Source code for proplot.axes.cartesian

#!/usr/bin/env python3
"""
The standard x-y axes used for most ProPlot figures.
"""
import matplotlib.dates as mdates
import matplotlib.ticker as mticker
import numpy as np

from .. import constructor
from .. import scale as pscale
from .. import ticker as pticker
from ..config import rc
from ..internals import ic  # noqa: F401
from ..internals import _not_none, docstring, rcsetup, warnings
from ..utils import units
from . import base

__all__ = ['CartesianAxes']


_alt_doc = """
Return an axes in the same location as this one but whose {x} axis is on
the {x2}. This is an alias and more intuitive name for
`~CartesianAxes.twin{y}`, which generates two *{x}* axes with
a shared ("twin") *{y}* axes.

Also enforces the following settings:

* Places the old *{x}* axis on the {x1} and the new *{x}* axis
  on the {x2}.
* Makes the old {x2} spine invisible and the new {x1}, {y1},
  and {y2} spines invisible.
* Adjusts the *{x}* axis tick, tick label, and axis label positions
  according to the visible spine positions.
* Locks the old and new *{y}* axis limits and scales, and makes the new
  {y} axis labels invisible.

Parameters
----------
{xargs} : optional
    Passed to `Axes.format`.
{args} : optional
    Prepended with ``'{x}'`` and passed to `Axes.format`.
"""

_alt_kwargs = (  # TODO: More systematic approach?
    'lim', 'reverse', 'scale', 'label',
    'tickdir', 'grid', 'gridminor',
    'tickminor', 'ticklabeldir', 'tickrange', 'wraprange',
    'rotation', 'formatter', 'ticklabels',
    'ticks', 'locator', 'minorticks', 'minorlocator',
    'bounds', 'margin', 'color',
    'ticklen', 'linewidth', 'gridcolor',
    'label_kw', 'scale_kw', 'locator_kw', 'formatter_kw', 'minorlocator_kw',
)

docstring.snippets['axes.altx'] = _alt_doc.format(
    x='x',
    x1='bottom',
    x2='top',
    y='y',
    y1='left',
    y2='right',
    args=', '.join(_alt_kwargs),
    xargs=', '.join('x' + key for key in _alt_kwargs),
)

docstring.snippets['axes.alty'] = _alt_doc.format(
    x='y',
    x1='left',
    x2='right',
    y='x',
    y1='bottom',
    y2='top',
    args=', '.join(_alt_kwargs),
    xargs=', '.join('y' + key for key in _alt_kwargs),
)

_dual_doc = """
Return a secondary *{x}* axis for denoting equivalent *{x}*
coordinates in *alternate units*.

Parameters
----------
funcscale : function, (function, function), or scale-spec
    Used to transform units from the parent axis to the secondary axis.
    This can be a `~proplot.scale.FuncScale` itself or a function,
    (function, function) tuple, or an axis scale specification interpreted
    by the `~proplot.constructor.Scale` constructor function, any of which will
    be used to build a `~proplot.scale.FuncScale` and applied to the dual axis
    (see `~proplot.scale.FuncScale` for details).
{args} : optional
    Prepended with ``'{x}'`` and passed to `Axes.format`.
"""

docstring.snippets['axes.dualx'] = _dual_doc.format(
    x='x',
    args=', '.join(_alt_kwargs),
    xargs=', '.join('x' + key for key in _alt_kwargs),
)
docstring.snippets['axes.dualy'] = _dual_doc.format(
    x='y',
    args=', '.join(_alt_kwargs),
    xargs=', '.join('y' + key for key in _alt_kwargs),
)

_twin_doc = """
Mimics the builtin `~matplotlib.axes.Axes.twin{y}` method.

Also enforces the following settings:

* Places the old *{x}* axis on the {x1} and the new *{x}* axis
  on the {x2}.
* Makes the old {x2} spine invisible and the new {x1}, {y1},
  and {y2} spines invisible.
* Adjusts the *{x}* axis tick, tick label, and axis label positions
  according to the visible spine positions.
* Locks the old and new *{y}* axis limits and scales, and makes the new
  {y} axis labels invisible.

Parameters
----------
{xargs} : optional
    Passed to `Axes.format`.
{args} : optional
    Prepended with ``'{x}'`` and passed to `Axes.format`.
"""

docstring.snippets['axes.twinx'] = _twin_doc.format(
    x='y', x1='left', x2='right',
    y='x', y1='bottom', y2='top',
    args=', '.join(_alt_kwargs),
    xargs=', '.join('y' + key for key in _alt_kwargs),
)
docstring.snippets['axes.twiny'] = _twin_doc.format(
    x='x', x1='bottom', x2='top',
    y='y', y1='left', y2='right',
    args=', '.join(_alt_kwargs),
    xargs=', '.join('x' + key for key in _alt_kwargs),
)


def _parse_alt(x, kwargs):
    """
    Interpret keyword args passed to all "twin axis" methods so they
    can be passed to Axes.format.
    """
    kw_bad, kw_out = {}, {}
    for key, value in kwargs.items():
        if key in _alt_kwargs:
            kw_out[x + key] = value
        elif key[0] == x and key[1:] in _alt_kwargs:
            # NOTE: We permit both e.g. 'locator' and 'xlocator' because
            # while is more elegant and consistent with e.g. colorbar() syntax
            # but latter is more consistent and easier to use when refactoring.
            kw_out[key] = value
        elif key in rcsetup._rc_nodots:
            kw_out[key] = value
        else:
            kw_bad[key] = value
    if kw_bad:
        raise TypeError(f'Unexpected keyword argument(s): {kw_bad!r}')
    return kw_out


def _parse_rcloc(x, string):  # figures out string location
    """
    Convert the *boolean* "left", "right", "top", and "bottom" rc settings
    to a location string. Returns ``None`` if settings are unchanged.
    """
    if x == 'x':
        top = rc.get(f'{string}.top', context=True)
        bottom = rc.get(f'{string}.bottom', context=True)
        if top is None and bottom is None:
            return None
        elif top and bottom:
            return 'both'
        elif top:
            return 'top'
        elif bottom:
            return 'bottom'
        else:
            return 'neither'
    else:
        left = rc.get(f'{string}.left', context=True)
        right = rc.get(f'{string}.right', context=True)
        if left is None and right is None:
            return None
        elif left and right:
            return 'both'
        elif left:
            return 'left'
        elif right:
            return 'right'
        else:
            return 'neither'


