mpl-plotter 4.0.0

Matplotlib-based plotting library

MPL Plotter

Making plots for technical documents can be a time sink. At some point I decided I might as well rid myself of that overhead and learn some Python along the way! This library is the result of that. It does the job for me and I expand it when it can't, plus it's somwhat opinionated, but it might still do the trick!

Hope you find some use in it :)

Antonio Lopez Rivera, 2020

Table of Contents

1. Introduction

2. Install

3. Map of the library

4. Capabilities

5. Getting started

5.1 2D Lines

5.2 3D Lines

6. Base methods: examples, status

6.1 2D

6.2 3D

6.3 Plot combination examples

7. Matplotlib compatibility

7.1 Retrieving axes, figures

7.2 Using Matplotlib's axis tiling

8. Advanced: Presets and custom_canvas

8.1 Custom presets

8.2 Standard presets

8.3 custom_canvas

9. Advanced: comparison and panes

9.1 comparison

9.2 n_pane_single

9.3 n_pane_comparison

9.4 Bunch of panes

1. Introduction

MPL Plotter is a Matplotlib based Python plotting library built with the goals of achieving publication-quality plots in an efficient and comprehensive way. What follows is a user's manual of MPL Plotter. The full Python API documentation is available here.

The fundamental premise of MPL Plotter is to:

• Generate publication quality plots in a single function call
• Allow for any and all further customization with regular Matplotlib if needed

As a result, MPL Plotter is built with Matplotlib compatibility in mind: its capabilities expand when used in combination. Keep reading to see them in action!

---

There's three ways to use MPL Plotter:

• Calls to the 2D and 3D plotting classes.
• Using presets, either those shipped with the library, or custom ones.
• Calling the "decorator" custom_canvas class. This class won't plot anything, but rather allow you to create a customized canvas on which to plot using Matplotlib.

The first will be covered in Sections 4 and 5, from basic usage to in depth customization. The base output and API stability of all base methods can be seen in Section 6. The latter two, in Section 8.

Say goodbye to hours getting your plots in shape!

2. Install

pip install mpl_plotter

All dependencies will be checked for and installed automatically. They can be found in setup.py under install_requires.

PyPi

TROUBLESHOOTING: If you're upgrading to the latest version of MPL Plotter, please make sure to check your dependencies are up to date with the repo. To do so, download requirements.txt above, activate your virtual environment (if you work with one, otherwise ignore that), and

pip install -r requirements.txt

Linux

PyQt5 may fail to install in Linux, prompting the following error:

FileNotFoundError: [Errno 2] No such file or directory: '/tmp/pip-build-4d8suz7p/PyQt5/setup.py'

To solve this, make sure pip is up to date and install PyQt5 5.14.0. Check this StackOverflow answer for further reference.

pip3 install --upgrade pip
pip3 install pyqt5==5.14.0

3. Map of the library

This is the map of the library. Mostly for import reference.

• Bold: package

• Code: methods

• Plain/: directories

• mpl_plotter

  • figure
  • get_available_fonts
  • two_d
    • line
    • scatter
    • heatmap
    • quiver
    • streamline
    • fill_area
    • floating_text
    • comparison/
      • comparison
    • panes/
      • n_pane_single
      • n_pane_comparison
  • three_d
    • line
    • scatter
    • surface
    • floating_text
  • canvas
    • custom_canvas
  • presets
    • publication
      • two_d
      • three_d
    • precision
      • two_d
      • three_d
    • custom
      • two_d
      • three_d
      • generate_preset_2d
      • generate_preset_3d
    • data/
      • publication
      • precision
  • color
    • maps
      • custom
    • schemes
      • one

4. Capabilities

With a single call, you can generate the following plots:

• 2D
  • Line plots
  • Scatter plots
  • Heatmaps
  • Quiver plots
  • Streamline plots
  • Area fills
  • Floating text
• 3D
  • Line plots
  • Scatter plots
  • Surface plots
  • Floating text

Furthermore, MPL Plotter also allows to:

• Use a custom_canvas function to define a cusomized figure and axis on which to draw using Matplotlib
• Generate, customize and use 2D and 3D presets in one or many function calls
• Use the pre-made publication and precision presets to immediately obtain valuable plots
• Easily create custom linear segmented colormaps, so you can use any sequence of colors you fancy
• Custom colorschemes (currently only 1, as it's enough to fit my needs, perhaps more in the future)

---

Each plot has specific parameters which can be modified, plus general ones which apply for all 2D and 3D plots respectively. The specific parameters for each plotting class are available in the docstrings of their __init__ methods. It's comfortable to access them from the interactive Python terminal. Eg:

>>> from mpl_plotter.two_d import line
>>> help(line)

In Section 11 at the end of this README, all general parameters for 2D and 3D plots can be seen.

