Adding new scales and projections to matplotlib

Matplotlib supports the addition of custom procedures that transform the data before it is displayed.

There is an important distinction between two kinds of transformations. Separable transformations, working on a single dimension, are called “scales”, and non-separable transformations, that handle data in two or more dimensions at a time, are called “projections”.

From the user’s perspective, the scale of a plot can be set with set_xscale() and set_yscale(). Projections can be chosen using the projection keyword argument to the plot() or subplot() functions, e.g.:

plot(x, y, projection="custom")

This document is intended for developers and advanced users who need to create new scales and projections for matplotlib. The necessary code for scales and projections can be included anywhere: directly within a plot script, in third-party code, or in the matplotlib source tree itself.

Creating a new scale

Adding a new scale consists of defining a subclass of matplotlib.scale.ScaleBase, that includes the following elements:

• A transformation from data coordinates into display coordinates.
• An inverse of that transformation. This is used, for example, to convert mouse positions from screen space back into data space.
• A function to limit the range of the axis to acceptable values (limit_range_for_scale()). A log scale, for instance, would prevent the range from including values less than or equal to zero.
• Locators (major and minor) that determine where to place ticks in the plot, and optionally, how to adjust the limits of the plot to some “good” values. Unlike limit_range_for_scale(), which is always enforced, the range setting here is only used when automatically setting the range of the plot.
• Formatters (major and minor) that specify how the tick labels should be drawn.

Once the class is defined, it must be registered with matplotlib so that the user can select it.

A full-fledged and heavily annotated example is in examples/api/custom_scale_example.py. There are also some classes in matplotlib.scale that may be used as starting points.

Creating a new projection

Adding a new projection consists of defining a projection axes which subclasses matplotlib.axes.Axes and includes the following elements:

• A transformation from data coordinates into display coordinates.
• An inverse of that transformation. This is used, for example, to convert mouse positions from screen space back into data space.
• Transformations for the gridlines, ticks and ticklabels. Custom projections will often need to place these elements in special locations, and matplotlib has a facility to help with doing so.
• Setting up default values (overriding cla()), since the defaults for a rectilinear axes may not be appropriate.
• Defining the shape of the axes, for example, an elliptical axes, that will be used to draw the background of the plot and for clipping any data elements.
• Defining custom locators and formatters for the projection. For example, in a geographic projection, it may be more convenient to display the grid in degrees, even if the data is in radians.
• Set up interactive panning and zooming. This is left as an “advanced” feature left to the reader, but there is an example of this for polar plots in matplotlib.projections.polar.
• Any additional methods for additional convenience or features.

Once the projection axes is defined, it can be used in one of two ways:

• By defining the class attribute name, the projection axes can be registered with matplotlib.projections.register_projection() and subsequently simply invoked by name:

  plt.axes(projection='my_proj_name')
  

• For more complex, parameterisable projections, a generic “projection” object may be defined which includes the method _as_mpl_axes. _as_mpl_axes should take no arguments and return the projection’s axes subclass and a dictionary of additional arguments to pass to the subclass’ __init__ method. Subsequently a parameterised projection can be initialised with:

  plt.axes(projection=MyProjection(param1=param1_value))
  

  where MyProjection is an object which implements a _as_mpl_axes method.

A full-fledged and heavily annotated example is in examples/api/custom_projection_example.py. The polar plot functionality in matplotlib.projections.polar may also be of interest.

API documentation

matplotlib.scale

class matplotlib.scale.LinearScale(axis, **kwargs)

Bases: matplotlib.scale.ScaleBase

The default linear scale.

get_transform()

The transform for linear scaling is just the IdentityTransform.

set_default_locators_and_formatters(axis)

Set the locators and formatters to reasonable defaults for linear scaling.

class matplotlib.scale.LogScale(axis, **kwargs)

Bases: matplotlib.scale.ScaleBase

A standard logarithmic scale. Care is taken so non-positive values are not plotted.

For computational efficiency (to push as much as possible to Numpy C code in the common cases), this scale provides different transforms depending on the base of the logarithm:

• base 10 (Log10Transform)
• base 2 (Log2Transform)
• base e (NaturalLogTransform)
• arbitrary base (LogTransform)

basex/basey:

The base of the logarithm

nonposx/nonposy: [‘mask’ | ‘clip’ ]

non-positive values in x or y can be masked as invalid, or clipped to a very small positive number

subsx/subsy:

Where to place the subticks between each major tick. Should be a sequence of integers. For example, in a log10 scale: [2, 3, 4, 5, 6, 7, 8, 9]

will place 8 logarithmically spaced minor ticks between each major tick.

get_transform()

Return a Transform instance appropriate for the given logarithm base.

limit_range_for_scale(vmin, vmax, minpos)

Limit the domain to positive values.

set_default_locators_and_formatters(axis)

Set the locators and formatters to specialized versions for log scaling.

class matplotlib.scale.LogitScale(axis, nonpos='mask')

Bases: matplotlib.scale.ScaleBase

Logit scale for data between zero and one, both excluded.

