Source code for proplot.config

#!/usr/bin/env python3
Tools for setting up proplot and configuring global settings.
See the :ref:`configuration guide <ug_config>` for details.
# NOTE: The matplotlib analogue to this file is actually
# but it makes more sense to have all the setup actions in a separate file
# so the namespace of the top-level module is unpolluted.
# NOTE: Why also load colormaps and cycles in this file and not
# Because I think it makes sense to have all the code that "runs" (i.e. not
# just definitions) in the same place, and I was having issues with circular
# dependencies and where import order of was affecting behavior.
import logging
import os
import re
import sys
from collections import namedtuple
from import MutableMapping
from numbers import Real

import cycler
import matplotlib as mpl
import matplotlib.colors as mcolors
import matplotlib.font_manager as mfonts
import matplotlib.mathtext  # noqa: F401
import as mstyle
import numpy as np
from matplotlib import RcParams

from .internals import ic  # noqa: F401
from .internals import _not_none, _translate_grid, docstring, rcsetup, warnings

    from IPython import get_ipython
except ImportError:
    def get_ipython():

# Suppress warnings emitted by ( in recent versions)
# when when substituting dummy unavailable glyph due to fallback disabled.

__all__ = [
    'RcConfigurator',  # deprecated
    'inline_backend_fmt',  # deprecated

# Constants

# Configurator docstrings
_rc_docstring = """
local : bool, optional
    Whether to load settings from the `~Configurator.local_files` file.
    Default is ``True``.
user : bool, optional
    Whether to load settings from the `~Configurator.user_file` file.
    Default is ``True``.
default : bool, optional
    Whether to reload built-in default proplot settings.
    Default is ``True``.
docstring._snippet_manager['rc.params'] = _rc_docstring

# Registration docstrings
_shared_docstring = """
*args : path-spec or `~proplot.colors.{type}Colormap`, optional
    The {objects} to register. These can be file paths containing
    RGB data or `~proplot.colors.{type}Colormap` instances. By default,
    if positional arguments are passed, then `user` is set to ``False``.

    Valid file extensions are listed in the below table. Note that {objects}
    are registered according to their filenames -- for example, ````
    will be registered as ``'name'``.
"""  # noqa: E501
_cmap_exts_docstring = """
    ===================  ==========================================
    Extension            Description
    ===================  ==========================================
    ``.json``            JSON database of the channel segment data.
    ``.hex``             Comma-delimited list of HEX strings.
    ``.rgb``, ``.txt``   3-4 column table of channel values.
    ===================  ==========================================
_cycle_exts_docstring = """
    ==================  ==========================================
    Extension           Description
    ==================  ==========================================
    ``.hex``            Comma-delimited list of HEX strings.
    ``.rgb``, ``.txt``  3-4 column table of channel values.
    ==================  ==========================================
_color_docstring = """
*args : path-like or dict, optional
    The colors to register. These can be file paths containing RGB data or
    dictionary mappings of names to RGB values. By default, if positional
    arguments are passed, then `user` is set to ``False``. Files must have
    the extension ``.txt`` and should contain one line per color in the
    format ``name : hex``. Whitespace is ignored.
_font_docstring = """
*args : path-like, optional
    The font files to add. By default, if positional arguments are passed, then
    `user` is set to ``False``. Files must have the extensions ``.ttf`` or ``.otf``.
    See `this link \
    for a guide on converting other font files to ``.ttf`` and ``.otf``.
_register_docstring = """
user : bool, optional
    Whether to reload {objects} from `~Configurator.user_folder`. Default is
    ``False`` if positional arguments were passed and ``True`` otherwise.
default : bool, optional
    Whether to reload the default {objects} packaged with proplot.
    Default is always ``False``.
docstring._snippet_manager['rc.cmap_params'] = _register_docstring.format(objects='colormaps')  # noqa: E501
docstring._snippet_manager['rc.cycle_params'] = _register_docstring.format(objects='color cycles')  # noqa: E501
docstring._snippet_manager['rc.color_params'] = _register_docstring.format(objects='colors')  # noqa: E501
docstring._snippet_manager['rc.font_params'] = _register_docstring.format(objects='fonts')  # noqa: E501
docstring._snippet_manager['rc.cmap_args'] = _shared_docstring.format(objects='colormaps', type='Continuous')  # noqa: E501
docstring._snippet_manager['rc.cycle_args'] = _shared_docstring.format(objects='color cycles', type='Discrete')  # noqa: E501
docstring._snippet_manager['rc.color_args'] = _color_docstring
docstring._snippet_manager['rc.font_args'] = _font_docstring
docstring._snippet_manager['rc.cmap_exts'] = _cmap_exts_docstring
docstring._snippet_manager['rc.cycle_exts'] = _cycle_exts_docstring

def _init_user_file():
    Initialize .proplotrc file.
    file = Configurator.user_file()
    if not os.path.exists(file):
        Configurator._save_yaml(file, comment=True)

def _init_user_folders():
    Initialize .proplot folder.
    for subfolder in ('', 'cmaps', 'cycles', 'colors', 'fonts'):
        folder = Configurator.user_folder(subfolder)
        if not os.path.isdir(folder):

def _get_data_folders(folder, user=True, default=True, reverse=False):
    Return data folder paths in reverse order of precedence.
    # When loading colormaps, cycles, and colors, files in the latter
    # directories overwrite files in the former directories. When loading
    # fonts, the resulting paths need to be *reversed*.
    paths = []
    if default:
        paths.append(os.path.join(os.path.dirname(__file__), folder))
    if user:
    if reverse:
        paths = paths[::-1]
    return paths