[docs]class CartesianAxes(base.Axes): """ Axes subclass for plotting in ordinary Cartesian coordinates. Adds the `~CartesianAxes.format` method and overrides several existing methods. """ #: The registered projection name. name = 'proplot_cartesian' def __init__(self, *args, **kwargs): """ See also -------- proplot.ui.subplots """ # Impose default formatter super().__init__(*args, **kwargs) formatter = pticker.AutoFormatter() self.xaxis.set_major_formatter(formatter) self.yaxis.set_major_formatter(formatter) self.xaxis.isDefault_majfmt = True self.yaxis.isDefault_majfmt = True self._altx_parent = None self._alty_parent = None self._datex_rotated = False # whether manual rotation has been applied self._dualx_funcscale = None # for scaling units on dual axes self._dualx_prevstate = None # prevent excess _dualy_scale calls self._dualy_funcscale = None self._dualy_prevstate = None def _apply_axis_sharing(self): """ Enforce the "shared" axis labels and axis tick labels. If this is not called at drawtime, "shared" labels can be inadvertantly turned off. """ # X axis # NOTE: The "panel sharing group" refers to axes and panels *above* the # bottommost or to the *right* of the leftmost panel. But the edge panel # sharing level is the *figure* sharing level. axis = self.xaxis if self._sharex is not None: level = 3 if self._panel_sharex_group else self.figure._sharex if level > 0: axis.label.set_visible(False) if level > 2: # WARNING: Cannot set NullFormatter because shared axes share the # same axis.Ticker classes. Instead use the approach copied from # matplotlib subplots(). axis.set_tick_params(which='both', labelbottom=False, labeltop=False) # Y axis axis = self.yaxis if self._sharey is not None: level = 3 if self._panel_sharey_group else self.figure._sharey if level > 0: axis.label.set_visible(False) if level > 2: axis.set_tick_params(which='both', labelleft=False, labelright=False) axis.set_minor_formatter(mticker.NullFormatter()) def _datex_rotate(self): """ Apply default rotation to datetime axis coordinates. """ # NOTE: Rotation is done *before* horizontal/vertical alignment, # cannot change alignment with set_tick_params. Must apply to text # objects. fig.autofmt_date calls subplots_adjust, so cannot use it. if ( not isinstance(self.xaxis.converter, mdates.DateConverter) or self._datex_rotated ): return rotation = rc['formatter.timerotation'] kw = {'rotation': rotation} if rotation not in (0, 90, -90): kw['ha'] = ('right' if rotation > 0 else 'left') for label in self.xaxis.get_ticklabels(): label.update(kw) self._datex_rotated = True # do not need to apply more than once def _dualx_scale(self): """ Lock the child "dual" *x* axis limits to the parent. """ # NOTE: We bypass autoscale_view because we set limits manually, and bypass # child.stale = True because that is done in call to set_xlim() below. # NOTE: We set the scale using private API to bypass application of # set_default_locators_and_formatters: only_if_default=True is critical # to prevent overriding user settings! # NOTE: Dual axis only needs to be constrained if the parent axis scale # and limits have changed, and limits are always applied before we reach # the child.draw() because always called after parent.draw() funcscale, parent, child = self._dualx_funcscale, self._altx_parent, self if funcscale is None or parent is None: return olim = parent.get_xlim() scale = parent.xaxis._scale if (scale, *olim) == child._dualx_prevstate: return funcscale = pscale.FuncScale(funcscale, invert=True, parent_scale=scale) child.xaxis._scale = funcscale child._update_transScale() funcscale.set_default_locators_and_formatters(child.xaxis, only_if_default=True) nlim = list(map(funcscale.functions[1], np.array(olim))) if np.sign(np.diff(olim)) != np.sign(np.diff(nlim)): nlim = nlim[::-1] # if function flips limits, so will set_xlim! child.set_xlim(nlim, emit=False) child._dualx_prevstate = (scale, *olim) def _dualy_scale(self): """ Lock the child "dual" *y* axis limits to the parent. """ # See _dualx_scale() comments funcscale, parent, child = self._dualy_funcscale, self._alty_parent, self if funcscale is None or parent is None: return olim = parent.get_ylim() scale = parent.yaxis._scale if (scale, *olim) == child._dualy_prevstate: return funcscale = pscale.FuncScale(funcscale, invert=True, parent_scale=scale) child.yaxis._scale = funcscale child._update_transScale() funcscale.set_default_locators_and_formatters(child.yaxis, only_if_default=True) nlim = list(map(funcscale.functions[1], np.array(olim))) if np.sign(np.diff(olim)) != np.sign(np.diff(nlim)): nlim = nlim[::-1] child.set_ylim(nlim, emit=False) child._dualy_prevstate = (scale, *olim) def _sharex_setup(self, sharex): """ Configure shared axes accounting for panels. The input is the 'parent' axes, from which this one will draw its properties. """ # Share *panels* across different subplots super()._sharex_setup(sharex) # Get sharing level level = ( 3 if self._panel_sharex_group and self._is_panel_group_member(sharex) else self.figure._sharex ) if level not in range(4): raise ValueError( 'Invalid sharing level sharex={value!r}. ' 'Axis sharing level can be 0 (share nothing), ' '1 (hide axis labels), ' '2 (share limits and hide axis labels), or ' '3 (share limits and hide axis and tick labels).' ) if sharex in (None, self) or not isinstance(sharex, CartesianAxes): return # Share future changes to axis labels # Proplot internally uses _sharex and _sharey for label sharing. Matplotlib # only uses these in __init__() and cla() to share tickers -- all other builtin # matplotlib axis sharing features derive from _shared_x_axes() group. if level > 0: self._sharex = sharex if not sharex.xaxis.label.get_text(): sharex.xaxis.label.set_text(self.xaxis.label.get_text()) # Share