This scale is similar to a log scale close to zero and to one, and almost linear around 0.5. It maps the interval ]0, 1[ onto ]-infty, +infty[.

nonpos: [‘mask’ | ‘clip’ ]

values beyond ]0, 1[ can be masked as invalid, or clipped to a number very close to 0 or 1

get_transform()

Return a LogitTransform instance.

limit_range_for_scale(vmin, vmax, minpos)

Limit the domain to values between 0 and 1 (excluded).

class matplotlib.scale.ScaleBase

Bases: object

The base class for all scales.

Scales are separable transformations, working on a single dimension.

Any subclasses will want to override:

• name
• get_transform()
• set_default_locators_and_formatters()

And optionally:

• limit_range_for_scale()

get_transform()

Return the Transform object associated with this scale.

limit_range_for_scale(vmin, vmax, minpos)

Returns the range vmin, vmax, possibly limited to the domain supported by this scale.

minpos should be the minimum positive value in the data.

This is used by log scales to determine a minimum value.

set_default_locators_and_formatters(axis)

Set the Locator and Formatter objects on the given axis to match this scale.

class matplotlib.scale.SymmetricalLogScale(axis, **kwargs)

Bases: matplotlib.scale.ScaleBase

The symmetrical logarithmic scale is logarithmic in both the positive and negative directions from the origin.

Since the values close to zero tend toward infinity, there is a need to have a range around zero that is linear. The parameter linthresh allows the user to specify the size of this range (-linthresh, linthresh).

basex/basey:

The base of the logarithm

linthreshx/linthreshy:

The range (-x, x) within which the plot is linear (to avoid having the plot go to infinity around zero).

subsx/subsy:

Where to place the subticks between each major tick. Should be a sequence of integers. For example, in a log10 scale: [2, 3, 4, 5, 6, 7, 8, 9]

will place 8 logarithmically spaced minor ticks between each major tick.

linscalex/linscaley:

This allows the linear range (-linthresh to linthresh) to be stretched relative to the logarithmic range. Its value is the number of decades to use for each half of the linear range. For example, when linscale == 1.0 (the default), the space used for the positive and negative halves of the linear range will be equal to one decade in the logarithmic range.

get_transform()

Return a SymmetricalLogTransform instance.

set_default_locators_and_formatters(axis)

Set the locators and formatters to specialized versions for symmetrical log scaling.

matplotlib.scale.get_scale_docs()

Helper function for generating docstrings related to scales.

matplotlib.scale.register_scale(scale_class)

Register a new kind of scale.

scale_class must be a subclass of ScaleBase.

matplotlib.scale.scale_factory(scale, axis, **kwargs)

Return a scale class by name.

ACCEPTS: [ linear | log | logit | symlog ]

matplotlib.projections

class matplotlib.projections.ProjectionRegistry

Bases: object

Manages the set of projections available to the system.

get_projection_class(name)

Get a projection class from its name.

get_projection_names()

Get a list of the names of all projections currently registered.

register(*projections)

Register a new set of projection(s).

matplotlib.projections.get_projection_class(projection=None)

Get a projection class from its name.

If projection is None, a standard rectilinear projection is returned.

matplotlib.projections.get_projection_names()

Get a list of acceptable projection names.

matplotlib.projections.process_projection_requirements(figure, *args, **kwargs)

Handle the args/kwargs to for add_axes/add_subplot/gca, returning:

(axes_proj_class, proj_class_kwargs, proj_stack_key)

Which can be used for new axes initialization/identification.

Note

kwargs is modified in place.

matplotlib.projections.polar

class matplotlib.projections.polar.InvertedPolarTransform(axis=None, use_rmin=True)

Bases: matplotlib.transforms.Transform

The inverse of the polar transform, mapping Cartesian coordinate space x and y back to theta and r.

inverted()

Return the corresponding inverse transformation.

The return value of this method should be treated as temporary. An update to self does not cause a corresponding update to its inverted copy.

x === self.inverted().transform(self.transform(x))

transform_non_affine(xy)

Performs only the non-affine part of the transformation.

transform(values) is always equivalent to transform_affine(transform_non_affine(values)).

