colorbar and logscale

Community matplotlib-users
1

Hi guys,

I'm struggling with colorbar since this morning. I'd like the colorbar being logscale

I'm experimenting some strange behavior with the colorbar as some 'labels' appear several times. For instance, 10^0 appears three times, 10^1 appears also three times, and so on. I believe the exponent is the digit of the float, while I'd like to see the exponent (of the scientific notation)

I attach a png of the plot

Here is my code:

k,m,fp = np.loadtxt(file, unpack=True)
ki = linspace(k.min(), k.max(), 37)
mi = linspace(m.min(), m.max(), 37)
Z = griddata(k, m, fp, ki, mi)
Z.shape
K, M = meshgrid(ki, mi)

pcolor(K, M, log10(Z))#, cmap=cm.gray)
cbar = colorbar(format=FormatStrFormatter('$10^{%d}$'))

semilogy()
axis([1,20,10000,500000], font2)
xlabel(r'\textrm{\# hash functions ($k$)}', font)
ylabel(r'\textrm{vector size ($m$)}', font)
cbar.ax.set_ylabel(r'$f_p$', font)

I obviously suspect my code is flawed somewhere but I can't figure out where. i have tested several format for the colorbar, like LogFormatterMathText, but it does not solve my problem.

I would appreciate any kind of help. Thanks in advance

Keep on Rockin'

Benoit

toto.png

2

Benoit,

Is there any particular reason why you can’t do a log10() of the data that is being pcolor()'d and then label the colorbar as having units of dB? That would seem to be the most straight-forward approach to me.

Ben Root

2010/6/24 Benoit Donnet <benoit.donnet@…1656…>

···

Hi guys,

I’m struggling with colorbar since this morning. I’d like the colorbar being logscale

I’m experimenting some strange behavior with the colorbar as some ‘labels’ appear several times. For instance, 10^0 appears three times, 10^1 appears also three times, and so on. I believe the exponent is the digit of the float, while I’d like to see the exponent (of the scientific notation)

I attach a png of the plot

Here is my code:

k,m,fp = np.loadtxt(file, unpack=True)

ki = linspace(k.min(), k.max(), 37)

mi = linspace(m.min(), m.max(), 37)

Z = griddata(k, m, fp, ki, mi)

Z.shape

K, M = meshgrid(ki, mi)

pcolor(K, M, log10(Z))#, cmap=cm.gray)

cbar = colorbar(format=FormatStrFormatter(‘10^{%d}’))

semilogy()

axis([1,20,10000,500000], font2)

xlabel(r’\textrm{# hash functions (k)}', font)

ylabel(r’\textrm{vector size (m)}', font)

cbar.ax.set_ylabel(r’f_p', font)

I obviously suspect my code is flawed somewhere but I can’t figure out where. i have tested several format for the colorbar, like LogFormatterMathText, but it does not solve my problem.

I would appreciate any kind of help. Thanks in advance

Keep on Rockin’

Benoit

Dr. Benoit Donnet

Université catholique de Louvain (UCL)

Ecole Polytechnique de Louvain (EPL) - Département d’Ingénierie Informatique (INGI)

Place Sainte Barbe, 2

B-1348 Louvain-la-Neuve

Phone: +32 10 47 87 18

Fax: +32 10 45 03 45

ThinkGeek and WIRED’s GeekDad team up for the Ultimate

GeekDad Father’s Day Giveaway. ONE MASSIVE PRIZE to the

lucky parental unit. See the prize list and enter to win:

http://p.sf.net/sfu/thinkgeek-promo

Matplotlib-users mailing list

Matplotlib-users@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-users

3

Hello Benjamin,

Thanks for your reply.

Is there any particular reason why you can’t do a log10() of the data that is being pcolor()'d and then label the colorbar as having units of dB? That would seem to be the most straight-forward approach to me.

That’s what I first tested.