def _get_matplotlib_defaults():
    Get the default rc parameters dictionary with deprecated parameters filtered.
    # NOTE: Use RcParams update to filter and translate deprecated settings
    # before actually applying them to rcParams down pipeline. This way we can
    # suppress warnings for deprecated default params but still issue warnings
    # when user-supplied stylesheets have deprecated params.
    # WARNING: Some deprecated rc params remain in dictionary as None so we
    # filter them out. Beware if hidden attribute changes.
    rcdict = _get_filtered_dict(mpl.rcParamsDefault, warn=False)
    with warnings.catch_warnings():
        warnings.simplefilter('ignore', mpl.MatplotlibDeprecationWarning)
        rcdict = dict(RcParams(rcdict))
    for attr in ('_deprecated_set', '_deprecated_remain_as_none'):
        if hasattr(mpl, attr):  # _deprecated_set is in matplotlib before v3
            for deprecated in getattr(mpl, attr):
                rcdict.pop(deprecated, None)
    return rcdict

def _get_filtered_dict(rcdict, warn=True):
    Filter out blacklisted style parameters.
    # NOTE: This implements bugfix:
    # This fix is *critical* for proplot because we always run style.use()
    # when the configurator is made. Without fix backend is reset every time
    # you import proplot in jupyter notebooks. So apply retroactively.
    rcdict_filtered = {}
    for key in rcdict:
        if key in mstyle.STYLE_BLACKLIST:
            if warn:
                    f'Dictionary includes a parameter, {key!r}, that is not related '
                    'to style. Ignoring.'
            rcdict_filtered[key] = rcdict[key]
    return rcdict_filtered

def _get_style_dicts(style, infer=False, filter=True):
    Return a dictionary of settings belonging to the requested style(s). If `infer`
    is ``True``, two dictionaries are returned, where the second contains custom
    proplot settings "inferred" from the matplotlib settings. If `filter` is ``True``,
    invalid style parameters like `backend` are filtered out.
    # NOTE: This is adapted from matplotlib source for the following changes:
    # 1. Add an 'original' pseudo style. Like rcParamsOrig except we also reload
    #    from the user matplotlibrc file.
    # 2. When the style is changed we reset to the default state ignoring matplotlibrc.
    #    By contrast matplotlib applies styles on top of current state (including
    #    matplotlibrc changes and runtime rcParams changes) but the word 'style'
    #    implies a rigid static format. This makes more sense.
    # 3. Add a separate function that returns lists of style dictionaries so that
    #    we can modify the active style in a context block. Proplot context is more
    #    conservative than matplotlib's rc_context because it gets called a lot
    #    (e.g. every time you make an axes and every format() call). Instead of
    #    copying the entire rcParams dict we just track the keys that were changed.
    style_aliases = {
        '538': 'fivethirtyeight',
        'mpl20': 'default',
        'mpl15': 'classic',
        'original': mpl.matplotlib_fname(),

    # Always apply the default style *first* so styles are rigid
    kw_matplotlib = _get_matplotlib_defaults()
    if style == 'default' or style is mpl.rcParamsDefault:
        if infer:
            kw_proplot = _infer_added_params(kw_matplotlib)
            return kw_matplotlib, kw_proplot
            return kw_matplotlib

    # Apply limited deviations from the matplotlib style that we want to propagate to
    # other styles. Want users selecting stylesheets to have few surprises, so
    # currently just enforce the new aesthetically pleasing fonts.
    kw_matplotlib[''] = 'sans-serif'
    for fmly in ('serif', 'sans-serif', 'monospace', 'cursive', 'fantasy'):
        kw_matplotlib['font.' + fmly] = rcsetup._rc_matplotlib_default['font.' + fmly]

    # Apply user input style(s) one by one
    if isinstance(style, str) or isinstance(style, dict):
        styles = [style]
        styles = style
    for style in styles:
        if isinstance(style, dict):
            kw = style
        elif isinstance(style, str):
            style = style_aliases.get(style, style)
            if style in mstyle.library:
                kw = mstyle.library[style]
                    kw = mpl.rc_params_from_file(style, use_default_template=False)
                except IOError:
                    raise IOError(
                        f'Style {style!r} not found in the style library and input '
                        'is not a valid URL or file path. Available styles are: '
                        + ', '.join(map(repr, mstyle.available))
                        + '.'
            raise ValueError(f'Invalid style {style!r}. Must be string or dictionary.')
        if filter:
            kw = _get_filtered_dict(kw, warn=True)

    # Infer proplot params from stylesheet params
    if infer:
        kw_proplot = _infer_added_params(kw_matplotlib)
        return kw_matplotlib, kw_proplot
        return kw_matplotlib

def _infer_added_params(kw_params):
    Infer values for proplot's "added" parameters from stylesheet parameters.
    kw_proplot = {}
    mpl_to_proplot = {
        'xtick.labelsize': (
            'tick.labelsize', 'grid.labelsize',
        'ytick.labelsize': (
            'tick.labelsize', 'grid.labelsize',
        'axes.titlesize': (
            'abc.size', 'suptitle.size', 'title.size',
            'leftlabel.size', 'rightlabel.size',
            'toplabel.size', 'bottomlabel.size',
        'text.color': (
            'abc.color', 'suptitle.color', 'title.color',
            'tick.labelcolor', 'grid.labelcolor',
            'leftlabel.color', 'rightlabel.color',
            'toplabel.color', 'bottomlabel.color',
    for key, params in mpl_to_proplot.items():
        if key in kw_params:
            value = kw_params[key]
            for param in params:
                kw_proplot[param] = value
    return kw_proplot

def _iter_data_objects(folder, *args, **kwargs):
    Iterate over input objects and files in the data folders that should be
    registered. Also yield an index indicating whether these are user files.
    i = 0  # default files
    for i, path in enumerate(_get_data_folders(folder, **kwargs)):
        for dirname, dirnames, filenames in os.walk(path):
            for filename in filenames:
                if filename[0] == '.':  # UNIX-style hidden files
                path = os.path.join(dirname, filename)
                yield i, path
    i += 1  # user files
    for path in args:
        path = os.path.expanduser(path)
        if os.path.exists(path):
            yield i, path
            raise FileNotFoundError(f'Invalid file path {path!r}.')

