Expand histogram gallery example to show usage of pattern as fill

[michaelgrund]

Description of proposed changes

Adresses #2323 and expands the histogram example to show the use of a pattern as fill.

Preview: https://pygmt-dev--2421.org.readthedocs.build/en/2421/gallery/histograms/histogram.html

Reminders

• Run make format and make check to make sure the code follows the style guide.
• Add tests for new features or tests that would have caught the bug that you're fixing.
• Add new public functions/methods/classes to doc/api/index.rst.
• Write detailed docstrings for all functions/methods.
• If wrapping a new module, open a 'Wrap new GMT module' issue and submit reasonably-sized PRs.
• If adding new functionality, add an example to docstrings or tutorials.
• Use underscores (not hyphens) in names of Python files and directories.

Slash Commands

You can write slash commands (/command) in the first line of a comment to perform
specific operations. Supported slash commands are:

• /format: automatically format and lint the code
• /test-gmt-dev: run full tests on the latest GMT development version

[michaelgrund]

/format

[seisman]

I'm OK with the changes, but I feel it would be better if we can generate two different random datasets and plot two overlapping histograms, just like @JingHuiTong's example in #2323 (comment):

The example also shows how to add legends for multiple histograms.

[michaelgrund]

I'm OK with the changes, but I feel it would be better if we can generate two different random datasets and plot two overlapping histograms, just like @JingHuiTong's example in #2323 (comment): The example also shows how to add legends for multiple histograms.

I'm fine with that changes although I'm not a big fan of overlapping histograms (even if pattern and color is used to distinguish them). What do others think @GenericMappingTools/pygmt-maintainers ?

[seisman]

I'm OK with the changes, but I feel it would be better if we can generate two different random datasets and plot two overlapping histograms, just like @JingHuiTong's example in #2323 (comment): The example also shows how to add legends for multiple histograms.

I'm fine with that changes although I'm not a big fan of overlapping histograms (even if pattern and color is used to distinguish them). What do others think @GenericMappingTools/pygmt-maintainers ?

Overlapping histograms are sometimes necessary and useful (e.g., comparing earthquake depth distributions of an initial catalog and a relocated catalog). But I agree that a gallery example for a simple histogram is still necessary.

What about keeping the current gallery example unchanged (i.e., closing this PR without merging. Sorry for that.) and having another gallery example for overlapping histograms filled with patterns?

[yvonnefroehlich]

I'm OK with the changes, but I feel it would be better if we can generate two different random datasets and plot two overlapping histograms, just like @JingHuiTong's example in #2323 (comment). The example also shows how to add legends for multiple histograms.

I'm fine with that changes although I'm not a big fan of overlapping histograms (even if pattern and color is used to distinguish them). What do others think @GenericMappingTools/pygmt-maintainers ?

Overlapping histograms are sometimes necessary and useful (e.g., comparing earthquake depth distributions of an initial catalog and a relocated catalog). But I agree that a gallery example for a simple histogram is still necessary.

What about keeping the current gallery example unchanged (i.e., closing this PR without merging. Sorry for that.) and having another gallery example for overlapping histograms filled with patterns?

In general, I prefer non-overlapping histograms. But probably it is fair to show how this histogram type can be generated.

Maybe a tutorial on histograms beside this gallery example is a solution herer? In a tutorial, we can show the different fill options for the bars, but also how to create cumulative, overlapping, grouped, stacked, horizontal etc. histograms (here as subplot to not post several figues). Maybe we can also state limitations or concerns regarding this kind of visualizations?

click to show the code

import numpy as np
import pygmt

np.random.seed(100)

# Generate random elevation data from a normal distribution
mean = 100  # mean of distribution
stddev = 20  # standard deviation of distribution
data01 = mean + stddev * np.random.randn(50)
data02 = mean + stddev*2 * np.random.randn(50)
data_merge = np.concatenate((data01, data02), axis=None)

R_histo = [0, 300, 0, 51]
bin_histo = 10

fig = pygmt.Figure()

with fig.subplot(
    nrows=2,
    ncols=3,
    figsize=("20c", "10c"),
    title="Histograms",
    frame=["WSne", "xaf10", "ya10"],
    sharex="b+lElevation in m",
    sharey="l+lConunts",
):
    
    with fig.set_panel(panel=0):
        fig.histogram(
            data=data01,
            region=R_histo,
            series=bin_histo,
            fill="red3",
            pen="1p",
            histtype=0,
            label="data01"
        )
        fig.legend()

    with fig.set_panel(panel=1):
        fig.histogram(
            data=data02,
            region=R_histo,
            series=bin_histo,
            fill="orange",
            pen="1p",
            histtype=0,
            label="data02",
        )
        fig.legend()

    with fig.set_panel(panel=2):
        fig.histogram(
            data=data01,
            region=R_histo,
            series=bin_histo,
            fill="red3",
            pen="1p",
            histtype=0,
        )
        fig.histogram(
            data=data02,
            region=R_histo,
            series=bin_histo,
            fill="orange@50",
            pen="1p",
            histtype=0,
            label="overlapping",
        )
        fig.legend()

    with fig.set_panel(panel=3):
        fig.histogram(
            data=data01,
            region=R_histo,
            series=bin_histo,
            fill="red3",
            pen="1p",
            histtype=0,
            label="data01"
        )
        fig.histogram(
            data=data01,
            region=R_histo,
            series=bin_histo,
            fill="p8+b",
            pen="1p",
            histtype=0,
            cumulative=True,
            label="cumulative"
        )
        fig.legend()

    with fig.set_panel(panel=4):
        fig.histogram(
            data=data_merge,
            region=R_histo,
            series=bin_histo,
            fill="orange",
            pen="1p",
            histtype=0,
            label="stacked",
        )
        fig.histogram(
            data=data01,
            region=R_histo,
            series=bin_histo,
            fill="red3",
            pen="1p",
            histtype=0,
        )
        fig.legend()

    with fig.set_panel(panel=5):
        fig.histogram(
            data=data01,
            region=R_histo,
            series=bin_histo,
            fill="red3",
            pen="1p",
            histtype=0,
            barwidth=str(bin_histo/2) + "+o-" + str(bin_histo/4),
        )
        fig.histogram(
            data=data02,
            region=R_histo,
            series=bin_histo,
            fill="orange",
            pen="1p",
            histtype=0,
            barwidth=str(bin_histo/2) + "+o" + str(bin_histo/4),
            label="grouped",
        )
        fig.legend()
    
fig.show()
# fig.savefig(fname="histo_versions.png")