5. Getting started

In this section we'll go from the the most basic line plot to a fairly customized version in 2D, and similarly for 3D. The line demo scripts can be found in _demo/scripts/line_demos/.

The MPL Plotter workflow is simple by design: the walkthrough below is sufficient to acquaint you with all functionality, for line plots as well as all others.

The base output of all available 2D and 3D plot follow in Section 6. By then, you will be able to pick anyone up and do your thing.

5.1 2D Lines

As follows from the map above, the import to use the 2D line class is:

from mpl_plotter.two_d import line

And the following is the most basic MPL Plotter call, which will generate the image below (no input, and sin wave respectively).

line(show=True)	x = np.linspace(0, 2*np.pi, 100) y = np.sin(x) line(x=x, y=y, show=True)

Two important features are apparent:

1. MPL Plotter provides mock plots for every plotting class, so you can get straight into action and see what each does
2. MPL Plotter is somewhat "opinionated" and sets up quite a few parameters by default. This is based purely on my preference. You may not agree and you're more than welcome to play around with them!

---

Two more examples (result in the table below):

1. We can add some customization to make our line look a bit better:

    line(show=True, demo_pad_plot=True, spines_removed=None)
   

   Our line has now some margins to breathe while the ticks are placed at the maximum and minimums of our curve, and no spines are removed.

2. Lastly, an example using some of the parameters you can change:

    line(norm=True, line_width=4,
         aspect=1,
         show=True, demo_pad_plot=True,
         x_label="x", x_label_size=30, x_label_pad=-0.05,
         y_label="$\Psi$", y_label_size=30, y_label_rotation=0, y_label_pad=20,
         title="Custom Line", title_font="Pump Triline", title_size=40, title_color="orange",
         tick_color="darkgrey", workspace_color="darkred", tick_ndecimals=4,
         x_tick_number=12, y_tick_number=12,
         x_tick_rotation=35,
         color_bar=True, cb_tick_number=5, cb_pad=0.05,
         grid=True, grid_color="grey")
   

1. Somewhat customized	2. Customization example

---

5.2 3D Lines

Much of the same follows for 3D plots. In this case however customization is somewhat more limited. This is due to the fact that

1. 3D plots are less useful in general (in my experience, and thus I've spent less time on them)
2. Matplotlib support for 3D plots is more limited

Basic	Somewhat customized	Customization example

6. Base methods: examples, status

Below can be seen the base output of all methods, their input variables, and an indication of how stable each method is. In tests/test_minimal, base calls (no arguments besides show=True) for all methods are available. For real-world reference, tests/tests_2D and tests/tests_3D contain an example using various parameters for every single method.

For method-specific customization options (say, the line_width or point_size attributes for lines and scatter plots respectively), please check each method's docstring .

6.1 2D

All plots generated in tests/test_minimal.py.

Method	Status	Input	Base output
line	Stable	x, y
scatter	Stable	x, y
heatmap	Stable	x, y, z
quiver	Stable	x, y, u, v
streamline	Stable	x, y, u, v
fill	Stable	x, y, z

6.2 3D

Once more, all plots generated in tests/test_minimal.py. Wireframe is included: note it's not a method per se, but a setting of surface (hover over the image to see it).

Method	Status	Input	Base output
line	Stable	x, y, z
scatter	Stable	x, y, z
surface	Stable	x, y, z
Wireframe	Stable	x, y, z

6.3 Plot combination examples

7. Matplotlib compatibility

7.1 Retrieving axes, figures

The axis and figure on which each class draws are instance attributes. To retrieve them and continue modifications using standard Matplotlib:

from mpl_plotter.two_d import line

my_plot = line()
ax, fig = my_plot.ax, my_plot.fig

With the axis and figure, most Matplotlib functions out there can be used to further modify your plots.