In that case, labels on the colorbar are the following (i don’t attach the plot to avoid spamming the entire mailing-list): 0.0, -0.4, -0.8, -1.2, -1.6, …, -3.6). It does not mean anything :s

Benoit

···

Ben Root

2010/6/24 Benoit Donnet <benoit.donnet@…1656…>

Hi guys,

I’m struggling with colorbar since this morning. I’d like the colorbar being logscale

I’m experimenting some strange behavior with the colorbar as some ‘labels’ appear several times. For instance, 10^0 appears three times, 10^1 appears also three times, and so on. I believe the exponent is the digit of the float, while I’d like to see the exponent (of the scientific notation)

I attach a png of the plot

Here is my code:

k,m,fp = np.loadtxt(file, unpack=True)

ki = linspace(k.min(), k.max(), 37)

mi = linspace(m.min(), m.max(), 37)

Z = griddata(k, m, fp, ki, mi)

Z.shape

K, M = meshgrid(ki, mi)

pcolor(K, M, log10(Z))#, cmap=cm.gray)

cbar = colorbar(format=FormatStrFormatter(‘10^{%d}’))

semilogy()

axis([1,20,10000,500000], font2)

xlabel(r’\textrm{# hash functions (k)}', font)

ylabel(r’\textrm{vector size (m)}', font)

cbar.ax.set_ylabel(r’f_p', font)

I obviously suspect my code is flawed somewhere but I can’t figure out where. i have tested several format for the colorbar, like LogFormatterMathText, but it does not solve my problem.

I would appreciate any kind of help. Thanks in advance

Keep on Rockin’

Benoit

Dr. Benoit Donnet

Université catholique de Louvain (UCL)

Ecole Polytechnique de Louvain (EPL) - Département d’Ingénierie Informatique (INGI)

Place Sainte Barbe, 2

B-1348 Louvain-la-Neuve

Phone: +32 10 47 87 18

Fax: +32 10 45 03 45

ThinkGeek and WIRED’s GeekDad team up for the Ultimate

GeekDad Father’s Day Giveaway. ONE MASSIVE PRIZE to the

lucky parental unit. See the prize list and enter to win:

http://p.sf.net/sfu/thinkgeek-promo

Matplotlib-users mailing list

Matplotlib-users@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-users

Dr. Benoit Donnet

Université catholique de Louvain (UCL)

Ecole Polytechnique de Louvain (EPL) - Département d’Ingénierie Informatique (INGI)

Place Sainte Barbe, 2

B-1348 Louvain-la-Neuve

Phone: +32 10 47 87 18

Fax: +32 10 45 03 45

4

Ah, I just noticed that.

Actually, I think I just figured out what is happening. The colorbar automatically chooses what values to display, and in your case, the values are 0.0, -0.4, -0.8, -1.2, -1.6, …, -3.6, which when turned into integers are 0, 0, 0, -1, -1, …, -3, which matches what your first plot had. All we have to do is specify the values that the colorbar should list. I am not very familiar with this aspect, but it has something to do with specifying your colormap and/or the normalizer.

Does anybody know of a good tutorial on creating colormaps and normalizers?

Ben Root

···

On Thu, Jun 24, 2010 at 8:53 AM, Benoit Donnet <Benoit.Donnet@…1656…> wrote:

Hello Benjamin,

Thanks for your reply.

Is there any particular reason why you can’t do a log10() of the data that is being pcolor()'d and then label the colorbar as having units of dB? That would seem to be the most straight-forward approach to me.

That’s what I first tested.

In that case, labels on the colorbar are the following (i don’t attach the plot to avoid spamming the entire mailing-list): 0.0, -0.4, -0.8, -1.2, -1.6, …, -3.6). It does not mean anything :s

Benoit

Ben Root

2010/6/24 Benoit Donnet <benoit.donnet@…1656…>

Hi guys,

I’m struggling with colorbar since this morning. I’d like the colorbar being logscale

I’m experimenting some strange behavior with the colorbar as some ‘labels’ appear several times. For instance, 10^0 appears three times, 10^1 appears also three times, and so on. I believe the exponent is the digit of the float, while I’d like to see the exponent (of the scientific notation)