[docs]def config_inline_backend(fmt=None): """ Set up the ipython `inline backend display format \ <>`__ and ensure that inline figures always look the same as saved figures. This runs the following ipython magic commands: .. code-block:: ipython %config InlineBackend.figure_formats = rc['inlinefmt'] %config InlineBackend.rc = {} # never override rc settings %config InlineBackend.close_figures = True # cells start with no active figures %config InlineBackend.print_figure_kwargs = {'bbox_inches': None} When the inline backend is inactive or unavailable, this has no effect. This function is called when you modify the :rcraw:`inlinefmt` property. Parameters ---------- fmt : str or sequence, optional The inline backend file format(s). Default is :rc:`inlinefmt`. Valid formats include ``'jpg'``, ``'png'``, ``'svg'``, ``'pdf'``, and ``'retina'``. See also -------- Configurator """ # Note if inline backend is unavailable this will fail silently ipython = get_ipython() if ipython is None: return fmt = _not_none(fmt, rc_proplot['inlinefmt']) if isinstance(fmt, str): fmt = [fmt] elif np.iterable(fmt): fmt = list(fmt) else: raise ValueError(f'Invalid inline backend format {fmt!r}. Must be string.') ipython.magic('config InlineBackend.figure_formats = ' + repr(fmt)) ipython.magic('config InlineBackend.rc = {}') ipython.magic('config InlineBackend.close_figures = True') ipython.magic("config InlineBackend.print_figure_kwargs = {'bbox_inches': None}")
[docs]def use_style(style): """ Apply the `matplotlib style(s) \ <>`__ with ``. This function is called when you modify the :rcraw:`style` property. Parameters ---------- style : str or sequence or dict-like The matplotlib style name(s) or stylesheet filename(s), or dictionary(s) of settings. Use ``'default'`` to apply matplotlib default settings and ``'original'`` to include settings from your user ``matplotlibrc`` file. See also -------- Configurator """ # NOTE: This function is not really necessary but makes proplot's # stylesheet-supporting features obvious. Plus changing the style does # so much *more* than changing rc params or quick settings, so it is # nice to have dedicated function instead of just another rc_param name. kw_matplotlib, kw_proplot = _get_style_dicts(style, infer=True) rc_matplotlib.update(kw_matplotlib) rc_proplot.update(kw_proplot)
[docs]@docstring._snippet_manager def register_cmaps(*args, user=None, default=False): """ Register named colormaps. This is called on import. Parameters ---------- %(rc.cmap_args)s %(rc.cmap_exts)s %(rc.cmap_params)s See also -------- register_cycles register_colors register_fonts proplot.demos.show_cmaps """ # Register input colormaps from . import colors as pcolors user = _not_none(user, not bool(args)) # skip user folder if input args passed paths = [] for arg in args: if isinstance(arg, mcolors.Colormap): pcolors._cmap_database[] = arg else: paths.append(arg) # Register data files for i, path in _iter_data_objects('cmaps', *paths, user=user, default=default): cmap = pcolors.ContinuousColormap.from_file(path, warn_on_failure=True) if not cmap: continue if i == 0 and in pcolors.CMAPS_CYCLIC: cmap.set_cyclic(True) pcolors._cmap_database[] = cmap
[docs]@docstring._snippet_manager def register_cycles(*args, user=None, default=False): """ Register named color cycles. This is called on import. Parameters ---------- %(rc.cycle_args)s %(rc.cycle_exts)s %(rc.cycle_params)s See also -------- register_cmaps register_colors register_fonts proplot.demos.show_cycles """ # Register input color cycles from . import colors as pcolors user = _not_none(user, not bool(args)) # skip user folder if input args passed paths = [] for arg in args: if isinstance(arg, mcolors.Colormap): pcolors._cmap_database[] = arg else: paths.append(arg) # Register data files for _, path in _iter_data_objects('cycles', *paths, user=user, default=default): cmap = pcolors.DiscreteColormap.from_file(path, warn_on_failure=True) if not cmap: continue pcolors._cmap_database[] = cmap
[docs]@docstring._snippet_manager def register_colors(*args, user=None, default=False, space=None, margin=None, **kwargs): """ Register named colors. This is called on import. Parameters ---------- %(rc.color_args)s %(rc.color_params)s space : {'hcl', 'hsl', 'hpl'}, optional The colorspace used to select "perceptually distinct" colors from the XKCD `color survey <>`__. Default is ``'hcl'``. If passed then `default` is set to ``True``. margin : float, optional The margin by which the normalized hue, saturation, and luminance of each XKCD color must differ from the channel values of the other XKCD colors to be deemed "perceptually distinct" and registered. Must fall between ``0`` and ``1`` (``0`` will register all colors). Default is ``0.1``. If passed then `default` is set to ``True``. **kwargs Additional color name specifications passed as keyword arguments rather than positional dictionaries. See also -------- register_cmaps register_cycles register_fonts proplot.demos.show_colors """ from . import colors as pcolors default = default or space is not None or margin is not None margin = _not_none(margin, 0.1) space = _not_none(space, 'hcl') # Remove previously registered colors # NOTE: Try not to touch matplotlib colors for compatibility srcs = {'xkcd': pcolors.COLORS_XKCD, 'opencolor': pcolors.COLORS_OPEN} if default: # possibly slow but not these dicts are empty on startup for src in srcs.values(): for key in src: if key not in COLORS_KEEP: pcolors._color_database.pop(key, None) # this also clears cache src.clear() # Register input colors user = _not_none(user, not bool(args) and not bool(kwargs)) # skip if args passed paths = [] for arg in args: if isinstance(arg, dict): kwargs.update(arg) else: paths.append(arg) for key, color in kwargs.items(): if mcolors.is_color_like(color): pcolors._color_database[key] = mcolors.to_rgba(color) else: raise ValueError(f'Invalid color {key}={color!r}.') # Load colors from file and get their HCL values # NOTE: Colors that come *later* overwrite colors that come earlier. for i, path in _iter_data_objects('colors', *paths, user=user, default=default): loaded = pcolors._load_colors(path, warn_on_failure=True) if i == 0: cat, _ = os.path.splitext(os.path.basename(path)) if cat not in srcs: raise RuntimeError(f'Unknown proplot color database {path!r}.') src = srcs[cat] if cat == 'xkcd': for key in COLORS_KEEP: loaded[key] = pcolors._color_database[key] # keep the same loaded = pcolors._standardize_colors(loaded, space, margin) src.clear() src.update(loaded) # needed for demos.show_colors() pcolors._color_database.update(loaded)
