π Geometric Computer Vision Library for Spatial AI
Python artificial-intelligencecomputer-visiondeep-learninghacktoberfestimage-processingmachine-learningneural-networkpythonpytorchroboticsspatial-ai
# LICENSE HEADER MANAGED BY add-license-header
#
# Copyright 2018 Kornia Team
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
import builtins
import html
import importlib.util
import inspect
import json
import os
import re
import sys
from datetime import UTC, datetime
# Monkey-patch for PyTorch compatibility with sphinx_autodoc_typehints
# Newer versions of PyTorch removed torch.jit.annotations.compiler_flag
import torch.jit.annotations
from sphinx.util import logging as sphinx_logging
if not hasattr(torch.jit.annotations, "compiler_flag"):
torch.jit.annotations.compiler_flag = None
# Let the library know it is being imported by the Sphinx build.
builtins.__sphinx_build__ = True
logger = sphinx_logging.getLogger(__name__)
# --- Patch sphinx_autodoc_defaultargs to not crash on torchscript/pybind11 callables ---
try:
import sphinx_autodoc_defaultargs
_orig_process_docstring = sphinx_autodoc_defaultargs.process_docstring
def _safe_process_docstring(app, what, name, obj, options, lines):
try:
return _orig_process_docstring(app, what, name, obj, options, lines)
except ValueError as e:
msg = str(e).lower()
if "no signature found for builtin" in msg or "pybind11" in msg:
return None # leave docstring unchanged
raise
sphinx_autodoc_defaultargs.process_docstring = _safe_process_docstring
except (ModuleNotFoundError, ImportError):
# Optional dependency not installed in some environments.
sphinx_autodoc_defaultargs = None
except AttributeError:
# Extension API changed; don't patch.
sphinx_autodoc_defaultargs = None
# ``generate_examples.main`` builds four pretrained models (KeyNet, DISK, ALIKED
# and XFeat), so this build fetches checkpoints from the same rate-limited hosts
# the test jobs do. CI restores the shared ``weights/`` cache for this job, but
# nothing here is running under ``conftest.py``, which is what points torch at
# it for the test and doctest runs -- so point at it here too, or the cache is
# restored and then ignored. Sphinx executes this file with the working
# directory set to its own folder, hence the path off ``__file__``.
import torch.hub # noqa: E402
_weights_cache = os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "..", "weights")
# Only when there is a restored cache to use and the developer has not pointed
# torch somewhere themselves: a local ``pixi run -e default build-docs`` in a fresh clone has
# neither, and redirecting it would re-download all four checkpoints past a warm
# ``~/.cache/torch/hub``.
if os.path.isdir(_weights_cache) and not os.environ.get("TORCH_HOME"):
torch.hub.set_dir(_weights_cache)
# readthedocs generated the whole documentation in an isolated environment
# by cloning the git repo. Thus, any on-the-fly operation will not effect
# on the resulting documentation. We therefore need to import and run the
# corresponding code here.
spec = importlib.util.spec_from_file_location("generate_examples", "../generate_examples.py")
generate_examples = importlib.util.module_from_spec(spec)
spec.loader.exec_module(generate_examples)
# Pre-generate the example images
generate_examples.main()
spec = importlib.util.spec_from_file_location("generate_benchmarks", "../generate_benchmarks.py")
generate_benchmarks = importlib.util.module_from_spec(spec)
spec.loader.exec_module(generate_benchmarks)
# Pre-generate the benchmark results page
generate_benchmarks.main()
spec = importlib.util.spec_from_file_location("generate_adoption", "../generate_adoption.py")
generate_adoption = importlib.util.module_from_spec(spec)
spec.loader.exec_module(generate_adoption)
# Pre-generate the adoption page from the committed dependents snapshot
_adoption = generate_adoption.main()
spec = importlib.util.spec_from_file_location("generate_export_support", "../generate_export_support.py")
generate_export_support = importlib.util.module_from_spec(spec)
spec.loader.exec_module(generate_export_support)
# Pre-generate the ONNX / torch.compile / torch.export support page: the promise from testing/conformance_promise.py
# and the conformance manifests, the survey from the committed snapshot
generate_export_support.main()
def _public_operator_count() -> int:
"""Public functions and ``nn.Module`` classes defined across kornia's public modules.