I attach a png of the plot

Here is my code:

k,m,fp = np.loadtxt(file, unpack=True)

ki = linspace(k.min(), k.max(), 37)

mi = linspace(m.min(), m.max(), 37)

Z = griddata(k, m, fp, ki, mi)

Z.shape

K, M = meshgrid(ki, mi)

pcolor(K, M, log10(Z))#, cmap=cm.gray)

cbar = colorbar(format=FormatStrFormatter(‘10^{%d}’))

semilogy()

axis([1,20,10000,500000], font2)

xlabel(r’\textrm{# hash functions (k)}', font)

ylabel(r’\textrm{vector size (m)}', font)

cbar.ax.set_ylabel(r’f_p', font)

I obviously suspect my code is flawed somewhere but I can’t figure out where. i have tested several format for the colorbar, like LogFormatterMathText, but it does not solve my problem.

I would appreciate any kind of help. Thanks in advance

Keep on Rockin’

Benoit

Dr. Benoit Donnet

Université catholique de Louvain (UCL)

Ecole Polytechnique de Louvain (EPL) - Département d’Ingénierie Informatique (INGI)

Place Sainte Barbe, 2

B-1348 Louvain-la-Neuve

Phone: +32 10 47 87 18

Fax: +32 10 45 03 45

ThinkGeek and WIRED’s GeekDad team up for the Ultimate

GeekDad Father’s Day Giveaway. ONE MASSIVE PRIZE to the

lucky parental unit. See the prize list and enter to win:

http://p.sf.net/sfu/thinkgeek-promo

Matplotlib-users mailing list

Matplotlib-users@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-users

Dr. Benoit Donnet

Université catholique de Louvain (UCL)

Ecole Polytechnique de Louvain (EPL) - Département d’Ingénierie Informatique (INGI)

Place Sainte Barbe, 2

B-1348 Louvain-la-Neuve

Phone: +32 10 47 87 18

Fax: +32 10 45 03 45

5

Ah, I just noticed that.

Actually, I think I just figured out what is happening. The colorbar
automatically chooses what values to display, and in your case, the
values are 0.0, -0.4, -0.8, -1.2, -1.6, ..., -3.6, which when turned
into integers are 0, 0, 0, -1, -1, ..., -3, which matches what your

Right, good eye!

first plot had. All we have to do is specify the values that the
colorbar should list. I am not very familiar with this aspect, but it
has something to do with specifying your colormap and/or the normalizer.

The tick locations can be specified via the ticks kwarg to colorbar. From the docstring:

         *ticks* [ None | list of ticks | Locator object ]
                       If None, ticks are determined automatically from the
                       input.
         *format* [ None | format string | Formatter object ]
                       If None, the
                       :class:`~matplotlib.ticker.ScalarFormatter` is used.
                       If a format string is given, e.g. '%.3f', that is
                       used. An alternative
                       :class:`~matplotlib.ticker.Formatter` object may be
                       given instead.

The OP may also wish to specify the range of the colormap via the vmin and vmax kwargs to pcolor (note that everything to do with the colormap and the norm is specified in pcolor, not in the call to colorbar.

The OP probably does not need to specify both the formatter and the ticker. If it is certain that the range of values will be substantially geater than one, and integer tick values are desired, then try this:

int_ticker = MaxNLocator(nbins=6, integer=True)
cbar = colorbar(ticks=int_ticker)

nbins is the maximum number of intervals (one less than the max number of ticks).

The default formatter will print integers as integers, so no custom formatter is needed. Specifying a custom format or formatter is risky because, as illustrated by your diagnosis of the OP's original problem, it can easily lead to labels that are not accurate representations of the tick values.

Does anybody know of a good tutorial on creating colormaps and normalizers?

Custom norms are rare, and I don't think there is much in the way of documentation or examples. The code in ticker.py is probably the best starting place for learning about customizing norms; it includes subclasses of Normalize. Colormaps are illustrated fairly well in the examples. See the second example in http://matplotlib.sourceforge.net/examples/pylab_examples/contourf_demo.html for ListedColormap, see http://matplotlib.sourceforge.net/examples/pylab_examples/contourf_demo.html for examples of how to use the somewhat complicated LinearSegmentedColormap to generate a custom map, and http://matplotlib.sourceforge.net/examples/pylab_examples/show_colormaps.html for a view of built-in colormaps.