[docs]@docstring._snippet_manager def register_fonts(*args, user=True, default=False): """ Register font names. Parameters ---------- %(rc.font_args)s %(rc.font_params)s See also -------- register_cmaps register_cycles register_colors proplot.demos.show_fonts """ # Find proplot fonts # WARNING: Must search data files in reverse because font manager will # not overwrite existing fonts with user-input fonts. # WARNING: If you include a font file with an unrecognized style, # matplotlib may use that font instead of the 'normal' one! Valid styles: # 'ultralight', 'light', 'normal', 'regular', 'book', 'medium', 'roman', # 'semibold', 'demibold', 'demi', 'bold', 'heavy', 'extra bold', 'black' # # For macOS the only fonts with 'Thin' in one of the .ttf file names # are Helvetica Neue and .SF NS Display Condensed. Never try to use these! paths_proplot = _get_data_folders('fonts', user=user, default=default, reverse=True) fnames_proplot = set(mfonts.findSystemFonts(paths_proplot)) for path in args: path = os.path.expanduser(path) if os.path.exists(path): fnames_proplot.add(path) else: raise FileNotFoundError(f'Invalid font file path {path!r}.') # Detect user-input ttc fonts and issue warning fnames_proplot_ttc = { file for file in fnames_proplot if os.path.splitext(file)[1] == '.ttc' } if fnames_proplot_ttc: warnings._warn_proplot( 'Ignoring the following .ttc fonts because they cannot be ' 'saved into PDF or EPS files (see matplotlib issue #3135): ' + ', '.join(map(repr, sorted(fnames_proplot_ttc))) + '. Please consider expanding them into separate .ttf files.' ) # Rebuild font cache only if necessary! Can be >50% of total import time! fnames_all = {font.fname for font in mfonts.fontManager.ttflist} fnames_proplot -= fnames_proplot_ttc if not fnames_all >= fnames_proplot: warnings._warn_proplot( 'Rebuilding font cache. This usually happens ' 'after installing or updating proplot.' ) if hasattr(mfonts.fontManager, 'addfont'): # Newer API lets us add font files manually and deprecates TTFPATH. However # to cache fonts added this way, we must call json_dump explicitly. # NOTE: Previously, cache filename was specified as _fmcache variable, but # recently became inaccessible. Must reproduce mpl code instead. # NOTE: Older mpl versions used fontList.json as the cache, but these # versions also did not have 'addfont', so makes no difference. for fname in fnames_proplot: mfonts.fontManager.addfont(fname) cache = os.path.join( mpl.get_cachedir(), f'fontlist-v{mfonts.FontManager.__version__}.json' ) mfonts.json_dump(mfonts.fontManager, cache) else: # Older API requires us to modify TTFPATH # NOTE: Previously we tried to modify TTFPATH before importing # font manager with hope that it would load proplot fonts on # initialization. But 99% of the time font manager just imports # the FontManager from cache, so we would have to rebuild anyway. paths = ':'.join(paths_proplot) if 'TTFPATH' not in os.environ: os.environ['TTFPATH'] = paths elif paths not in os.environ['TTFPATH']: os.environ['TTFPATH'] += ':' + paths mfonts._rebuild() # Remove ttc files and 'Thin' fonts *after* rebuild # NOTE: 'Thin' filter is ugly kludge but without this matplotlib picks up on # Roboto thin ttf files installed on the RTD server when compiling docs. mfonts.fontManager.ttflist = [ font for font in mfonts.fontManager.ttflist if os.path.splitext(font.fname)[1] != '.ttc' or 'Thin' in os.path.basename(font.fname) ]
[docs]class Configurator(MutableMapping, dict): """ A dictionary-like class for managing `matplotlib settings <>`__ stored in `rc_matplotlib` and :ref:`proplot settings <ug_rcproplot>` stored in `rc_proplot`. This class is instantiated as the `rc` object on import. See the :ref:`user guide <ug_config>` for details. """ def __repr__(self): cls = type('rc', (dict,), {}) # temporary class with short name src = cls({key: val for key, val in rc_proplot.items() if '.' not in key}) return type(rc_matplotlib).__repr__(src).strip()[:-1] + ',\n ...\n })' def __str__(self): cls = type('rc', (dict,), {}) # temporary class with short name src = cls({key: val for key, val in rc_proplot.items() if '.' not in key}) return type(rc_matplotlib).__str__(src) + '\n...' def __iter__(self): yield from rc_proplot # sorted proplot settings, ignoring deprecations yield from rc_matplotlib # sorted matplotlib settings, ignoring deprecations def __len__(self): return len(tuple(iter(self))) def __delitem__(self, key): # noqa: U100 raise RuntimeError('rc settings cannot be deleted.') def __delattr__(self, attr): # noqa: U100 raise RuntimeError('rc settings cannot be deleted.') @docstring._snippet_manager def __init__(self, local=True, user=True, default=True, **kwargs): """ Parameters ---------- %(rc.params)s """ self._context = [] self._init(local=local, user=user, default=default, **kwargs)
[docs] def __getitem__(self, key): """ Return an `rc_matplotlib` or `rc_proplot` setting. """ key = self._validate_key(key) # might issue proplot removed/renamed error try: return rc_proplot[key] except KeyError: pass return rc_matplotlib[key] # might issue matplotlib removed/renamed error
[docs] def __setitem__(self, key, value): """ Modify an `rc_matplotlib` or `rc_proplot` setting. """ kw_proplot, kw_matplotlib = self._get_params(key, value) rc_proplot.update(kw_proplot) rc_matplotlib.update(kw_matplotlib)