future axis tickers, limits, and scales # NOTE: Only difference between levels 2 and 3 is level 3 hides # tick labels. But this is done after the fact -- tickers are still shared. if level > 1: # Initial limits and scales should be shared both ways for (ax1, ax2) in ((self, sharex), (sharex, self)): if ax1.get_xscale() == 'linear' and ax2.get_xscale() != 'linear': ax1.set_xscale(ax2.get_xscale()) if ax1.get_autoscalex_on() and not ax2.get_autoscalex_on(): ax1.set_xlim(ax2.get_xlim()) # Locators and formatters only need to be shared from children # to parent, because this is done automatically when we assign # parent sharex tickers to child. self._shared_x_axes.join(self, sharex) # share limit/scale changes if sharex.xaxis.isDefault_majloc and not self.xaxis.isDefault_majloc: sharex.xaxis.set_major_locator(self.xaxis.get_major_locator()) if sharex.xaxis.isDefault_minloc and not self.xaxis.isDefault_minloc: sharex.xaxis.set_minor_locator(self.xaxis.get_minor_locator()) if sharex.xaxis.isDefault_majfmt and not self.xaxis.isDefault_majfmt: sharex.xaxis.set_major_formatter(self.xaxis.get_major_formatter()) if sharex.xaxis.isDefault_minfmt and not self.xaxis.isDefault_minfmt: sharex.xaxis.set_minor_formatter(self.xaxis.get_minor_formatter()) self.xaxis.major = sharex.xaxis.major self.xaxis.minor = sharex.xaxis.minor def _sharey_setup(self, sharey): """ Configure shared axes accounting for panels. The input is the 'parent' axes, from which this one will draw its properties. """ # Share *panels* across different subplots super()._sharey_setup(sharey) # Get sharing level level = ( 3 if self._panel_sharey_group and self._is_panel_group_member(sharey) else self.figure._sharey ) if level not in range(4): raise ValueError( 'Invalid sharing level sharey={value!r}. ' 'Axis sharing level can be 0 (share nothing), ' '1 (hide axis labels), ' '2 (share limits and hide axis labels), or ' '3 (share limits and hide axis and tick labels).' ) if sharey in (None, self) or not isinstance(sharey, CartesianAxes): return # Share future changes to axis labels if level > 0: self._sharey = sharey if not sharey.yaxis.label.get_text(): sharey.yaxis.label.set_text(self.yaxis.label.get_text()) # Share future axis tickers, limits, and scales if level > 1: # Initial limits and scales should be shared both ways for (ax1, ax2) in ((self, sharey), (sharey, self)): if ax1.get_yscale() == 'linear' and ax2.get_yscale() != 'linear': ax1.set_yscale(ax2.get_yscale()) if ax1.get_autoscaley_on() and not ax2.get_autoscaley_on(): ax1.set_ylim(ax2.get_ylim()) # Locators and formatters only need to be shared from children # to parent, because this is done automatically when we assign # parent sharey tickers to child. self._shared_y_axes.join(self, sharey) # share limit/scale changes if sharey.yaxis.isDefault_majloc and not self.yaxis.isDefault_majloc: sharey.yaxis.set_major_locator(self.yaxis.get_major_locator()) if sharey.yaxis.isDefault_minloc and not self.yaxis.isDefault_minloc: sharey.yaxis.set_minor_locator(self.yaxis.get_minor_locator()) if sharey.yaxis.isDefault_majfmt and not self.yaxis.isDefault_majfmt: sharey.yaxis.set_major_formatter(self.yaxis.get_major_formatter()) if sharey.yaxis.isDefault_minfmt and not self.yaxis.isDefault_minfmt: sharey.yaxis.set_minor_formatter(self.yaxis.get_minor_formatter()) self.yaxis.major = sharey.yaxis.major self.yaxis.minor = sharey.yaxis.minor def _update_bounds(self, x, fixticks=False): """ Ensure there are no out-of-bounds labels. Mostly a brute-force version of `~matplotlib.axis.Axis.set_smart_bounds` (which I couldn't get to work). """ # NOTE: Previously triggered this every time FixedFormatter was found # on axis but 1) that seems heavy-handed + strange and 2) internal # application of FixedFormatter by boxplot resulted in subsequent format() # successfully calling this and messing up the ticks for some reason. # So avoid using this when possible, and try to make behavior consistent # by cacheing the locators before we use them for ticks. axis = getattr(self, x + 'axis') sides = ('bottom', 'top') if x == 'x' else ('left', 'right') bounds = tuple(tuple(self.spines[side].get_bounds() or ()) for side in sides) if fixticks or any(bounds) or axis.get_scale() == 'cutoff': # Major locator lim = bounds[0] or bounds[1] or getattr(self, 'get_' + x + 'lim')() locator = getattr(axis, '_major_locator_cached', None) if locator is None: locator = axis._major_locator_cached = axis.get_major_locator() locator = constructor.Locator([x for x in locator() if lim[0] <= x <= lim[1]]) # noqa: E501 axis.set_major_locator(locator) # Minor locator locator = getattr(axis, '_minor_locator_cached', None) if locator is None: locator = axis._minor_locator_cached = axis.get_minor_locator() locator = constructor.Locator([x for x in locator() if lim[0] <= x <= lim[1]]) # noqa: E501 axis.set_minor_locator(locator) def _update_formatter( self, x, formatter=None, *, formatter_kw=None, tickrange=None, wraprange=None, ): """ Update the axis formatter. Passes `formatter` through `Formatter` with kwargs. """ # Is this a date axis? # NOTE: Make sure to get this *after* lims set! # See: https://matplotlib.org/api/units_api.html # And: https://matplotlib.org/api/dates_api.html # Also see: https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/axis.py # noqa # The axis_date() method just applies DateConverter axis = getattr(self, x + 'axis') date = isinstance(axis.converter, mdates.DateConverter) # Major formatter # NOTE: The only reliable way to disable ticks labels and then # restore them is by messing with the *formatter*, rather than # setting labelleft=False, labelright=False, etc. formatter_kw = formatter_kw or {} formatter_kw = formatter_kw.copy() if formatter is not None or tickrange is not None or wraprange