7.2 Using Matplotlib's axis tiling

Matplotlib allows for subplot composition using subplot2grid. This can be used in combination with MPL Plotter:

Importantly:

• The auxiliary function figure (from mpl_plotter.setup import figure) sets up a figure in a chosen backend. This is convenient, as if the figure is created with plt.figure(), only the default non-interactive Matplotlib backend will be available, unless matplotlib.use(<backend>) is specified before importing pyplot.

    backend = "Qt5Agg"  # None -> regular non-interactive matplotlib output
    
    fig = figure(figsize=(10, 10), backend=backend)
    
    ax0 = plt.subplot2grid((2, 2), (0, 0), rowspan=1, aspect=1, fig=fig)
    ax1 = plt.subplot2grid((2, 2), (1, 0), rowspan=1, aspect=1, fig=fig)
    ax2 = plt.subplot2grid((2, 2), (0, 1), rowspan=1, aspect=1, fig=fig)
    ax3 = plt.subplot2grid((2, 2), (1, 1), rowspan=1, aspect=12, fig=fig)
    
    axes = [ax0, ax1, ax2, ax3]
    plots = [line, quiver, streamline, fill_area]
    
    for i in range(len(plots)):
        plots[i](fig=fig, ax=axes[i],
                 backend=backend
                 )
    
    plt.show()
  

8. Advanced: Presets and custom_canvas

The following are alternative ways to use MPL Plotter. Presets are currently implemented for the 2D and 3D line and scatter plot classes. More might be implemented in the future.

8.1 Custom presets

Presets enable you to create plots without barely writing any code. An example workflow follows.

1. Use a preset creation function (generate_preset_2d or generate_preset_3d) to create a preset

    from mpl_plotter.presets.custom import generate_preset_2d
    
    generate_preset_2d(preset_dest="presets", preset_name="MYPRESET", disable_warning=True, overwrite=True)
   

   A MYPRESET.py file will be created in a new (or not) presets/ directory within your project's root directory.

   • If no preset_dest is provided, MYPRESET.py will be saved in your root directory.
   • If no preset_name is provided, the preset will be saved as preset_2d.py.
   • By setting disable_warning=True, console output reminding you of the risk of rewriting your preset will be suppressed.
   • By setting overwrite=True, every time your run the preset creation function, it will overwrite the previously created preset with the same name (rather inconvenient, but who knows when it can come in handy).

   This file has a preset dictionary inside, with all editable parameters inside it, and commented out. Eg:

    preset = { 
        ## Basic 
        # "plot_label": None, 
        ## Backend 
        # "backend": "Qt5Agg", 
        ## Fonts 
        # "font": "serif",
        ...
    }
   

   By uncommenting certain lines, those parameters will be read and used to shape your plots.

2. Modify MYPRESET.py according to your needs.

3. Import mpl_plotter.presets.custom.two_d (or three_d if working with a 3D preset) and initiate it with MYPRESET

    from mpl_plotter.presets.custom import two_d
    
    my_preset_plot_family = two_d(preset_dir="presets", preset_name="MYPRESET")
    
    my_preset_line = my_plot_family.line
    
    # You can create further plotting classes spawning from my_preset_plot_family:
    # Eg        --->        my_preset_scatter = my_plot_family.scatter
   

4. Call a plotting function child of two_d, setting any extra parameters appropriately (plot title, etc.)

    my_preset_line(show=True, demo_pad_plot=True, color="blue", title="TITLE", title_size=200, aspect=1)
   

   The result of this example, its 3D version, and demos for all other available 2D and 3D plots can be seen in the table below.

   2D
   3D

5. Make as many plots as you need. Tiling is supported as well (see panes in Section 9)

8.2 Standard presets

Standard presets are available to remove overhead. They're tailored for my needs and desires, but perhaps you find them useful too.