[docs] def __getattr__(self, attr): """ Return an `rc_matplotlib` or `rc_proplot` setting. """ if attr[:1] == '_': return super().__getattribute__(attr) # raise built-in error else: return self.__getitem__(attr)
[docs] def __setattr__(self, attr, value): """ Modify an `rc_matplotlib` or `rc_proplot` setting. """ if attr[:1] == '_': super().__setattr__(attr, value) else: self.__setitem__(attr, value)
def __enter__(self): """ Apply settings from the most recent context block. """ if not self._context: raise RuntimeError( 'rc object must be initialized for context block using rc.context().' ) context = self._context[-1] kwargs = context.kwargs rc_new = context.rc_new # used for context-based _get_with_context rc_old = context.rc_old # used to re-apply settings without copying whole dict for key, value in kwargs.items(): kw_proplot, kw_matplotlib = self._get_params(key, value) for rc_dict, kw_new in zip( (rc_proplot, rc_matplotlib), (kw_proplot, kw_matplotlib), ): for key, value in kw_new.items(): rc_old[key] = rc_dict[key] rc_new[key] = rc_dict[key] = value def __exit__(self, *args): # noqa: U100 """ Restore settings from the most recent context block. """ if not self._context: raise RuntimeError( 'rc object must be initialized for context block using rc.context().' ) context = self._context[-1] for key, value in context.rc_old.items(): kw_proplot, kw_matplotlib = self._get_params(key, value) rc_proplot.update(kw_proplot) rc_matplotlib.update(kw_matplotlib) del self._context[-1] @staticmethod def _validate_key(key): """ Ensure string and convert keys with omitted dots. """ if not isinstance(key, str): raise KeyError(f'Invalid key {key!r}. Must be string.') if '.' not in key: key = rcsetup._rc_nodots.get(key, key) key = rc_proplot._check_key(key) # may issue deprecation warning return key.lower() @staticmethod def _validate_value(key, value): """ Convert numpy ndarray to list and validate if possible. """ # NOTE: Ideally would implicitly validate on subsequent assignment to rc # dictionaries, but must explicitly do it here, so _get_params can 'sync' # settings with children (e.g. multiplying tick widths by width ratios) # and so _load_file() can catch errors and emit warnings before assignment. if isinstance(value, np.ndarray): value = value.item() if value.size == 1 else value.tolist() validate = getattr(rc_matplotlib, 'validate', None) validate_proplot = rc_proplot._validate if validate is not None and key in validate: # guard against API change value = validate[key](value) elif key in validate_proplot: value = validate_proplot[key](value) return value def _init(self, *, local, user, default, skip_cycle=False): """ Initialize the configurator. """ # Always remove context objects self._context.clear() # Update from default settings # NOTE: see _remove_blacklisted_style_params bugfix if default: rc_matplotlib.update(_get_style_dicts('original', filter=False)) rc_matplotlib.update(rcsetup._rc_matplotlib_default) rc_proplot.update(rcsetup._rc_proplot_default) for key, value in rc_proplot.items(): kw_proplot, kw_matplotlib = self._get_params( key, value, skip_cycle=skip_cycle ) rc_matplotlib.update(kw_matplotlib) rc_proplot.update(kw_proplot) # Update from user home user_path = None if user: user_path = self.user_file() if os.path.isfile(user_path): self.load(user_path) # Update from local paths if local: local_paths = self.local_files() for path in local_paths: if path == user_path: # local files always have precedence continue self.load(path) def _get_params(self, key, value, skip_cycle=False): """ Return dictionaries for updating the `rc_proplot` and `rc_matplotlib` properties associated with this key. Used when setting items, entering context blocks, or loading files. """ # Get validated key, value, and child keys key = self._validate_key(key) value = self._validate_value(key, value) keys = (key,) + rcsetup._rc_children.get(key, ()) # settings to change contains = lambda *args: any(arg in keys for arg in args) # noqa: E731 # Fill dictionaries of matplotlib and proplot settings # NOTE: Raise key error right away so it can be caught by _load_file(). # Also ignore deprecation warnings so we only get them *once* on assignment kw_proplot = {} # custom properties kw_matplotlib = {} # builtin properties with warnings.catch_warnings(): warnings.simplefilter('ignore', mpl.MatplotlibDeprecationWarning) warnings.simplefilter('ignore', warnings.ProplotWarning) for key in keys: if key in rc_matplotlib: kw_matplotlib[key] = value elif key in rc_proplot: kw_proplot[key] = value else: raise KeyError(f'Invalid rc setting {key!r}.') # Special key: configure inline backend if contains('inlinefmt'): config_inline_backend(value) # Special key: apply stylesheet elif contains('style'): if value is not None: ikw_matplotlib, ikw_proplot = _get_style_dicts(value, infer=True) kw_matplotlib.update(ikw_matplotlib) kw_proplot.update(ikw_proplot) # Cycler # NOTE: Have to skip this step during initial proplot import elif contains('cycle') and not skip_cycle: from .colors import _get_cmap_subtype cmap = _get_cmap_subtype(value, 'discrete') kw_matplotlib['axes.prop_cycle'] = cycler.cycler('color', cmap.colors) kw_matplotlib['patch.facecolor'] = 'C0' # Turning bounding box on should turn border off and vice versa elif contains('abc.bbox', 'title.bbox', 'abc.border', 'title.border'): if value: name, this = key.split('.') other = 'border' if this == 'bbox' else 'bbox' kw_proplot[name + '.' + other] = False # Fontsize # NOTE: Re-application of e.g. size='small' uses the updated 'font.size' elif contains('font.size'): kw_proplot.update( { key: value for key, value in rc_proplot.items() if key in rcsetup.FONT_KEYS and value in mfonts.font_scalings } ) kw_matplotlib.update( { key: value for key, value in rc_matplotlib.items() if key in rcsetup.FONT_KEYS and value in