is not None: # Tick range if tickrange is not None or wraprange is not None: if formatter not in (None, 'auto'): warnings._warn_proplot( 'The tickrange and autorange features require ' 'proplot.AutoFormatter formatter. Overriding the input.' ) formatter = 'auto' if tickrange is not None: formatter_kw.setdefault('tickrange', tickrange) if wraprange is not None: formatter_kw.setdefault('wraprange', wraprange) # Set the formatter # Note some formatters require 'locator' as keyword arg if formatter in ('date', 'concise'): locator = axis.get_major_locator() formatter_kw.setdefault('locator', locator) formatter = constructor.Formatter(formatter, date=date, **formatter_kw) axis.set_major_formatter(formatter) def _update_labels(self, x, *, label=None, color=None, **kwargs): """ Apply axis labels to the relevant shared axis. If spanning labels are toggled this keeps the labels synced for all subplots in the same row or column. Label positions will be adjusted at draw-time with figure._align_axislabels. """ # First get settings kw = rc.fill( { 'color': 'axes.labelcolor', 'weight': 'axes.labelweight', 'fontsize': 'axes.labelsize', 'fontfamily': 'font.family', }, context=True, ) if label is not None: kw['text'] = label if color: kw['color'] = color kw.update(kwargs) if not kw: return # Get axes in 3 step process # 1. Walk to parent if it is a main axes # 2. Get spanning main axes in this row or column (ignore short panel edges) # 3. Walk to parent if it exists (may be a panel long edge) # NOTE: Initially we keep spanning labels off # NOTE: Axis sharing between "main" axes is only ever one level deep. # NOTE: Critical to apply labels to *shared* axes attributes rather # than testing extents or we end up sharing labels with twin axes. ax = self if getattr(self.figure, '_share' + x) > 0: share = getattr(ax, '_share' + x) or ax if not share._panel_parent: ax = share # Get spanning axes axs = [ax] if getattr(ax.figure, '_span' + x): side = getattr(self, x + 'axis').get_label_position() if side in ('left', 'bottom'): axs = ax._get_side_axes(side, panels=False) # Update axes with label for ax in axs: ax = getattr(ax, '_share' + x) or ax # defer to panel axis = getattr(ax, x + 'axis') axis.label.update(kw) def _update_locators( self, x, locator=None, minorlocator=None, *, tickminor=None, locator_kw=None, minorlocator_kw=None, ): """ Update the locators. Requires `Locator` instances. """ # Major locator # NOTE: Parts of API (dualxy) rely on minor tick toggling preserving the # isDefault_minloc setting. In future should override mpl minorticks_on() # NOTE: Unlike matplotlib when "turning on" minor ticks we *always* use the # scale default, thanks to scale classes refactoring with _ScaleBase. axis = getattr(self, x + 'axis') locator_kw = locator_kw or {} if locator is not None: locator = constructor.Locator(locator, **locator_kw) if locator is not None: axis.set_major_locator(locator) if isinstance(locator, mticker.IndexLocator): tickminor = False # 'index' minor ticks make no sense # Minor locator minorlocator_kw = minorlocator_kw or {} if minorlocator is not None: minorlocator = constructor.Locator(minorlocator, **minorlocator_kw) if tickminor or minorlocator: isdefault = minorlocator is None if isdefault: minorlocator = getattr(axis._scale, '_default_minor_locator', None) if not minorlocator: minorlocator = constructor.Locator('minor') axis.set_minor_locator(minorlocator) axis.isDefault_minloc = isdefault elif tickminor is not None and not tickminor: # NOTE: Generally if you *enable* minor ticks on a dual axis, want to # allow FuncScale updates to change the minor tick locators. If you # *disable* minor ticks, do not want FuncScale applications to turn them # on. So we allow below to set isDefault_minloc to False. axis.set_minor_locator(constructor.Locator('null')) def _update_limits(self, x, *, min_=None, max_=None, lim=None, reverse=None): """ Update the axis limits. """ # Set limits for just one side or both at once axis = getattr(self, x + 'axis') if min_ is not None or max_ is not None: if lim is not None: warnings._warn_proplot( f'Overriding {x}lim={lim!r} ' f'with {x}min={min_!r} and {x}max={max_!r}.' ) lim = (min_, max_) if lim is not None: getattr(self, 'set_' + x + 'lim')(lim) # Reverse direction # NOTE: 3.1+ has axis.set_inverted(), below is from source code if reverse is not None: lo, hi = axis.get_view_interval() if reverse: lim = (max(lo, hi), min(lo, hi)) else: lim = (min(lo, hi), max(lo, hi)) axis.set_view_interval(*lim, ignore=True) def _update_spines(self, x, *, loc='both', bounds=None, color=None, linewidth=None): """ Update the spine settings. """ # Get rc properties kw = rc.fill( { 'color': 'axes.edgecolor', 'linewidth': 'axes.linewidth', }, context=True, ) if color is not None: kw['color'] = color if linewidth is not None: kw['linewidth'] = linewidth # Iterate over spines associated with this axis sides = ('bottom', 'top') if x == 'x' else ('left', 'right') for side in sides: # Line properties. Override if we're settings spine bounds # In this case just have spines on edges by default spine = self.spines[side] if bounds is not None and loc not in sides: loc = sides[0] # Eliminate sides if loc == 'neither': spine.set_visible(False) elif loc == 'both': spine.set_visible(True) elif loc in sides: # make relevant spine visible spine.set_visible(side == loc) # Special spine location, usually 'zero', 'center', or tuple with # (units, location) where 'units' can be 'axes', 'data', or 'outward' elif loc is not None: if side == sides[1]: spine.set_visible(False) else: spine.set_visible(True) try: spine.set_position(loc) except ValueError: raise ValueError( f'Invalid {x} spine location {loc!r}. ' 'Options are: ' + ', '.join(map( repr, (*sides, 'both', 'neither') )) + '.' ) # Apply