Walks every ``kornia.*`` module without a private path component and counts the functions and
``torch.nn.Module`` subclasses it defines under a public name. Model architecture internals
(``kornia.models.*.architecture``) and ``kornia.core`` are left out so the
figure is about operators rather than plumbing.
"""
import contextlib
import io
import pkgutil
import warnings
import torch
import kornia
skip = re.compile(r"^kornia\.core(\.|$)|\.architecture(\.|$)")
seen: set[tuple[str, str]] = set()
with (
warnings.catch_warnings(),
contextlib.redirect_stdout(io.StringIO()),
contextlib.redirect_stderr(io.StringIO()),
):
warnings.simplefilter("ignore")
for info in pkgutil.walk_packages(kornia.__path__, "kornia.", onerror=lambda name: None):
if skip.search(info.name) or any(part.startswith("_") for part in info.name.split(".")[1:]):
continue
try:
module = importlib.import_module(info.name)
except Exception as exc: # noqa: BLE001 - an optional dependency is missing; nothing to count there
logger.debug("skipping %s while counting operators: %r", info.name, exc)
continue
for name, obj in vars(module).items():
# Type checks first: attribute access on a ``kornia.core.external.LazyLoader`` would
# try to import the optional dependency behind it (raising, or installing it in the "ask" and "auto"
# installation modes).
if not (inspect.isfunction(obj) or (inspect.isclass(obj) and issubclass(obj, torch.nn.Module))):
continue
if name.startswith("_") or obj.__name__ != name:
continue
defined_in = getattr(obj, "__module__", "") or ""
if defined_in.startswith("kornia.") and not skip.search(defined_in):
seen.add((defined_in, obj.__qualname__))
return len(seen)
def _landing_page_counts() -> dict[str, str]:
"""Figures the landing and ONNX pages quote, derived from the library rather than typed in.
The public-name counts read the stable-core inventory that ``tests/test_api_surface.py`` pins;
``kornia.feature`` and ``kornia.models`` are outside it, so they are counted from the live
package and from the model pages under ``docs/source/models/``. Each count names exactly the
module the card it appears on links to. ``operators-floor`` is the library-wide total from
:func:`_public_operator_count`, floored to the hundred below it so the page can say "1,000+"
and stay true between releases.
"""
with open("../../tests/api_surface.json", encoding="utf-8") as handle:
surface = {name: len(names) for name, names in json.load(handle).items()}
import kornia.feature
operators = _public_operator_count()
model_pages = [entry for entry in os.listdir("models") if entry.endswith(".rst") and entry != "index.rst"]
# Every entry here is quoted by a page; a substitution nothing references is dead weight that
# reads as an available figure, so add one when a page needs it rather than ahead of time.
return {
"geometry": str(surface["kornia.geometry"]),
"feature": str(len(kornia.feature.__all__)),
"filters": str(surface["kornia.filters"]),
"augmentation": str(surface["kornia.augmentation"]),
"models": str(len(model_pages)),
"operators-floor": f"{operators // 100 * 100:,}",
"dependents": f"{_adoption['repositories']:,}",
"dependent-packages": f"{_adoption['packages']:,}",
}
# Substitutions such as |count-geometry| and |count-operators-floor| for the landing page.
rst_epilog = "\n".join(f".. |count-{key}| replace:: {value}" for key, value in _landing_page_counts().items())
# If extensions (or modules to document with autodoc) are in another directory,
# add these directories to sys.path here. If the directory is relative to the
# documentation root, use os.path.abspath to make it absolute, like shown here.
#
current_path = os.path.abspath(os.path.join(__file__, "..", "..", ".."))