mfonts.font_scalings } ) # Tick length/major-minor tick length ratio elif contains('tick.len', 'tick.lenratio'): if contains('tick.len'): ticklen = value ratio = rc_proplot['tick.lenratio'] else: ticklen = rc_proplot['tick.len'] ratio = value kw_matplotlib['xtick.minor.size'] = ticklen * ratio kw_matplotlib['ytick.minor.size'] = ticklen * ratio # Spine width/major-minor tick width ratio elif contains('tick.width', 'tick.widthratio'): if contains('tick.width'): tickwidth = value ratio = rc_proplot['tick.widthratio'] else: tickwidth = rc_proplot['tick.width'] ratio = value kw_matplotlib['xtick.minor.width'] = tickwidth * ratio kw_matplotlib['ytick.minor.width'] = tickwidth * ratio # Gridline width elif contains('grid.width', 'grid.widthratio'): if contains('grid.width'): gridwidth = value ratio = rc_proplot['grid.widthratio'] else: gridwidth = rc_proplot['grid.width'] ratio = value kw_proplot['gridminor.linewidth'] = gridwidth * ratio kw_proplot['gridminor.width'] = gridwidth * ratio # Gridline toggling elif contains('grid', 'gridminor'): b, which = _translate_grid( value, 'gridminor' if contains('gridminor') else 'grid' ) kw_matplotlib['axes.grid'] = b kw_matplotlib['axes.grid.which'] = which return kw_proplot, kw_matplotlib def _get_with_context(self, key, mode=None): """ As with `~Configurator.__getitem__` but the search is limited based on the context mode and ``None`` is returned if the key is not found. """ key = self._validate_key(key) if mode is None: mode = self._context_mode cache = tuple(context.rc_new for context in self._context) if mode == 0: rcdicts = (*cache, rc_proplot, rc_matplotlib) elif mode == 1: rcdicts = (*cache, rc_proplot) # added settings only! elif mode == 2: rcdicts = (*cache,) else: raise ValueError(f'Invalid caching mode {mode!r}.') for rcdict in rcdicts: if not rcdict: continue try: return rcdict[key] except KeyError: continue if mode == 0: raise KeyError(f'Invalid rc setting {key!r}.')
[docs] @staticmethod def local_files(): """ Return locations of local proplotrc files in this directory and in parent directories. See also -------- Configurator.user_file Configurator.user_folder """ cdir = os.getcwd() paths = [] # Loop until we reach root while cdir: # Look for hidden and unhidden proplotrc files for name in ('proplotrc', '.proplotrc'): path = os.path.join(cdir, name) if os.path.exists(path): paths.append(path) # Move on to next parent directory ndir = os.path.dirname(cdir) if ndir == cdir: # root break cdir = ndir return paths[::-1] # sort from decreasing to increasing importantce
@staticmethod def _config_folder(): """ Get the XDG proplot folder. """ home = os.path.expanduser('~') base = os.environ.get('XDG_CONFIG_HOME') if not base: base = os.path.join(home, '.config') if sys.platform.startswith(('linux', 'freebsd')) and os.path.exists(base): configdir = os.path.join(base, 'proplot') else: configdir = os.path.join(home, '.proplot') return configdir
[docs] @staticmethod def user_file(): """ Return location of the default proplotrc file. On Linux, this is either ``$XDG_CONFIG_HOME/proplot/proplotrc`` or ``~/.config/proplot/proplotrc`` if the `XDG directory <>`__ is unset. On other operating systems, this is ``~/.proplot/proplotrc``. The location ``~/.proplotrc`` or ``~/.proplot/proplotrc`` is always returned if the file exists, regardless of the operating system. If multiple valid locations are found, a warning is raised. See also -------- Configurator.user_folder Configurator.local_files """ # Support both loose files and files inside .proplot file = os.path.join(Configurator.user_folder(), 'proplotrc') universal = os.path.join(os.path.expanduser('~'), '.proplotrc') if os.path.isfile(universal): if file != universal and os.path.isfile(file): warnings._warn_proplot( 'Found conflicting default user proplotrc files at ' f'{universal!r} and {file!r}. Ignoring the second one.' ) file = universal return file
[docs] @staticmethod def user_folder(subfolder=None): """ Return location of the default proplot folder. On Linux, this is either ``$XDG_CONFIG_HOME/proplot`` or ``~/.config/proplot`` if the `XDG directory <>`__ is unset. On other operating systems, this is ``~/.proplot``. The location ``~/.proplot`` is always returned if the folder exists, regardless of the operating system. If multiple valid locations are found, a warning is raised. See also -------- Configurator.user_file Configurator.local_files """ # Try the XDG standard location # NOTE: This is borrowed from matplotlib.get_configdir home = os.path.expanduser('~') universal = folder = os.path.join(home, '.proplot') if sys.platform.startswith(('linux', 'freebsd')): xdg = os.environ.get('XDG_CONFIG_HOME') xdg = xdg or os.path.join(home, '.config') folder = os.path.join(xdg, 'proplot') # Fallback to the loose ~/.proplot if it is present # NOTE: This is critical or we might ignore previously stored settings! if os.path.isdir(universal): if folder != universal and os.path.isdir(folder): warnings._warn_proplot( 'Found conflicting default user proplot folders at ' f'{universal!r} and {folder!r}. Ignoring the second one.' ) folder = universal # Return the folder if subfolder: folder = os.path.join(folder, subfolder) return folder