spine bounds if bounds is not None and spine.get_visible(): spine.set_bounds(*bounds) spine.update(kw) def _update_ticks( self, x, *, grid=None, gridminor=None, color=None, gridcolor=None, ticklen=None, tickloc=None, ticklabelloc=None, labelloc=None, tickdir=None, ticklabeldir=None, rotation=None, ): """ Update the ticks and gridlines. """ # Initial stuff axis = getattr(self, x + 'axis') sides = ('bottom', 'top') if x == 'x' else ('left', 'right') sides_active = tuple(side for side in sides if self.spines[side].get_visible()) # Tick and grid settings for major and minor ticks separately # Override is just a "new default", but user can override this for which, igrid in zip(('major', 'minor'), (grid, gridminor)): # Tick properties # NOTE: This loads xtick.major.size, xtick.major.width, # xtick.major.pad, xtick.major.bottom, and xtick.major.top # For all the x/y major/minor tick types kwticks = rc.category(x + 'tick.' + which, context=True) if kwticks is None: kwticks = {} else: kwticks.pop('visible', None) # invalid setting if ticklen is not None: kwticks['size'] = units(ticklen, 'pt') if which == 'minor': kwticks['size'] *= rc['tick.lenratio'] # Grid style and toggling name = 'grid' if which == 'major' else 'gridminor' if igrid is not None: axis.grid(igrid, which=which) kwgrid = rc.fill( { 'grid_color': name + '.color', 'grid_alpha': name + '.alpha', 'grid_linewidth': name + '.linewidth', 'grid_linestyle': name + '.linestyle', }, context=True, ) if gridcolor is not None: # override for specific x/y axes kwgrid['grid_color'] = gridcolor axis.set_tick_params(which=which, **kwgrid, **kwticks) # Tick and ticklabel properties that apply equally for major/minor lines # Weird issue causes set_tick_params to reset/forget grid is turned on if # you access tick.gridOn directly, instead of passing through tick_params. # Since gridOn is undocumented feature, don't use it. So calling _format_axes # a second time will remove the lines. First determine tick sides, avoiding # situation where we draw ticks on top of invisible spine. kw = {} loc2sides = { None: None, 'both': sides, 'none': (), 'neither': (), } if tickloc not in sides and any(self.spines[_].get_bounds() is not None for _ in sides): # noqa: E501 tickloc = sides[0] # override to just one side ticklocs = loc2sides.get(tickloc, (tickloc,)) if ticklocs is not None: kw.update({side: side in ticklocs for side in sides}) kw.update({side: False for side in sides if side not in sides_active}) # Tick label sides # Will override to make sure only appear where ticks are ticklabellocs = loc2sides.get(ticklabelloc, (ticklabelloc,)) if ticklabellocs is not None: kw.update({'label' + side: (side in ticklabellocs) for side in sides}) kw.update( { 'label' + side: False for side in sides if side not in sides_active or (ticklocs is not None and side not in ticklocs) } ) # The axis label side if labelloc is None: if ticklocs is not None: options = tuple( side for side in sides if side in ticklocs and side in sides_active ) if len(options) == 1: labelloc = options[0] elif labelloc not in sides: raise ValueError( f'Got labelloc {labelloc!r}, valid options are ' + ', '.join(map(repr, sides)) + '.' ) # Apply axis.set_tick_params(which='both', **kw) if labelloc is not None: axis.set_label_position(labelloc) # Tick label settings kw = rc.fill( { 'labelcolor': 'tick.labelcolor', # new props 'labelsize': 'tick.labelsize', 'color': x + 'tick.color', }, context=True, ) if color: kw['color'] = color kw['labelcolor'] = color # Tick label direction and rotation if tickdir == 'in': # ticklabels should be much closer kw['pad'] = 1.0 if ticklabeldir == 'in': # put tick labels inside the plot tickdir = 'in' kw['pad'] = -rc[f'{x}tick.major.size'] - rc[f'{x}tick.major.pad'] kw['pad'] -= rc._scale_font(rc[f'{x}tick.labelsize']) if tickdir is not None: kw['direction'] = tickdir axis.set_tick_params(which='both', **kw) # Settings that can't be controlled by set_tick_params # Also set rotation and alignment here kw = rc.fill( { 'fontfamily': 'font.family', 'weight': 'tick.labelweight' }, context=True, ) if rotation is not None: kw = {'rotation': rotation} if x == 'x': self._datex_rotated = True if rotation not in (0, 90, -90): kw['ha'] = ('right' if rotation > 0 else 'left') for t in axis.get_ticklabels(): t.update(kw)
[docs] @docstring.add_snippets def format( self, *, aspect=None, xloc=None, yloc=None, xspineloc=None, yspineloc=None, xtickloc=None, ytickloc=None, fixticks=False, xlabelloc=None, ylabelloc=None, xticklabelloc=None, yticklabelloc=None, xtickdir=None, ytickdir=None, xgrid=None, ygrid=None, xgridminor=None, ygridminor=None, xtickminor=None, ytickminor=None, xticklabeldir=None, yticklabeldir=None, xtickrange=None, ytickrange=None, xwraprange=None, ywraprange=None, xreverse=None, yreverse=None, xlabel=None, ylabel=None, xlim=None, ylim=None, xmin=None, ymin=None, xmax=None, ymax=None, xscale=None, yscale=None, xrotation=None, yrotation=None, xformatter=None, yformatter=None, xticklabels=None, yticklabels=None, xticks=None, yticks=None, xlocator=None, ylocator=None, xminorticks=None, yminorticks=None, xminorlocator=None, yminorlocator=None, xbounds=None, ybounds=None, xmargin=None, ymargin=None, xcolor=None, ycolor=None, xlinewidth=None, ylinewidth=None, xgridcolor=None, ygridcolor=None, xticklen=None, yticklen=None, xlabel_kw=None, ylabel_kw=None, xscale_kw=None, yscale_kw=None, xlocator_kw=None, ylocator_kw=None, xformatter_kw=None, yformatter_kw=None, xminorlocator_kw=None, yminorlocator_kw=None, patch_kw=None, **kwargs ): """ Modify the *x* and *y* axis labels, tick locations, tick labels, axis scales, spine settings, and more. Unknown keyword arguments are passed to `Axes.format` and `~proplot.config.RcConfigurator.context`. Parameters ---------- aspect : {'auto', 'equal'}, optional The aspect ratio mode. See `~matplotlib.axes.Axes.set_aspect` for details. xlabel, ylabel : str, optional The *x* and *y* axis labels. Applied with `~matplotlib.axes.Axes.set_xlabel` and `~matplotlib.axes.Axes.set_ylabel`. xlabel_kw, ylabel_kw : dict-like, optional The *x* and *y* axis label settings. Applied with the `~matplotlib.artist.Artist.update` method on the `~matplotlib.text.Text` instance. Options include ``'color'``, ``'size'``, and ``'weight'``. xlim, ylim : 2-tuple of floats or None, optional The *x* and *y* axis data limits. Applied with `~matplotlib.axes.Axes.set_xlim` and `~matplotlib.axes.Axes.set_ylim`. xmin, ymin : float, optional The *x* and *y* minimum data limits. Useful if you do not want to set the maximum limits. xmax, ymax : float, optional The *x* and *y* maximum data limits. Useful if you do not want to set the minimum limits. xreverse, yreverse : bool, optional Sets whether the *x* and *y* axis are oriented in the "reverse" direction. The "normal" direction is increasing to the right for the *x* axis and to the top for the *y* axis. The "reverse" direction is increasing to the left for the *x* axis and to the bottom for the *y* axis. xscale, yscale : axis scale spec, optional The *x* and *y* axis scales. Passed to the `~proplot.scale.Scale` constructor. For example, ``xscale='log'`` applies logarithmic scaling, and ``xscale=('cutoff', 0.5, 2)`` applies a custom `~proplot.scale.CutoffScale`. xscale_kw, yscale_kw : dict-like, optional The *x* and *y* axis scale settings. Passed to `~proplot.scale.Scale`. xspineloc, yspineloc \ : {'both', 'bottom', 'top', 'left', 'right', 'neither', 'center', 'zero'}, optional The *x* and *y* axis spine locations. xloc, yloc : optional Aliases for `xspineloc`, `yspineloc`. xtickloc, ytickloc \ : {'both', 'bottom', 'top', 'left', 'right', 'neither'}, optional Which *x* and *y* axis spines should have major and minor tick marks. xtickminor, ytickminor : bool, optional Whether to draw minor ticks on the *x* and *y* axes. xtickdir, ytickdir : {'out', 'in', 'inout'} Direction that major and minor tick marks point for the *x* and *y* axis. xgrid, ygrid : bool, optional Whether to draw major gridlines on the *x* and *y* axis. Use `grid` to toggle both. xgridminor, ygridminor : bool, optional Whether to draw minor gridlines for the *x* and *y* axis. Use `gridminor` to toggle both. xticklabeldir, yticklabeldir : {'in', 'out'} Whether to place *x* and *y* axis tick label text inside or outside the axes. xlocator, ylocator : locator spec, optional Used to determine the *x* and *y* axis tick mark positions. Passed to the `~proplot.constructor.Locator` constructor. Can be float, list of float, string, or `matplotlib.ticker.Locator` instance. xticks, yticks : optional Aliases for `xlocator`, `ylocator`. xlocator_kw, ylocator_kw : dict-like, optional Keyword arguments passed to the `matplotlib.ticker.Locator` class. xminorlocator, yminorlocator : optional As for `xlocator`, `ylocator`, but for the minor ticks. xminorticks, yminorticks : optional Aliases for `xminorlocator`, `yminorlocator`. xminorlocator_kw, yminorlocator_kw As for `xlocator_kw`, `ylocator_kw`, but for the minor locator. xformatter, yformatter : formatter spec, optional Used to determine the *x* and *y* axis tick label string format. Passed to the `~proplot.constructor.Formatter` constructor. Can be string, list of strings, or `matplotlib.ticker.Formatter` instance. Use ``[]`` or ``'null'`` for no ticks. xticklabels, yticklabels : optional Aliases for `xformatter`, `yformatter`. xformatter_kw, yformatter_kw : dict-like, optional Keyword arguments passed to the `matplotlib.ticker.Formatter` class. xrotation, yrotation : float, optional The rotation for *x* and *y* axis tick labels. Default is ``0`` for normal axes, :rc:`formatter.timerotation` for time *x* axes. xtickrange, ytickrange : (float, float), optional The *x* and *y* axis data ranges within which major tick marks are labelled. For example, the tick range ``(-1, 1)`` with axis range ``(-5, 5)`` and a tick interval of 1 will only label the ticks marks at -1, 0, and 1. See `~proplot.ticker.AutoFormatter` for details. xwraprange, ywraprange : (float, float), optional The *x* and *y* axis data ranges with which major tick mark values are *wrapped*. For example, the wrap range ``(0, 3)`` causes the values 0 through 9 to be formatted as 0, 1, 2, 0, 1, 2, 0, 1, 2, 0. See `~proplot.ticker.AutoFormatter` for details. xmargin, ymargin : float, optional The default margin between plotted content and the *x* and *y* axis spines. Value is proportional to the width, height of the axes. Use this if you want whitespace between plotted content and the spines, but don't want to explicitly set `xlim` or `ylim`. xbounds, ybounds : (float, float), optional The *x* and *y* axis data bounds within which to draw the spines. For example, the axis range ``(0, 4)`` with bounds ``(1, 4)`` will prevent the spines from meeting at the origin. xcolor, ycolor : color-spec, optional Color for the *x* and *y* axis spines, ticks, tick labels, and axis labels. Default is :rc:`color`. Use e.g. ``ax.format(color='red')`` to set for both axes. xlinewidth, ylinewidth : color-spec, optional Line width for the *x* and *y* axis spines and major ticks. Default is :rc:`linewidth`. Use e.g. ``ax.format(linewidth=2)`` to set for both axes. xgridcolor, ygridcolor : color-spec, optional Color for the *x* and *y* axis major and minor gridlines. Default is :rc:`grid.color`. Use e.g. ``ax.format(gridcolor='r')`` to set for both axes. xticklen, yticklen : float or str, optional Tick lengths for the *x* and *y* axis. Units are interpreted by `~proplot.utils.units`, with "points" as the numeric unit. Default is :rc:`ticklen`. Minor tick lengths are scaled according to :rc:`tick.lenratio`. Use e.g. ``ax.format(ticklen=1)`` to set for both axes. fixticks : bool, optional Whether to always transform the tick locators to a `~matplotlib.ticker.FixedLocator` instance. Default is ``False``. If your axis ticks are doing weird things (for example, ticks drawn outside of the axis spine), try setting this to ``True``. %(axes.patch_kw)s Other parameters ---------------- %(axes.other)s See also -------- proplot.config.RcConfigurator.context proplot.axes.Axes.format Note ---- If you plot something with a `datetime64 \ <https://docs.scipy.org/doc/numpy/reference/arrays.datetime.html>`__, `pandas.Timestamp`, `pandas.DatetimeIndex`, `datetime.date`, `datetime.time`, or `datetime.datetime` array as the *x* or *y* axis coordinate, the axis ticks and tick labels will be automatically formatted as dates. """ rc_kw, rc_mode, kwargs = self._parse_format(**kwargs) with rc.context(rc_kw, mode=rc_mode): # Background patch kw_face = rc.fill( { 'facecolor': 'axes.facecolor', 'alpha': 'axes.alpha' }, context=True, ) patch_kw = patch_kw or {} kw_face.update(patch_kw) self.patch.update(kw_face) # No mutable default args xlabel_kw = xlabel_kw or {} ylabel_kw = ylabel_kw or {} xscale_kw = xscale_kw or {} yscale_kw = yscale_kw or {} xlocator_kw = xlocator_kw or {} ylocator_kw = ylocator_kw or {} xformatter_kw = xformatter_kw or {} yformatter_kw = yformatter_kw or {} xminorlocator_kw = xminorlocator_kw or {} yminorlocator_kw = yminorlocator_kw or {} # Flexible keyword args, declare defaults xmargin = _not_none(xmargin, rc.get('axes.xmargin', context=True)) ymargin = _not_none(ymargin, rc.get('axes.ymargin', context=True)) xtickdir = _not_none(xtickdir, rc.get('xtick.direction', context=True)) ytickdir = _not_none(ytickdir, rc.get('ytick.direction', context=True)) xformatter = _not_none(xformatter=xformatter, xticklabels=xticklabels) yformatter = _not_none(yformatter=yformatter, yticklabels=yticklabels) xlocator = _not_none(xlocator=xlocator, xticks=xticks) ylocator = _not_none(ylocator=ylocator, yticks=yticks) xtickminor = _not_none( xtickminor, rc.get('xtick.minor.visible', context=True) ) ytickminor = _not_none( ytickminor, rc.get('ytick.minor.visible', context=True) ) xminorlocator = _not_none( xminorlocator=xminorlocator, xminorticks=xminorticks, ) yminorlocator = _not_none( yminorlocator=yminorlocator, yminorticks=yminorticks, ) # Grid defaults are more complicated grid = rc.get('axes.grid', context=True) which = rc.get('axes.grid.which', context=True) if which is not None or grid is not None: # if *one* was changed axis = rc['axes.grid.axis'] # always need this property if grid is None: grid = rc['axes.grid'] elif which is None: which = rc['axes.grid.which'] xgrid = _not_none( xgrid, grid and axis in ('x', 'both') and which in ('major', 'both') ) ygrid = _not_none( ygrid, grid and axis in ('y', 'both') and which in ('major', 'both') ) xgridminor = _not_none( xgridminor, grid and axis in ('x', 'both') and which in ('minor', 'both') ) ygridminor = _not_none( ygridminor, grid and axis in ('y', 'both') and which in ('minor', 'both') ) # Sensible defaults for spine, tick, tick label, and label locs # NOTE: Allow tick labels to be present without ticks! User may # want this sometimes! Same goes for spines! xspineloc = _not_none(xloc=xloc, xspineloc=xspineloc) yspineloc = _not_none(yloc=yloc, yspineloc=yspineloc) xtickloc = _not_none(xtickloc, xspineloc, _parse_rcloc('x', 'xtick')) ytickloc = _not_none(ytickloc, yspineloc, _parse_rcloc('y', 'ytick')) xspineloc = _not_none(xspineloc, _parse_rcloc('x', 'axes.spines')) yspineloc = _not_none(yspineloc, _parse_rcloc('y', 'axes.spines')) if xtickloc != 'both': xticklabelloc = _not_none(xticklabelloc, xtickloc) xlabelloc = _not_none(xlabelloc, xticklabelloc) if xlabelloc not in (None, 'bottom', 'top'): # e.g. "both" xlabelloc = 'bottom' if ytickloc != 'both': yticklabelloc = _not_none(yticklabelloc, ytickloc) ylabelloc = _not_none(ylabelloc, yticklabelloc) if ylabelloc not in (None, 'left', 'right'): ylabelloc = 'left' # Begin loop for ( x, axis, label, color, linewidth, gridcolor, ticklen, margin, bounds, tickloc, spineloc, ticklabelloc, labelloc, grid, gridminor, tickminor, minorlocator, min_, max_, lim, reverse, scale, locator, tickrange, wraprange, formatter, tickdir, ticklabeldir, rotation, label_kw, scale_kw, locator_kw, minorlocator_kw, formatter_kw ) in zip( ('x', 'y'), (self.xaxis, self.yaxis), (xlabel, ylabel), (xcolor, ycolor), (xlinewidth, ylinewidth), (xgridcolor, ygridcolor), (xticklen, yticklen), (xmargin, ymargin), (xbounds, ybounds), (xtickloc, ytickloc), (xspineloc, yspineloc), (xticklabelloc, yticklabelloc), (xlabelloc, ylabelloc), (xgrid, ygrid), (xgridminor, ygridminor), (xtickminor, ytickminor), (xminorlocator, yminorlocator), (xmin, ymin), (xmax, ymax), (xlim, ylim), (xreverse, yreverse), (xscale, yscale), (xlocator, ylocator), (xtickrange, ytickrange), (xwraprange, ywraprange), (xformatter, yformatter), (xtickdir, ytickdir), (xticklabeldir, yticklabeldir), (xrotation, yrotation), (xlabel_kw, ylabel_kw), (xscale_kw, yscale_kw), (xlocator_kw, ylocator_kw), (xminorlocator_kw, yminorlocator_kw), (xformatter_kw, yformatter_kw), ): # Axis scale # WARNING: This relies on monkey patch of mscale.scale_factory # that allows it to accept a custom scale class! # WARNING: Changing axis scale also changes default locators # and formatters, and restricts possible range of axis limits, # so critical to do it first. if scale is not None: scale = constructor.Scale(scale, **scale_kw) getattr(self, 'set_' + x + 'scale')(scale) # Axis limits self._update_limits( x, min_=min_, max_=max_, lim=lim, reverse=reverse ) if margin is not None: self.margins(**{x: margin}) # Axis