sys.path.append(current_path)
# -- General configuration ------------------------------------------------
# If your documentation needs a minimal Sphinx version, state it here.
#
# needs_sphinx = '1.0'
# Add any Sphinx extension module names here, as strings. They can be
# extensions coming with Sphinx (named 'sphinx.ext.*') or your custom
# ones.
extensions = [
"sphinx.ext.autodoc",
"sphinx.ext.autosummary",
"sphinx.ext.doctest",
"sphinx.ext.intersphinx",
"sphinx.ext.mathjax",
"sphinx.ext.napoleon",
"sphinx_autodoc_typehints",
"sphinx_autodoc_defaultargs",
"sphinx_copybutton",
"sphinx.ext.linkcode",
"sphinx.ext.githubpages",
"sphinxcontrib.bibtex",
"sphinxcontrib.gtagjs",
"sphinxcontrib.youtube",
"sphinx_design",
"notfound.extension",
]
# substitutes the default values
docstring_default_arg_substitution = "Default: "
autodoc_preserve_defaults = True
bibtex_bibfiles = ["references.bib"]
napoleon_use_ivar = True
# Add any paths that contain templates here, relative to this directory.
templates_path = ["_templates"]
# The suffix(es) of source filenames.
# You can specify multiple suffix as a list of string:
#
# source_suffix = ['.rst', '.md']
source_suffix = [".rst", ".ipynb"]
# The master toctree document.
master_doc = "index"
# General information about the project.
project = "Kornia"
author = f"{project} developers"
copyright = f"{datetime.now(tz=UTC).year}, {author}"
# The version info for the project you're documenting, acts as replacement for
# |version| and |release|, also used in various other places throughout the
# built documents.
# version = 'master (' + kornia.__version__ + ' )'
version = ""
if "READTHEDOCS" not in os.environ:
# if developing locally, use kornia.__version__ as version
from kornia import __version__
version = __version__
# release = 'master'
release = version
# The language for content autogenerated by Sphinx. Refer to documentation
# for a list of supported languages.
#
# This is also used if you do content translation via gettext catalogs.
# Usually you set "language" from the command line for these cases.
language = "en"
# List of patterns, relative to source directory, that match files and
# directories to ignore when looking for source files.
# This patterns also effect to html_static_path and html_extra_path
exclude_patterns = ["_build", ".ipynb_checkpoints"]
# The name of the Pygments (syntax highlighting) style to use. The dark counterpart is a
# pydata-sphinx-theme option, set in ``_PYDATA_THEME_OPTIONS`` below.
pygments_style = "friendly"
# The documentation builds on pydata-sphinx-theme (top navbar with Learn / API / Models and the
# Ecosystem / About / Support dropdowns).
html_theme = "pydata_sphinx_theme"
# Theme options are theme-specific and customize the look and feel of a theme
# further. For a list of options available for each theme, see the
# documentation.
#
# TODO(jian): make to work with https://docs.kornia.org
html_baseurl = "https://kornia.readthedocs.io/en/latest/"
# Git ref that the "view/edit source" links and the ``linkcode`` extension point at.
rtd_version = os.environ.get("READTHEDOCS_VERSION")
code_ref = rtd_version if rtd_version and rtd_version not in {"latest", "stable"} else "main"
# Changing sidebar title to Kornia
html_title = "Kornia"