[docs] def context(self, *args, mode=0, file=None, **kwargs): """ Temporarily modify the rc settings in a "with as" block. Parameters ---------- *args Dictionaries of `rc` keys and values. file : path-like, optional Filename from which settings should be loaded. **kwargs `rc` names and values passed as keyword arguments. If the name has dots, simply omit them. Other parameters ---------------- mode : {0, 1, 2}, optional The context mode. Dictates the behavior of `~Configurator.find`, `~Configurator.fill`, and `~Configurator.category` within a "with as" block when called with ``context=True``. The options are as follows: 0. Matplotlib's `rc_matplotlib` settings and proplots `rc_proplot` settings are all returned, whether or not `~Configurator.context` has changed them. 1. Unchanged `rc_matplotlib` settings return ``None`` but `rc_proplot` settings are returned whether or not `~Configurator.context` has changed them. This is used in the `~proplot.axes.Axes.__init__` call to `~proplot.axes.Axes.format`. When a lookup returns ``None``, `~proplot.axes.Axes.format` does not apply it. 2. All unchanged settings return ``None``. This is used during user calls to `~proplot.axes.Axes.format`. Note ---- This is used by proplot internally but may also be useful for power users. It was invented to prevent successive calls to `~proplot.axes.Axes.format` from constantly looking up and re-applying unchanged settings. These gratuitous lookups increased runtime significantly, and resulted in successive calls to `~proplot.axes.Axes.format` overwriting the previous calls. Example ------- The below applies settings to axes in a specific figure using `~Configurator.context`. >>> import proplot as pplt >>> with pplt.rc.context(ticklen=5, metalinewidth=2): >>> fig, ax = pplt.subplots() >>> ax.plot(data) The below applies settings to a specific axes using `~proplot.axes.Axes.format`, which uses `~Configurator.context` internally. >>> import proplot as pplt >>> fig, ax = pplt.subplots() >>> ax.format(ticklen=5, metalinewidth=2) """ # Add input dictionaries for arg in args: if not isinstance(arg, dict): raise ValueError('Non-dictionary argument {arg!r}.') kwargs.update(arg) # Add settings from file if file is not None: kw_proplot, kw_matplotlib = self._load_file(file) kwargs.update(kw_proplot) kwargs.update(kw_matplotlib) # Activate context object if mode not in range(3): raise ValueError(f'Invalid mode {mode!r}.') cls = namedtuple('RcContext', ('mode', 'kwargs', 'rc_new', 'rc_old')) context = cls(mode=mode, kwargs=kwargs, rc_new={}, rc_old={}) self._context.append(context) return self
[docs] def category(self, cat, *, trimcat=True, context=False): """ Return a dictionary of settings beginning with the substring ``cat + '.'``. Optionally limit the search to the context level. Parameters ---------- cat : str, optional The `rc` setting category. trimcat : bool, optional Whether to trim ``cat`` from the key names in the output dictionary. Default is ``True``. context : bool, optional If ``True``, then each category setting that is not found in the context mode dictionaries is omitted from the output dictionary. See `~Configurator.context`. See also -------- Configurator.find Configurator.fill """ kw = {} if cat not in rcsetup._rc_categories: raise ValueError( f'Invalid rc category {cat!r}. Valid categories are: ' + ', '.join(map(repr, rcsetup._rc_categories)) + '.' ) for key in self: if not re.match(fr'\A{cat}\.[^.]+\Z', key): continue value = self._get_with_context(key, None if context else 0) if value is None: continue if trimcat: key = re.sub(fr'\A{cat}\.', '', key) kw[key] = value return kw
[docs] def fill(self, props, *, context=False): """ Return a dictionary filled with settings whose names match the string values in the input dictionary. Optionally limit the search to the context level. Parameters ---------- props : dict-like Dictionary whose values are setting names. context : bool, optional If ``True``, then settings not found in the context mode dictionaries are excluded from the output dictionary. See `~Configurator.context`. See also -------- Configurator.category Configurator.find """ kw = {} for key, value in props.items(): item = self._get_with_context(value, None if context else 0) if item is not None: kw[key] = item return kw
[docs] def find(self, key, *, context=False): """ Return a single setting. Optionally limit the search to the context level. Parameters ---------- key : str The single setting name. context : bool, optional If ``True``, then ``None`` is returned if the setting is not found in the context mode dictionaries. See `~Configurator.context`. See also -------- Configurator.category Configurator.fill """ return self._get_with_context(key, None if context else 0)
[docs] def update(self, *args, **kwargs): """ Update several settings at once. Parameters ---------- *args : str or dict-like, optional A dictionary containing `rc` keys and values. You can also pass a "category" name as the first argument, in which case all settings are prepended with ``'category.'``. For example, ``rc.update('axes', labelsize=20, titlesize=20)`` changes the :rcraw:`axes.labelsize` and :rcraw:`axes.titlesize` settings. **kwargs, optional `rc` keys and values passed as keyword arguments. If the name has dots, simply omit them. See also -------- Configurator.category Configurator.fill """ prefix, kw = '', {} if not args: pass elif len(args) == 1 and isinstance(args[0], str): prefix = args[0] elif len(args) == 1 and isinstance(args[0], dict): kw = args[0] elif len(args) == 2 and isinstance(args[0], str) and isinstance(args[1], dict): prefix, kw = args else: raise ValueError( f'Invalid arguments {args!r}. Usage is either ' 'rc.update(dict), rc.update(kwy=value, ...), ' 'rc.update(category, dict), or rc.update(category, key=value, ...).' ) prefix = prefix and prefix + '.' kw.update(kwargs) for key, value in kw.items(): self.__setitem__(prefix + key, value)
[docs] @docstring._snippet_manager def reset(self, local=True, user=True, default=True, **kwargs): """ Reset the configurator to its initial state. Parameters ---------- %(rc.params)s """ self._init(local=local, user=user, default=default, **kwargs)
