(y > 0) & (y < 1)]
y.sort()
x = np.arange(len(y))

for i, fmt in enumerate([old_formatter, formatter]):
    # plot with various axes scales
    fig, axs = plt.subplots(2, 2)

    # linear
    ax = axs[0, 0]
    ax.plot(x, y)
    ax.set_yscale('linear')
    ax.set_title('linear')
    ax.grid(True)

    # log
    ax = axs[0, 1]
    ax.plot(x, y)
    ax.set_yscale('log')
    ax.set_title('log')
    ax.yaxis.set_major_formatter(fmt)

    ax.grid(True)

    # symmetric log
    ax = axs[1, 1]
    ax.plot(x, y - y.mean())
    ax.set_yscale('symlog', linthreshy=0.05)
    ax.set_title('symlog')
    ax.yaxis.set_major_formatter(fmt)
    ax.grid(True)

    # logit
    ax = axs[1, 0]
    ax.plot(x, y)
    ax.set_yscale('logit')
    ax.set_title('logit')
    #ax.yaxis.set_major_formatter(formatter)
    ax.yaxis.set_minor_formatter(ticker.NullFormatter())
    ax.grid(True)

    plt.show()
    if i == 0:
        fig.savefig("images/example_old_formatter.png")
    else:
        fig.savefig("images/example_disabling_sublabels.png")

3. some random tests

import numpy as np

import matplotlib.pyplot as plt
from matplotlib import ticker, colors

formatter = ticker.LogFormatterSciNotation(
    10, disable_sublabel=True,
    labelOnlyBase=False)

X_1 = np.random.randint(10000, 90000, size=(1000,))
X_1.sort()
X_2 = np.random.randint(1e5, 1e12, size=(1000,))
X_2.sort()
X_3 = np.random.randint(1e5, 1e16, size=(1000,))
X_3.sort()

X = [X_1, X_2, X_3]

fig, ax = plt.subplots(ncols=len(X), nrows=2, figsize=(12, 6))
for i, x in enumerate(X):
    ax[0, i].plot(x)
    ax[0, i].set_yscale("symlog")

    ax[1, i].plot(x)
    ax[1, i].set_yscale("symlog")
    ax[1, i].yaxis.set_major_formatter(formatter)


ax[0, 1].set_title("Symmetrical log")

fig, ax = plt.subplots(ncols=len(X), nrows=2, figsize=(12, 6))
for i, x in enumerate(X):
    ax[0, i].plot(x)
    ax[0, i].set_yscale("log")
    ax[1, i].plot(x)
    ax[1, i].set_yscale("log")
    ax[1, i].yaxis.set_major_formatter(formatter)

ax[0, 1].set_title("Log")
l_f = ax[0, 1].yaxis.get_major_formatter()

NelleV · 2016-11-05T21:08:40Z

Note that by default, labelOnlyBase is True for LogFormatter but False for LogFormatterSciNotation.

egaudry mentioned this issue Sep 30, 2016

symlog scale no longer shows labels on the negative side #7146

Closed

egaudry mentioned this issue Oct 3, 2016

Colorbar with SymmetricalLogLocator now handles negative values (Pull request issue for #7202) #7211

Closed

NelleV mentioned this issue Oct 24, 2016

FIX symlog scale now shows negative labels. #7337

Merged

NelleV added the Release critical For bugs that make the library unusable (segfaults, incorrect plots, etc) and major regressions. label Oct 25, 2016

NelleV added this to the 2.0 (style change major release) milestone Oct 25, 2016

This was referenced Nov 5, 2016

[WIP] FIX log formatters now work with colorbars #7414

Closed

ENH ticks sublabel display is now an option #7438

Merged

tacaswell closed this as completed in 9b98582 Dec 2, 2016

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Colorbar with SymmetricalLogLocator : issue when handling only negative values #7202

Colorbar with SymmetricalLogLocator : issue when handling only negative values #7202

egaudry commented Sep 30, 2016

egaudry commented Sep 30, 2016 •

edited

Loading

egaudry commented Sep 30, 2016

tacaswell commented Oct 3, 2016

NelleV commented Oct 24, 2016 •

edited

Loading

tacaswell commented Oct 31, 2016

NelleV commented Oct 31, 2016

NelleV commented Nov 5, 2016

NelleV commented Nov 5, 2016

Colorbar with SymmetricalLogLocator : issue when handling only negative values #7202

Colorbar with SymmetricalLogLocator : issue when handling only negative values #7202

Comments

egaudry commented Sep 30, 2016

egaudry commented Sep 30, 2016 • edited Loading

egaudry commented Sep 30, 2016

tacaswell commented Oct 3, 2016

NelleV commented Oct 24, 2016 • edited Loading

tacaswell commented Oct 31, 2016

NelleV commented Oct 31, 2016

NelleV commented Nov 5, 2016

NelleV commented Nov 5, 2016

egaudry commented Sep 30, 2016 •

edited

Loading

NelleV commented Oct 24, 2016 •

edited

Loading