In non-affine transformations, this is generally equivalent to transform(values). In affine transformations, this is always a no-op.

Accepts a numpy array of shape (N x input_dims) and returns a numpy array of shape (N x output_dims).

Alternatively, accepts a numpy array of length input_dims and returns a numpy array of length output_dims.

class matplotlib.projections.polar.PolarAffine(scale_transform, limits)

Bases: matplotlib.transforms.Affine2DBase

The affine part of the polar projection. Scales the output so that maximum radius rests on the edge of the axes circle.

limits is the view limit of the data. The only part of its bounds that is used is ymax (for the radius maximum). The theta range is always fixed to (0, 2pi).

get_matrix()

Get the Affine transformation array for the affine part of this transform.

class matplotlib.projections.polar.PolarAxes(*args, **kwargs)

Bases: matplotlib.axes._axes.Axes

A polar graph projection, where the input dimensions are theta, r.

Theta starts pointing east and goes anti-clockwise.

class InvertedPolarTransform(axis=None, use_rmin=True)

Bases: matplotlib.transforms.Transform

The inverse of the polar transform, mapping Cartesian coordinate space x and y back to theta and r.

inverted()

Return the corresponding inverse transformation.

The return value of this method should be treated as temporary. An update to self does not cause a corresponding update to its inverted copy.

x === self.inverted().transform(self.transform(x))

transform_non_affine(xy)

Performs only the non-affine part of the transformation.

transform(values) is always equivalent to transform_affine(transform_non_affine(values)).

In non-affine transformations, this is generally equivalent to transform(values). In affine transformations, this is always a no-op.

Accepts a numpy array of shape (N x input_dims) and returns a numpy array of shape (N x output_dims).

Alternatively, accepts a numpy array of length input_dims and returns a numpy array of length output_dims.

class PolarAxes.PolarAffine(scale_transform, limits)

Bases: matplotlib.transforms.Affine2DBase

The affine part of the polar projection. Scales the output so that maximum radius rests on the edge of the axes circle.

limits is the view limit of the data. The only part of its bounds that is used is ymax (for the radius maximum). The theta range is always fixed to (0, 2pi).

get_matrix()

Get the Affine transformation array for the affine part of this transform.

class PolarAxes.PolarTransform(axis=None, use_rmin=True)

Bases: matplotlib.transforms.Transform

The base polar transform. This handles projection theta and r into Cartesian coordinate space x and y, but does not perform the ultimate affine transformation into the correct position.

inverted()

Return the corresponding inverse transformation.

The return value of this method should be treated as temporary. An update to self does not cause a corresponding update to its inverted copy.

x === self.inverted().transform(self.transform(x))

transform_non_affine(tr)

Performs only the non-affine part of the transformation.

transform(values) is always equivalent to transform_affine(transform_non_affine(values)).

In non-affine transformations, this is generally equivalent to transform(values). In affine transformations, this is always a no-op.

Accepts a numpy array of shape (N x input_dims) and returns a numpy array of shape (N x output_dims).

Alternatively, accepts a numpy array of length input_dims and returns a numpy array of length output_dims.

transform_path_non_affine(path)

Returns a path, transformed only by the non-affine part of this transform.

path: a Path instance.

transform_path(path) is equivalent to transform_path_affine(transform_path_non_affine(values)).

class PolarAxes.RadialLocator(base)

Bases: matplotlib.ticker.Locator

Used to locate radius ticks.

Ensures that all ticks are strictly positive. For all other tasks, it delegates to the base Locator (which may be different depending on the scale of the r-axis.

class PolarAxes.ThetaFormatter

Bases: matplotlib.ticker.Formatter

Used to format the theta tick labels. Converts the native unit of radians into degrees and adds a degree symbol.

PolarAxes.can_pan()

Return True if this axes supports the pan/zoom button functionality.

For polar axes, this is slightly misleading. Both panning and zooming are performed by the same button. Panning is performed in azimuth while zooming is done along the radial.

PolarAxes.can_zoom()

Return True if this axes supports the zoom box button functionality.

Polar axes do not support zoom boxes.

PolarAxes.format_coord(theta, r)

Return a format string formatting the coordinate using Unicode characters.

PolarAxes.get_data_ratio()

Return the aspect ratio of the data itself. For a polar plot, this should always be 1.0

PolarAxes.get_rlabel_position()

Returns:

float

The theta position of the radius labels in degrees.