def _load_file(self, path): """ Return dictionaries of proplot and matplotlib settings loaded from the file. """ added = set() path = os.path.expanduser(path) kw_proplot = {} kw_matplotlib = {} with open(path, 'r') as fh: for idx, line in enumerate(fh): # Strip comments message = f'line #{idx + 1} in file {path!r}' stripped = line.split('#', 1)[0].strip() if not stripped: pass # no warning continue # Parse the pair pair = stripped.split(':', 1) if len(pair) != 2: warnings._warn_proplot(f'Illegal {message}:\n{line}"') continue # Detect duplicates key, val = map(str.strip, pair) if key in added: warnings._warn_proplot(f'Duplicate rc key {key!r} on {message}.') added.add(key) # Get child dictionaries. Careful to have informative messages with warnings.catch_warnings(): warnings.simplefilter('error', warnings.ProplotWarning) try: ikw_proplot, ikw_matplotlib = self._get_params(key, val) except KeyError: warnings._warn_proplot( f'Invalid rc key {key!r} on {message}.', 'default' ) continue except ValueError as err: warnings._warn_proplot( f'Invalid rc val {val!r} for key {key!r} on {message}: {err}', 'default' # noqa: E501 ) continue except warnings.ProplotWarning as err: warnings._warn_proplot( f'Outdated rc key {key!r} on {message}: {err}', 'default' ) warnings.simplefilter('ignore', warnings.ProplotWarning) ikw_proplot, ikw_matplotlib = self._get_params(key, val) # Update the settings kw_proplot.update(ikw_proplot) kw_matplotlib.update(ikw_matplotlib) return kw_proplot, kw_matplotlib
[docs] def load(self, path): """ Load settings from the specified file. Parameters ---------- path : path-like The file path. See also -------- """ kw_proplot, kw_matplotlib = self._load_file(path) rc_proplot.update(kw_proplot) rc_matplotlib.update(kw_matplotlib)
@staticmethod def _save_rst(path): """ Create an RST table file. Used for online docs. """ string = rcsetup._rst_table() with open(path, 'w') as fh: fh.write(string) @staticmethod def _save_yaml(path, user_dict=None, *, comment=False, description=False): """ Create a YAML file. Used for online docs and default and user-generated proplotrc files. Extra settings can be passed with the input dictionary. """ user_table = () if user_dict: # add always-uncommented user settings user_table = rcsetup._yaml_table(user_dict, comment=False) user_table = ('# Changed settings', user_table, '') proplot_dict = rcsetup._rc_proplot_table if description else rcsetup._rc_proplot_default # noqa: E501 proplot_table = rcsetup._yaml_table(proplot_dict, comment=comment, description=description) # noqa: E501 proplot_table = ('# Proplot settings', proplot_table, '') matplotlib_dict = rcsetup._rc_matplotlib_default matplotlib_table = rcsetup._yaml_table(matplotlib_dict, comment=comment) matplotlib_table = ('# Matplotlib settings', matplotlib_table) parts = ( '#--------------------------------------------------------------------', '# Use this file to change the default proplot and matplotlib settings.', '# The syntax is identical to matplotlibrc syntax. For details see:', '#', '#', '#--------------------------------------------------------------------', *user_table, # empty if nothing passed *proplot_table, *matplotlib_table, ) with open(path, 'w') as fh: fh.write('\n'.join(parts))
[docs] def save(self, path=None, user=True, comment=None, backup=True, description=False): """ Save the current settings to a ``proplotrc`` file. This writes the default values commented out plus the values that *differ* from the defaults at the top of the file. Parameters ---------- path : path-like, optional The path. The default file name is ``proplotrc`` and the default directory is the current directory. user : bool, optional If ``True`` (the default), the settings you changed since importing proplot are shown uncommented at the very top of the file. backup : bool, optional If the file already exists and this is set to ``True``, it is moved to a backup file with the suffix ``.bak``. comment : bool, optional Whether to comment out the default settings. Default is the value of `user`. description : bool, optional Whether to include descriptions of each setting as comments. Default is ``False``. See also -------- Configurator.load """ path = os.path.expanduser(path or '.') if os.path.isdir(path): # includes '' path = os.path.join(path, 'proplotrc') if os.path.isfile(path) and backup: backup = path + '.bak' os.rename(path, backup) warnings._warn_proplot(f'Existing file {path!r} was moved to {backup!r}.') comment = _not_none(comment, user) user_dict = self.changed if user else None self._save_yaml(path, user_dict, comment=comment, description=description)
@property def _context_mode(self): """ Return the highest (least permissive) context mode. """ return max((context.mode for context in self._context), default=0) @property def changed(self): """ A dictionary of settings that have been changed from the proplot defaults. """ # Carefully detect changed settings rcdict = {} for key, value in self.items(): default = rcsetup._get_default_param(key) if isinstance(value, Real) and isinstance(default, Real) and np.isclose(value, default): # noqa: E501 pass elif value == default: pass else: rcdict[key] = value # Ignore non-style-related settings. See mstyle.STYLE_BLACKLIST # TODO: For now not sure how to detect if prop cycle changed since # we cannot load it from _cmap_database in rcsetup. rcdict.pop('interactive', None) # changed by backend rcdict.pop('axes.prop_cycle', None) return _get_filtered_dict(rcdict, warn=False) # Renamed methods load_file = warnings._rename_objs('0.8', load_file=load)
# Initialize locations _init_user_folders() _init_user_file() #: A dictionary-like container of matplotlib settings. Assignments are #: validated and restricted to recognized setting names. rc_matplotlib = mpl.rcParams # PEP8 4 lyfe #: A dictionary-like container of proplot settings. Assignments are #: validated and restricted to recognized setting names. rc_proplot = rcsetup._rc_proplot_default.copy() # a validated rcParams-style dict #: Instance of `Configurator`. This controls both `rc_matplotlib` and `rc_proplot` #: settings. See the :ref:`configuration guide <ug_config>` for details. rc = Configurator(skip_cycle=True) # Deprecated RcConfigurator = warnings._rename_objs( '0.8', RcConfigurator=Configurator, ) inline_backend_fmt = warnings._rename_objs( '0.6', inline_backend_fmt=config_inline_backend )