Publication

It is a common mistake to make a figure for a paper with unreadable labels. This preset tries to solve that, generating plots optimized to be printed on a small format, in side-by-side plots or embedded in a column of text.

from mpl_plotter.presets.precision import two_d
from mpl_plotter.color.schemes import one           # Custom colorscheme

x = np.linspace(0, 4, 1000)
y = np.exp(x)
z = abs(np.sin(x)*np.exp(x))

two_d.line(x, z, aspect=0.05, color=one()[-2], show=True)

Precision

Made to plot functions large on the screen, with equal x and y scales to avoid skewing the variables, and many ticks to visually inspect a signal.

from mpl_plotter.presets.precision import two_d

two_d.line(x, z, aspect=0.05, color=one()[-2], show=True)

Publication	Precision

And below, all remaining plots (publication preset above, precision below):

8.3 custom_canvas

Lastly, MPL Plotter can be used to create a "custom canvas" on which to draw with Matplotlib. - custom_canvas creates a figure and 1. By retrieving the figure, more axes may be created. - If you wish custom_canvas to resize your axes, it must be given the x and y of (one) of your plots

NOTE: functionality might not be at 100% yet when using custom_canvas+Matplotlib as compared to plotting with MPL Plotter directly.

from mpl_plotter.setup import custom_canvas

x = np.linspace(0, 2*np.pi, 100)
y = np.sin(x)

c = custom_canvas(x=x, y=y, spines_removed=None, font_color="darkred")  # x and y provided: axes are resized
ax, fig = c.ax, c.fig

# Regular Matplotlib stuff

plt.plot(x, y)

plt.show()

9. Advanced: comparison and panes

Disclaimer: The following are utilities which combine presets and axis tiling to create n-pane plots. The API is very volatile, and flexibility must be improved. In any case, I find them practical from time to time, perhaps you too.

MPL Plotter includes a panes package for line plots, via the Lines class. The method "map" is as follows:

• mpl_plotter
  • two_d
    • comparison
      • comparison
    • panes
      • n_pane_single
      • n_pane_comparison

9.1 comparison

The comparison function facilitates including any number of curves in a single plot. The axis limits will be automatically set so no data lies outside.

Lines().comparison([x, x, x],
                   [u, v, w],
                   plot_labels=["sin", "cos", "tan"],
                   x_custom_tick_labels=[0, r"$\frac{\pi}{8}$", r"$\frac{\pi}{4}$"],
                   show=show,
                   )

A plotting function of choice can be specified for each of the arrays to be plotted. This is especially useful to easily combine lines with scatter plots, among other uses. Below you can see an example in which:

1. Three plotting functions are defined making use of the MPL Plotter line and scatter plotting classes.

2. The plotting functions are input in a list in the comparison call

    from mpl_plotter.two_d import comparison, line, scatter
    
    
    def f(x, y, **kwargs):
        line(x, y,
             line_width=2,
             **kwargs)
    def g(x, y, **kwargs):
        scatter(x, y,
                marker="D",
                point_size=10,
                **kwargs)
    def h(x, y, **kwargs):
        scatter(x, y,
                marker="s",
                point_size=5,
                **kwargs)
   
    comparison([x, x, x],
               [u, v, w],
               [f, g, h],
               plot_labels=["sin", "cos", "tan"],
               zorders=[1, 2, 3],
               colors=['C1', 'C2', 'C3'],
               alphas=[0.5, 0.5, 1],
               x_custom_tick_labels=[0, r"$\frac{\pi}{8}$", r"$\frac{\pi}{4}$"],
               show=show, backend=backend
               )
   

9.2 n_pane_single

This function takes in a number n of curves, and generates an n-pane panel plot with them.

 Lines(preset=preset).n_pane_single(x,                   # Horizontal vector
                                    [u, v, y],           # List of curves to be plotted
                                    ["u", "v", "y"],     # List of vertical axis labels
                                    ["a", "b", "c"]      # List of legend labels 
                                    )  

9.3 n_pane_comparison

In turn, this function takes in a number n of lists of m curves (where m=2 in the example below), to be plotted in the same pane for comparison.

Lines(preset=preset).n_pane_comparison(x,                               # Horizontal vector
                                       [[u, uu], [v, vv], [y, yy]],     # List of pairs of curves to be compared
                                       ["u", "v", "y"],                 # List of vertical axis labels
                                       ["a", "b"]                       # List of legend labels
                                       )

9.4 Bunch of panes

Why not.

And more of the same with n m-curve comparisons.