_PYDATA_THEME_OPTIONS = {
"logo": {
"image_light": "_static/img/kornia_logo_only_light.svg",
"image_dark": "_static/img/kornia_logo_only_dark.svg",
"text": "Kornia",
},
# Top navbar: one entry per top-level toctree item of index.rst (Guide / API / Models / Community),
# plus the external tutorials site. Each section gets its own left sidebar.
# Navbar: logo + search on the left, flexible space, then the nav links,
# theme switcher and icon links on the right.
"navbar_align": "right",
"navbar_start": ["navbar-logo", "search-button-field"],
"navbar_center": [],
"navbar_end": ["navbar-nav", "theme-switcher", "navbar-icon-links"],
"navbar_persistent": [],
"header_links_before_dropdown": 6,
"icon_links": [
{"name": "GitHub", "url": "https://github.com/kornia/kornia", "icon": "fa-brands fa-github"},
{"name": "Discord", "url": "https://discord.gg/HfnywwpBnD", "icon": "fa-brands fa-discord"},
{"name": "Twitter", "url": "https://twitter.com/kornia_foss", "icon": "fa-brands fa-x-twitter"},
],
"use_edit_page_button": True,
"show_nav_level": 1,
"navigation_depth": 4,
"show_toc_level": 2,
"collapse_navigation": False,
# Right sidebar: page TOC and edit links, with the sponsor box (docs/source/_templates/sponsors.html)
# below. The landing page is a designed hero page and carries no right rail at all.
"secondary_sidebar_items": {
"index": [],
"**": ["page-toc", "edit-this-page", "sourcelink", "sponsors"],
},
"pygments_light_style": "friendly",
"pygments_dark_style": "monokai",
"footer_start": [],
"footer_end": [],
}
html_theme_options = _PYDATA_THEME_OPTIONS
# Navbar dropdown menus, rendered by ``_static/js/custom.js``: "Support" and "About" hang off the
# toctree entries of the same name in index.rst; "Ecosystem" is a grouped panel inserted before
# "About". Entries are ``[label, target]`` where the target is a docname (checked against the build,
# so a renamed page fails ``-W`` instead of 404ing from the menu), an external URL, or ``None`` for
# a "coming soon" placeholder.
NAVBAR_MENUS = {
"Support": [
["Sponsor", "community/sponsor"],
["Contribute", "community/contribute"],
],
"About": [
["FAQ", "community/faqs"],
["Team", "get-started/governance"],
["Community Guide", "community/community"],
["Code of Conduct", "https://github.com/kornia/kornia/blob/main/CODE_OF_CONDUCT.md"],
["Citing Kornia", "get-started/about"],
["Adoption", "community/adoption"],
["API Stability Policy", "get-started/stability"],
],
"Ecosystem": {
"official libs": [
["kornia", "https://github.com/kornia/kornia"],
["kornia-rs", "https://github.com/kornia/kornia-rs"],
],
"help": [
["Discord chat", "https://discord.gg/HfnywwpBnD"],
["GitHub discussions", "https://github.com/kornia/kornia/discussions"],
["Issue tracker", "https://github.com/kornia/kornia/issues"],
],
"news": [
["Twitter / X", "https://twitter.com/kornia_foss"],
["LinkedIn", "https://www.linkedin.com/company/kornia/"],
["Newsletter", None],
],
},
}
# Feeds the "Edit this page" button and the source links.
html_context = {
"github_user": "kornia",
"github_repo": "kornia",
"github_version": code_ref,
"doc_path": "docs/source",
}
# The landing page has its own card grid; a section sidebar next to it would be empty noise.
html_sidebars = {"index": []}
# html_logo = '_static/img/kornia_logo.svg'
# html_logo = '_static/img/kornia_logo_only.png'
html_favicon = "_static/img/kornia_logo_favicon.png"
# Show the build date in the footer so readers can tell a stale mirror from the live docs.
html_last_updated_fmt = "%b %d, %Y"
# sphinx-copybutton: strip ``>>>`` / ``...`` / ``$`` prompts and skip output lines, so the
# clipboard receives runnable code instead of a transcript.
copybutton_prompt_text = r">>> |\.\.\. |\$ |In \[\d*\]: | {2,5}\.\.\.: | {5,8}: "
copybutton_prompt_is_regexp = True
copybutton_only_copy_prompt_lines = True
copybutton_line_continuation_character = "\\"
# Config the `sphinxcontrib.gtagjs` extension
# NOTE: if this didn't work, we can remove the extension itself
gtagjs_ids = [
"G-YSCFZB2WDV", # Shouldn't be necessary if the readthedocs autoinjection work
]
# Add any paths that contain custom static files (such as style sheets) here,
# relative to this directory. They are copied after the builtin static files,
# so a file named "default.css" will overwrite the builtin "default.css".
html_static_path = ["_static"]
html_extra_path = ["_extra"]
# Output file base name for HTML help builder.
htmlhelp_basename = "Kornia"
html_css_files = ["css/pydata.css", "css/blueprint.css", "css/playground.css"]
html_js_files = ["js/custom.js"]