I am fine with not expanding this gallery example. But in case we still want to do so, would this be an option? I feel here the overlapping is visually not misleading.
Code example: (so far, only code, no comments)

import numpy as np
import pygmt

np.random.seed(100)

# Generate random elevation data from a normal distribution
mean = 100  # mean of distribution
stddev = 20  # standard deviation of distribution
data = mean + stddev * np.random.randn(50)

region_histo = [0, 200, 0, 51]
bin_histo = 10

fig = pygmt.Figure()

with fig.subplot(
    nrows=1,
    ncols=2,
    figsize=("20c", "10c"),
    title="Histograms",
    frame=["WSne", "xaf10", "ya10"],
    sharex="b+lElevation in m",
    sharey="l+lConunts",
):
    
    with fig.set_panel(panel=0):
        fig.histogram(
            data=data,
            region=region_histo,
            series=bin_histo,
            fill="red3",
            pen="1p",
            histtype=0,
        )

    with fig.set_panel(panel=1):
        fig.histogram(
            data=data,
            region=region_histo,
            series=bin_histo,
            fill="red3",
            pen="1p",
            histtype=0,
            label="data",
        )
        fig.histogram(
            data=data,
            region=region_histo,
            series=bin_histo,
            fill="p8+b",
            pen="1p",
            histtype=0,
            cumulative=True,
            label="data cumulative",
        )
        fig.legend(position="jLT")
    
fig.show()
# fig.savefig(fname="histo_cumulative.png")

Output figure:

[seisman]

Maybe a tutorial on histograms beside this gallery example is a solution herer? In a tutorial, we can show the different fill options for the bars, but also how to create cumulative, overlapping, grouped, stacked, horizontal etc. histograms (here as subplot to not post several figues). Maybe we can also state limitations or concerns regarding this kind of visualizations?

A tutorial for histograms is definitely a good idea, although writing a tutorial usually takes more time than writing an example.

[yvonnefroehlich]

Maybe a tutorial on histograms beside this gallery example is a solution herer? In a tutorial, we can show the different fill options for the bars, but also how to create cumulative, overlapping, grouped, stacked, horizontal etc. histograms (here as subplot to not post several figues). Maybe we can also state limitations or concerns regarding this kind of visualizations?

A tutorial for histograms is definitely a good idea, although writing a tutorial usually takes more time than writing an example.

I would be pleased to work on such a histogram tutorial. @michaelgrund are you fine with this?

Regarding this gallery example I have currently unfortunately no better / other ideas as those I have already posted 🙁.

[yvonnefroehlich]

Maybe a tutorial on histograms beside this gallery example is a solution herer? In a tutorial, we can show the different fill options for the bars, but also how to create cumulative, overlapping, grouped, stacked, horizontal etc. histograms (here as subplot to not post several figues). Maybe we can also state limitations or concerns regarding this kind of visualizations?

A tutorial for histograms is definitely a good idea, although writing a tutorial usually takes more time than writing an example.

I would be pleased to work on such a histogram tutorial. @michaelgrund are you fine with this?

Great! I started working on a tutorial regarding cartesian histograms in PR #2445.

[weiji14]

So, what's the decision for this PR? Are we keeping the original 1-histogram example and closing this PR in favour of #2445?

[yvonnefroehlich]

So, what's the decision for this PR? Are we keeping the original 1-histogram example and closing this PR in favour of #2445?

Thanks for asking @weiji14! Hm. I am also a bit unsure at the moment 🙈. I thought, plotting the same histogram with a pattern is a bit too simple (please see #2323 (comment)). But after @seisman approved this PR now, it looked like we keep the changes as is 😂. As already mentioned, I am fine with both ways. Regarding showing a histogram with overlapping bars (please see #2421 (comment)), I posted an alternative plot (please see #2421 (comment)). Here, I feel overlaid bars are visually OK and not misleading (see comment #2421 (comment)). However, this figure, is currently also / now used in the tutorial (please see PR #2445). @michaelgrund, as this is your PR, how do you feel you / we want to continue here 🙂?

[seisman]

So, what's the decision for this PR? Are we keeping the original 1-histogram example and closing this PR in favour of #2445?

Thanks for asking @weiji14! Hm. I am also a bit unsure at the moment 🙈. I thought, plotting the same histogram with a pattern is a bit too simple (please see #2323 (comment)). But after @seisman approved this PR now, it looked like we keep the changes as is 😂. As already mentioned, I am fine with both ways. Regarding showing a histogram with overlapping bars (please see #2421 (comment)), I posted an alternative plot (please see #2421 (comment)). Here, I feel overlaid bars are visually OK and not misleading (see comment #2421 (comment)). However, this figure, is currently also / now used in the tutorial (please see PR #2445). @michaelgrund, as this is your PR, how do you feel you / we want to continue here 🙂?

Although I approved this PR, I still prefer the original 1-histogram example, which is much simpler than the current two-subplot example.

[michaelgrund]

Fine with closing this PR.