PolarAxes.get_theta_direction()

Get the direction in which theta increases.

-1:

Theta increases in the clockwise direction

1:

Theta increases in the counterclockwise direction

PolarAxes.get_theta_offset()

Get the offset for the location of 0 in radians.

PolarAxes.set_rgrids(radii, labels=None, angle=None, fmt=None, **kwargs)

Set the radial locations and labels of the r grids.

The labels will appear at radial distances radii at the given angle in degrees.

labels, if not None, is a len(radii) list of strings of the labels to use at each radius.

If labels is None, the built-in formatter will be used.

Return value is a list of tuples (line, label), where line is Line2D instances and the label is Text instances.

kwargs are optional text properties for the labels:

Property	Description
agg_filter	unknown
alpha	float (0.0 transparent through 1.0 opaque)
animated	[True | False]
axes	an Axes instance
backgroundcolor	any matplotlib color
bbox	FancyBboxPatch prop dict
bottom_margin	unknown
clip_box	a matplotlib.transforms.Bbox instance
clip_on	[True | False]
clip_path	[ (Path, Transform) | Patch | None ]
color	any matplotlib color
contains	a callable function
family or fontfamily or fontname or name	[FONTNAME | ‘serif’ | ‘sans-serif’ | ‘cursive’ | ‘fantasy’ | ‘monospace’ ]
figure	a matplotlib.figure.Figure instance
fontproperties or font_properties	a matplotlib.font_manager.FontProperties instance
gid	an id string
horizontalalignment or ha	[ ‘center’ | ‘right’ | ‘left’ ]
label	string or anything printable with ‘%s’ conversion.
left_margin	unknown
linespacing	float (multiple of font size)
margins	unknown
multialignment	[‘left’ | ‘right’ | ‘center’ ]
path_effects	unknown
picker	[None|float|boolean|callable]
position	(x,y)
rasterized	[True | False | None]
right_margin	unknown
rotation	[ angle in degrees | ‘vertical’ | ‘horizontal’ ]
rotation_mode	unknown
size or fontsize	[size in points | ‘xx-small’ | ‘x-small’ | ‘small’ | ‘medium’ | ‘large’ | ‘x-large’ | ‘xx-large’ ]
sketch_params	unknown
snap	unknown
stretch or fontstretch	[a numeric value in range 0-1000 | ‘ultra-condensed’ | ‘extra-condensed’ | ‘condensed’ | ‘semi-condensed’ | ‘normal’ | ‘semi-expanded’ | ‘expanded’ | ‘extra-expanded’ | ‘ultra-expanded’ ]
style or fontstyle	[ ‘normal’ | ‘italic’ | ‘oblique’]
text	string or anything printable with ‘%s’ conversion.
top_margin	unknown
transform	Transform instance
url	a url string
usetex	unknown
variant or fontvariant	[ ‘normal’ | ‘small-caps’ ]
verticalalignment or ma or va	[ ‘center’ | ‘top’ | ‘bottom’ | ‘baseline’ ]
visible	[True | False]
weight or fontweight	[a numeric value in range 0-1000 | ‘ultralight’ | ‘light’ | ‘normal’ | ‘regular’ | ‘book’ | ‘medium’ | ‘roman’ | ‘semibold’ | ‘demibold’ | ‘demi’ | ‘bold’ | ‘heavy’ | ‘extra bold’ | ‘black’ ]
wrap	unknown
x	float
y	float
zorder	any number

ACCEPTS: sequence of floats

PolarAxes.set_rlabel_position(value)

Updates the theta position of the radius labels.

Parameters:

value : number

The angular position of the radius labels in degrees.

PolarAxes.set_theta_direction(direction)

Set the direction in which theta increases.

clockwise, -1:

Theta increases in the clockwise direction

counterclockwise, anticlockwise, 1:

Theta increases in the counterclockwise direction

PolarAxes.set_theta_offset(offset)

Set the offset for the location of 0 in radians.

PolarAxes.set_theta_zero_location(loc)

Sets the location of theta’s zero. (Calls set_theta_offset with the correct value in radians under the hood.)

May be one of “N”, “NW”, “W”, “SW”, “S”, “SE”, “E”, or “NE”.

PolarAxes.set_thetagrids(angles, labels=None, frac=None, fmt=None, **kwargs)

Set the angles at which to place the theta grids (these gridlines are equal along the theta dimension). angles is in degrees.

labels, if not None, is a len(angles) list of strings of the labels to use at each angle.

If labels is None, the labels will be fmt % angle

frac is the fraction of the polar axes radius at which to place the label (1 is the edge). e.g., 1.05 is outside the axes and 0.95 is inside the axes.