# Configure viewcode extension.
# based on https://github.com/readthedocs/sphinx-autoapi/issues/202
code_url = f"https://github.com/kornia/kornia/blob/{code_ref}"
def linkcode_resolve(domain, info):
if domain != "py":
return None
modname = info.get("module")
fullname = info.get("fullname")
if not modname or not fullname:
return None
try:
mod = importlib.import_module(modname)
except (ImportError, ModuleNotFoundError):
return None
obj = mod
for part in fullname.split("."):
try:
obj = getattr(obj, part)
except AttributeError:
return None
obj = inspect.unwrap(obj)
try:
fn = inspect.getsourcefile(obj)
src, start = inspect.getsourcelines(obj)
except (TypeError, OSError, ValueError):
return None
if not fn:
return None
fn = os.path.abspath(fn).replace("\\", "/")
marker = "/kornia/"
idx = fn.rfind(marker)
if idx == -1:
return None
file_rel = fn[idx + 1 :] # -> "kornia/....py"
end = start + len(src) - 1
return f"{code_url}/{file_rel}#L{start}-L{end}"
# -- Options for LaTeX output ---------------------------------------------
latex_elements = {
# The paper size ('letterpaper' or 'a4paper').
#
# 'papersize': 'letterpaper',
# The font size ('10pt', '11pt' or '12pt').
#
# 'pointsize': '10pt',
# Additional stuff for the LaTeX preamble.
#
# 'preamble': '',
# Latex figure (float) alignment
#
# 'figure_align': 'htbp',
}
# Grouping the document tree into LaTeX files. List of tuples
# (source start file, target name, title,
# author, documentclass [howto, manual, or own class]).
latex_documents = [(master_doc, "kornia.tex", "Kornia", "manual")]
# -- Options for manual page output ---------------------------------------
# One entry per manual page. List of tuples
# (source start file, name, description, authors, manual section).
man_pages = [(master_doc, "Kornia", "Kornia Documentation", [author], 1)]
# -- Options for Texinfo output -------------------------------------------
# Grouping the document tree into Texinfo files. List of tuples
# (source start file, target name, title, author,
# dir menu entry, description, category)
texinfo_documents = [
(
master_doc,
"kornia",
"Kornia Documentation",
author,
"Kornia",
"Differentiable Computer Vision in Pytorch.",
"Miscellaneous",
)
]
# Cross-references into the Python, NumPy and PyTorch docs. Each project is fetched from its live site first; the copy
# in _intersphinx/ is used only when that fetch fails, so an outage of one of those sites does not fail the -W build.
# The links still point at the live site. Refresh a copy with
# `curl -fsSLo docs/source/_intersphinx/<name>.inv <site>objects.inv`.
intersphinx_mapping = {
"python": ("https://docs.python.org/3/", (None, "_intersphinx/python.inv")),
"numpy": ("https://numpy.org/doc/stable/", (None, "_intersphinx/numpy.inv")),
"torch": ("https://pytorch.org/docs/stable/", (None, "_intersphinx/torch.inv")),
}
# Optional third-party modules that autodoc must not try to import. Empty: the packages the documented
# API still reaches lazily (diffusers, onnx, onnxruntime) go through kornia.core.external.LazyLoader,
# which returns a placeholder under Sphinx (see the note there), so nothing needs mocking.
autodoc_mock_imports: list[str] = []
# -- Social / SEO metadata --------------------------------------------------
# Used when a page does not declare its own ``.. meta:: :name: description``.
_DEFAULT_DESCRIPTION = (
"Kornia is a differentiable computer vision library for PyTorch: batched, GPU-ready, "
"autograd-friendly image transforms, filters, color conversions, camera geometry, augmentations "
"and curated deep learning models."