Eric

···

On 06/24/2010 04:03 AM, Benjamin Root wrote:

Ben Root

On Thu, Jun 24, 2010 at 8:53 AM, Benoit Donnet > <Benoit.Donnet@...1656... <mailto:Benoit.Donnet@…1656…>> wrote:

    Hello Benjamin,

    Thanks for your reply.

    Is there any particular reason why you can't do a log10() of the
    data that is being pcolor()'d and then label the colorbar as
    having units of dB? That would seem to be the most
    straight-forward approach to me.

    That's what I first tested.

    In that case, labels on the colorbar are the following (i don't
    attach the plot to avoid spamming the entire mailing-list): 0.0,
    -0.4, -0.8, -1.2, -1.6, ..., -3.6). It does not mean anything :s

    Benoit

    Ben Root

    2010/6/24 Benoit Donnet <benoit.donnet@...1656...
    <mailto:benoit.donnet@…1656…>>

        Hi guys,

        I'm struggling with colorbar since this morning. I'd like the
        colorbar being logscale

        I'm experimenting some strange behavior with the colorbar as
        some 'labels' appear several times. For instance, 10^0
        appears three times, 10^1 appears also three times, and so on.
         I believe the exponent is the digit of the float, while I'd
        like to see the exponent (of the scientific notation)

        I attach a png of the plot

        Here is my code:

        k,m,fp = np.loadtxt(file, unpack=True)
        ki = linspace(k.min(), k.max(), 37)
        mi = linspace(m.min(), m.max(), 37)
        Z = griddata(k, m, fp, ki, mi)
        Z.shape
        K, M = meshgrid(ki, mi)

        pcolor(K, M, log10(Z))#, cmap=cm.gray)
        cbar = colorbar(format=FormatStrFormatter('10^\{%d\}'))

        semilogy()
        axis([1,20,10000,500000], font2)
        xlabel(r'\textrm{\# hash functions (k)}', font)
        ylabel(r'\textrm{vector size (m)}', font)
        cbar.ax.set_ylabel(r'f\_p', font)

        I obviously suspect my code is flawed somewhere but I can't
        figure out where. i have tested several format for the
        colorbar, like LogFormatterMathText, but it does not solve my
        problem.

        I would appreciate any kind of help. Thanks in advance

        Keep on Rockin'

        Benoit

        ---
        Dr. Benoit Donnet
        Université catholique de Louvain (UCL)
        Ecole Polytechnique de Louvain (EPL) - Département
        d'Ingénierie Informatique (INGI)
        Place Sainte Barbe, 2
        B-1348 Louvain-la-Neuve
        Phone: +32 10 47 87 18
        Fax: +32 10 45 03 45

        ------------------------------------------------------------------------------
        ThinkGeek and WIRED's GeekDad team up for the Ultimate
        GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the
        lucky parental unit. See the prize list and enter to win:
        http://p.sf.net/sfu/thinkgeek-promo
        _______________________________________________
        Matplotlib-users mailing list
        Matplotlib-users@lists.sourceforge.net
        <mailto:Matplotlib-users@lists.sourceforge.net>
        matplotlib-users List Signup and Options

    ---
    Dr. Benoit Donnet
    Université catholique de Louvain (UCL)
    Ecole Polytechnique de Louvain (EPL) - Département d'Ingénierie
    Informatique (INGI)
    Place Sainte Barbe, 2
    B-1348 Louvain-la-Neuve
    Phone: +32 10 47 87 18
    Fax: +32 10 45 03 45

------------------------------------------------------------------------------
ThinkGeek and WIRED's GeekDad team up for the Ultimate
GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the
lucky parental unit. See the prize list and enter to win:
http://p.sf.net/sfu/thinkgeek-promo

_______________________________________________
Matplotlib-users mailing list
Matplotlib-users@lists.sourceforge.net
matplotlib-users List Signup and Options