Return value is a list of tuples (line, label), where line is Line2D instances and the label is Text instances.

kwargs are optional text properties for the labels:

Property	Description
agg_filter	unknown
alpha	float (0.0 transparent through 1.0 opaque)
animated	[True | False]
axes	an Axes instance
backgroundcolor	any matplotlib color
bbox	FancyBboxPatch prop dict
bottom_margin	unknown
clip_box	a matplotlib.transforms.Bbox instance
clip_on	[True | False]
clip_path	[ (Path, Transform) | Patch | None ]
color	any matplotlib color
contains	a callable function
family or fontfamily or fontname or name	[FONTNAME | ‘serif’ | ‘sans-serif’ | ‘cursive’ | ‘fantasy’ | ‘monospace’ ]
figure	a matplotlib.figure.Figure instance
fontproperties or font_properties	a matplotlib.font_manager.FontProperties instance
gid	an id string
horizontalalignment or ha	[ ‘center’ | ‘right’ | ‘left’ ]
label	string or anything printable with ‘%s’ conversion.
left_margin	unknown
linespacing	float (multiple of font size)
margins	unknown
multialignment	[‘left’ | ‘right’ | ‘center’ ]
path_effects	unknown
picker	[None|float|boolean|callable]
position	(x,y)
rasterized	[True | False | None]
right_margin	unknown
rotation	[ angle in degrees | ‘vertical’ | ‘horizontal’ ]
rotation_mode	unknown
size or fontsize	[size in points | ‘xx-small’ | ‘x-small’ | ‘small’ | ‘medium’ | ‘large’ | ‘x-large’ | ‘xx-large’ ]
sketch_params	unknown
snap	unknown
stretch or fontstretch	[a numeric value in range 0-1000 | ‘ultra-condensed’ | ‘extra-condensed’ | ‘condensed’ | ‘semi-condensed’ | ‘normal’ | ‘semi-expanded’ | ‘expanded’ | ‘extra-expanded’ | ‘ultra-expanded’ ]
style or fontstyle	[ ‘normal’ | ‘italic’ | ‘oblique’]
text	string or anything printable with ‘%s’ conversion.
top_margin	unknown
transform	Transform instance
url	a url string
usetex	unknown
variant or fontvariant	[ ‘normal’ | ‘small-caps’ ]
verticalalignment or ma or va	[ ‘center’ | ‘top’ | ‘bottom’ | ‘baseline’ ]
visible	[True | False]
weight or fontweight	[a numeric value in range 0-1000 | ‘ultralight’ | ‘light’ | ‘normal’ | ‘regular’ | ‘book’ | ‘medium’ | ‘roman’ | ‘semibold’ | ‘demibold’ | ‘demi’ | ‘bold’ | ‘heavy’ | ‘extra bold’ | ‘black’ ]
wrap	unknown
x	float
y	float
zorder	any number

ACCEPTS: sequence of floats

class matplotlib.projections.polar.PolarTransform(axis=None, use_rmin=True)

Bases: matplotlib.transforms.Transform

The base polar transform. This handles projection theta and r into Cartesian coordinate space x and y, but does not perform the ultimate affine transformation into the correct position.

inverted()

Return the corresponding inverse transformation.

The return value of this method should be treated as temporary. An update to self does not cause a corresponding update to its inverted copy.

x === self.inverted().transform(self.transform(x))

transform_non_affine(tr)

Performs only the non-affine part of the transformation.

transform(values) is always equivalent to transform_affine(transform_non_affine(values)).

In non-affine transformations, this is generally equivalent to transform(values). In affine transformations, this is always a no-op.

Accepts a numpy array of shape (N x input_dims) and returns a numpy array of shape (N x output_dims).

Alternatively, accepts a numpy array of length input_dims and returns a numpy array of length output_dims.

transform_path_non_affine(path)

Returns a path, transformed only by the non-affine part of this transform.

path: a Path instance.

transform_path(path) is equivalent to transform_path_affine(transform_path_non_affine(values)).

class matplotlib.projections.polar.RadialLocator(base)

Bases: matplotlib.ticker.Locator

Used to locate radius ticks.

Ensures that all ticks are strictly positive. For all other tasks, it delegates to the base Locator (which may be different depending on the scale of the r-axis.

class matplotlib.projections.polar.ThetaFormatter

Bases: matplotlib.ticker.Formatter

Used to format the theta tick labels. Converts the native unit of radians into degrees and adds a degree symbol.