)
_SOCIAL_IMAGE = "https://github.com/kornia/data/raw/main/kornia_banner_pixie.png"
_PAGE_REDIRECTS = {"get-started/highlights": "get-started/introduction"}
_NOINDEX_PAGES = {"404", "genindex", "py-modindex", "search"}
_META_TAG_RE = re.compile(r"<meta\b[^>]*>", re.IGNORECASE)
_META_ATTR_RE = re.compile(r'(\w+)="([^"]*)"')
def _inject_social_metatags(app, pagename, templatename, context, doctree):
"""Add Open Graph / Twitter card tags (and a fallback description) to every HTML page.
Search engines and chat apps render these when a docs link is shared; Sphinx and the theme
emit none of them by default. The description is taken from the page's own ``.. meta::``
directive when it has one, so page authors keep control of the snippet.
"""
metatags = context.get("metatags", "") or ""
match = None
for tag in _META_TAG_RE.findall(metatags):
attrs = dict(_META_ATTR_RE.findall(tag))
if attrs.get("name", "").lower() == "description" and attrs.get("content"):
match = attrs["content"]
break
# Sphinx hands the title over already HTML-escaped and it can contain theme markup.
page_title = re.sub(r"<[^>]+>", "", html.unescape(context.get("title") or project))
page_title = " ".join(page_title.split())
redirect_target = _PAGE_REDIRECTS.get(pagename)
if match and not redirect_target:
description = html.unescape(match).strip().strip('"')
elif context.get("title") and pagename not in _NOINDEX_PAGES and not redirect_target:
description = f"{page_title}: documentation for Kornia, the differentiable computer vision library for PyTorch."
else:
description = _DEFAULT_DESCRIPTION
description = " ".join(description.split())
if len(description) > 300:
description = description[:297].rsplit(" ", 1)[0] + "..."
title = f"{page_title} - {project}"
if pagename == master_doc:
title = f"{project} β Differentiable Computer Vision for PyTorch"
url = context.get("pageurl") or f"{html_baseurl}{app.builder.get_target_uri(pagename)}"
if pagename == master_doc:
url = html_baseurl
context["pageurl"] = url
refresh_url = None
if redirect_target:
# Keep obsolete docs URLs out of results while sending users and canonical signals to their replacement.
url = f"{html_baseurl}{app.builder.get_target_uri(redirect_target)}"
context["pageurl"] = url
refresh_url = app.builder.get_relative_uri(pagename, redirect_target)
esc = html.escape
extra = []
if not match:
extra.append(f'<meta name="description" content="{esc(description)}" />')
if pagename in _NOINDEX_PAGES:
extra.append('<meta name="robots" content="noindex, follow" />')
if redirect_target:
extra += [
f'<meta http-equiv="refresh" content="0; url={esc(refresh_url)}" />',
]
extra += [
'<meta property="og:type" content="website" />',
f'<meta property="og:site_name" content="{esc(project)}" />',
f'<meta property="og:title" content="{esc(title)}" />',
f'<meta property="og:description" content="{esc(description)}" />',
f'<meta property="og:url" content="{esc(url)}" />',
f'<meta property="og:image" content="{_SOCIAL_IMAGE}" />',
'<meta name="twitter:card" content="summary_large_image" />',
'<meta name="twitter:site" content="@kornia_foss" />',
f'<meta name="twitter:title" content="{esc(title)}" />',
f'<meta name="twitter:description" content="{esc(description)}" />',
f'<meta name="twitter:image" content="{_SOCIAL_IMAGE}" />',
]