spine settings self._update_spines( x, loc=spineloc, bounds=bounds, linewidth=linewidth, color=color, ) # Axis tick settings self._update_ticks( x, grid=grid, gridminor=gridminor, color=color, gridcolor=gridcolor, ticklen=ticklen, tickloc=tickloc, ticklabelloc=ticklabelloc, labelloc=labelloc, tickdir=tickdir, ticklabeldir=ticklabeldir, rotation=rotation, ) # Axis label settings # NOTE: This has to come after set_label_position, or # ha or va overrides in label_kw are overwritten self._update_labels( x, label=label, color=color, **label_kw ) # Axis locator if minorlocator is True or minorlocator is False: # must test identity warnings._warn_proplot( f'You passed {x}minorticks={minorlocator}, but this ' 'argument is used to specify tick *locations*. If ' 'you just want to *toggle* minor ticks on and off, ' f'please use {x}tickminor=True or {x}tickminor=False.' ) minorlocator = None self._update_locators( x, locator, minorlocator, tickminor=tickminor, locator_kw=locator_kw, minorlocator_kw=minorlocator_kw, ) # Axis formatter self._update_formatter( x, formatter, formatter_kw=formatter_kw, tickrange=tickrange, wraprange=wraprange, ) # Ensure ticks are within axis bounds self._update_bounds(x, fixticks=fixticks) # Call parent if aspect is not None: self.set_aspect(aspect) super().format(**kwargs)
[docs] @docstring.add_snippets def altx(self, **kwargs): """ %(axes.altx)s """ # NOTE: Cannot *wrap* twiny() because we want to use CartesianAxes, not # matplotlib Axes. Instead use hidden method SubplotBase._make_twin_axes. # WARNING: This repairs a matplotlib bug where twins fail to inherit the minor # locator due to application of `AutoMinorLocator` when `ytick.minor.visible` # is ``True`` in `Axes.cla` and due to the fact that passing ``sharey=self`` # to the alternate axes means that they share the same major and minor Tickers. # >>> import matplotlib.pyplot as plt # ... fig, ax = plt.subplots() # ... ax.set_yscale('log') # ... ax.twiny() # WARNING: We add axes as children for tight layout algorithm convenience and # to support eventual paradigm of arbitrarily many duplicates with spines # arranged in an edge stack. However this means all artists drawn there take # on zorder of their axes when drawn inside the "parent" (see Axes.draw()). # To restore matplotlib behavior, which draws "child" artists on top simply # because the axes was created after the "parent" one, use the inset_axes # zorder of 4 and make the background transparent. minorlocator = self.yaxis.get_minor_locator() with self.figure._context_authorize_add_subplot(): ax = self._make_twin_axes(sharey=self, projection='proplot_cartesian') # Child defaults ax._altx_parent = self ax.yaxis.set_minor_locator(minorlocator) ax.yaxis.isDefault_minloc = True for side, spine in ax.spines.items(): spine.set_visible(side == 'top') ax.xaxis.tick_top() ax.xaxis.set_label_position('top') ax.yaxis.set_visible(False) ax.patch.set_visible(False) ax.grid(False) ax.set_zorder(4) ax.set_autoscaley_on(self.get_autoscaley_on()) # Parent defaults self.spines['top'].set_visible(False) self.spines['bottom'].set_visible(True) self.xaxis.tick_bottom() self.xaxis.set_label_position('bottom') # Add axes self.add_child_axes(ax) # to facilitate tight layout self.figure._axstack.remove(ax) # or gets drawn twice! ax.format(**_parse_alt('x', kwargs)) return ax
[docs] @docstring.add_snippets def alty(self, **kwargs): """ %(axes.alty)s """ # See altx() comments minorlocator = self.xaxis.get_minor_locator() with self.figure._context_authorize_add_subplot(): ax = self._make_twin_axes(sharex=self, projection='proplot_cartesian') # Child defaults ax._alty_parent = self ax.xaxis.set_minor_locator(minorlocator) ax.xaxis.isDefault_minloc = True for side, spine in ax.spines.items(): spine.set_visible(side == 'right') ax.yaxis.tick_right() ax.yaxis.set_label_position('right') ax.xaxis.set_visible(False) ax.patch.set_visible(False) ax.grid(False) ax.set_zorder(4) ax.set_autoscalex_on(self.get_autoscalex_on()) # Parent defaults self.spines['right'].set_visible(False) self.spines['left'].set_visible(True) self.yaxis.tick_left() self.yaxis.set_label_position('left') # Add axes self.add_child_axes(ax) # to facilitate tight layout self.figure._axstack.remove(ax) # or gets drawn twice! ax.format(**_parse_alt('y', kwargs)) return ax
[docs] @docstring.add_snippets def dualx(self, funcscale, **kwargs): """ %(axes.dualx)s """ # NOTE: Matplotlib 3.1 has a 'secondary axis' feature. For the time # being, our version is more robust (see FuncScale) and simpler, since # we do not create an entirely separate _SecondaryAxis class. ax = self.altx(**kwargs) ax._dualx_funcscale = funcscale ax._dualx_scale() return ax
[docs] @docstring.add_snippets def dualy(self, funcscale, **kwargs): """ %(axes.dualy)s """ ax = self.alty(**kwargs) ax._dualy_funcscale = funcscale ax._dualy_scale() return ax
def draw(self, renderer=None, *args, **kwargs): # Perform extra post-processing steps # NOTE: This mimics matplotlib API, which calls identical # post-processing steps in both draw() and get_tightbbox() self._dualx_scale() self._dualy_scale() self._datex_rotate() self._apply_axis_sharing() if self._inset_parent is not None and self._inset_zoom: self.indicate_inset_zoom() super().draw(renderer, *args, **kwargs) def get_tightbbox(self, renderer, *args, **kwargs): # Perform extra post-processing steps self._dualx_scale() self._dualy_scale() self._datex_rotate() self._apply_axis_sharing() if self._inset_parent is not None and self._inset_zoom: self.indicate_inset_zoom() return super().get_tightbbox(renderer, *args, **kwargs)
[docs] @docstring.add_snippets def twinx(self): """ %(axes.twinx)s """ return self.alty()
[docs] @docstring.add_snippets def twiny(self): """ %(axes.twiny)s """ return self.altx()