context["metatags"] = "\n".join(extra) + "\n" + metatags
# -- Deep-link compatibility ------------------------------------------------
# Pages that used to hold a whole module's reference and are now section indexes, with the
# objects spread over per-topic subpages. Years of external links point at anchors on these
# pages (``augmentation.module.html#kornia.augmentation.RandomAffine``), and a server-side
# redirect cannot see the fragment. Each of these pages therefore embeds a JSON map from
# object name to the page that documents it now, and ``_static/js/custom.js`` redirects when
# the requested fragment is in the map. Map the page to its old ``kornia.<module>`` prefix.
_SPLIT_INDEX_PAGES = {
"augmentation.module": "kornia.augmentation.",
"color": "kornia.color.",
"enhance": "kornia.enhance.",
"feature": "kornia.feature.",
"filters": "kornia.filters.",
"image": "kornia.image.",
"losses": "kornia.losses.",
"metrics": "kornia.metrics.",
}
def _inject_anchor_redirects(app, pagename, templatename, context, doctree):
"""Embed ``{new docname: [object names]}`` on the pages listed in ``_SPLIT_INDEX_PAGES``."""
prefix = _SPLIT_INDEX_PAGES.get(pagename)
if prefix is None:
return
moved: dict[str, list[str]] = {}
for fullname, entry in app.env.domaindata["py"]["objects"].items():
if fullname.startswith(prefix) and entry.docname != pagename:
moved.setdefault(entry.docname, []).append(fullname)
if not moved:
return
context["metatags"] = (context.get("metatags", "") or "") + "\n" + _json_script("kornia-anchor-redirects", moved)
def _json_script(element_id, data):
payload = json.dumps(data, separators=(",", ":")).replace("</", "<\\/")
return f'<script type="application/json" id="{element_id}">{payload}</script>'
# -- Navbar menus -----------------------------------------------------------
def _navbar_menu_entries():
for menu in NAVBAR_MENUS.values():
groups = menu.values() if isinstance(menu, dict) else [menu]
for group in groups:
yield from group
def _check_navbar_menus(app, env):
"""Fail the build (under ``-W``) when a menu entry points at a page that no longer exists."""
for label, target in _navbar_menu_entries():
if target is None or "://" in target:
continue
if target not in env.found_docs:
logger.warning("NAVBAR_MENUS entry %r points at unknown document %r", label, target)
def _inject_navbar_menus(app, pagename, templatename, context, doctree):
context["metatags"] = (context.get("metatags", "") or "") + "\n" + _json_script("kornia-navbar-menus", NAVBAR_MENUS)
# --- "Try in browser" badges: link an autodoc'd API to its interactive playground page ----------
# The map (fully-qualified name -> page URL) is a snapshot of the kornia.org playground; refresh it
# with docs/generate_playground_links.py. Functions, their nn.Module counterparts and browser models
# that have a page get a small badge after their summary line.
try:
with open(os.path.join(os.path.dirname(__file__), "_playground_links.json"), encoding="utf-8") as _f:
_PLAYGROUND_LINKS = json.load(_f)
except (OSError, json.JSONDecodeError): # the docs must still build without the snapshot
_PLAYGROUND_LINKS = {}
def _inject_playground_link(app, what, name, obj, options, lines):
if what not in ("function", "class"):
return
url = _PLAYGROUND_LINKS.get(name)
if not url:
return
badge = ["", ".. container:: kornia-try", "", f" `Try in browser <{url}>`__", ""]
# keep the summary as the first line (autosummary uses it) and drop the badge in just after it
cut = next((i for i, line in enumerate(lines) if line.strip() == ""), len(lines))
lines[cut:cut] = badge
def setup(app):
app.connect("autodoc-process-docstring", _inject_playground_link)
# Run after the theme's canonical URL normalizer so redirects can replace ``pageurl``.
app.connect("html-page-context", _inject_social_metatags, priority=900)
app.connect("html-page-context", _inject_anchor_redirects)
app.connect("html-page-context", _inject_navbar_menus)
app.connect("env-check-consistency", _check_navbar_menus)