plugin.py

# Copyright (c) 2016-2023 Martin Donath <martin.donath@squidfunk.com>

# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to
# deal in the Software without restriction, including without limitation the
# rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
# sell copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:

# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.

# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
# IN THE SOFTWARE.

import json
import logging
import os
import regex as re

from html import escape
from html.parser import HTMLParser
from mkdocs import utils
from mkdocs.plugins import BasePlugin

from .config import SearchConfig

try:
    import jieba
except ImportError:
    jieba = None

# -----------------------------------------------------------------------------
# Classes
# -----------------------------------------------------------------------------

# Search plugin
class SearchPlugin(BasePlugin[SearchConfig]):

    # Initialize plugin
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)

        # Initialize incremental builds
        self.is_dirty = False
        self.is_dirtyreload = False

        # Initialize search index cache
        self.search_index_prev = None

    # Determine whether we're serving the site
    def on_startup(self, *, command, dirty):
        self.is_dirty = dirty

    # Initialize plugin
    def on_config(self, config):
        if not self.config.enabled:
            return

        # Retrieve default value for language
        if not self.config.lang:
            self.config.lang = [self._translate(
                config, "search.config.lang"
            )]

        # Retrieve default value for separator
        if not self.config.separator:
            self.config.separator = self._translate(
                config, "search.config.separator"
            )

        # Retrieve default value for pipeline
        if not self.config.pipeline:
            self.config.pipeline = list(filter(len, re.split(
                r"\s*,\s*", self._translate(config, "search.config.pipeline")
            )))

        # Initialize search index
        self.search_index = SearchIndex(**self.config)

        # Set jieba dictionary, if given
        if self.config.jieba_dict:
            path = os.path.normpath(self.config.jieba_dict)
            if os.path.isfile(path):
                jieba.set_dictionary(path)
                log.debug(f"Loading jieba dictionary: {path}")
            else:
                log.warning(
                    f"Configuration error for 'search.jieba_dict': "
                    f"'{self.config.jieba_dict}' does not exist."
                )

        # Set jieba user dictionary, if given
        if self.config.jieba_dict_user:
            path = os.path.normpath(self.config.jieba_dict_user)
            if os.path.isfile(path):
                jieba.load_userdict(path)
                log.debug(f"Loading jieba user dictionary: {path}")
            else:
                log.warning(
                    f"Configuration error for 'search.jieba_dict_user': "
                    f"'{self.config.jieba_dict_user}' does not exist."
                )

    # Add page to search index
    def on_page_context(self, context, *, page, config, nav):
        if not self.config.enabled:
            return

        # Index page
        self.search_index.add_entry_from_context(page)
        page.content = re.sub(
            r"\s?data-search-\w+=\"[^\"]+\"",
            "",
            page.content
        )

    # Generate search index
    def on_post_build(self, *, config):
        if not self.config.enabled:
            return

        # Write search index
        base = os.path.join(config.site_dir, "search")
        path = os.path.join(base, "search_index.json")

        # Generate and write search index to file
        data = self.search_index.generate_search_index(self.search_index_prev)
        utils.write_file(data.encode("utf-8"), path)

        # Persist search index for repeated invocation
        if self.is_dirty:
            self.search_index_prev = self.search_index

    # Determine whether we're running under dirty reload
    def on_serve(self, server, *, config, builder):
        self.is_dirtyreload = self.is_dirty

    # -------------------------------------------------------------------------

    # Translate the given placeholder value
    def _translate(self, config, value):
        env = config.theme.get_env()

        # Load language template and return translation for placeholder
        language = "partials/language.html"
        template = env.get_template(language, None, { "config": config })
        return template.module.t(value)

# -----------------------------------------------------------------------------

# Search index with support for additional fields
class SearchIndex:

    # Initialize search index
    def __init__(self, **config):
        self.config = config
        self.entries = []

    # Add page to search index
    def add_entry_from_context(self, page):
        search = page.meta.get("search", {})
        if search.get("exclude"):
            return

        # Divide page content into sections
        parser = Parser()
        parser.feed(page.content)
        parser.close()

        # Add sections to index
        for section in parser.data:
            if not section.is_excluded():
                self.create_entry_for_section(section, page.toc, page.url, page)

    # Override: graceful indexing and additional fields
    def create_entry_for_section(self, section, toc, url, page):
        item = self._find_toc_by_id(toc, section.id)
        if item:
            url = url + item.url
        elif section.id:
            url = url + "#" + section.id

        # Set page title as section title if none was given, which happens when
        # the first headline in a Markdown document is not a h1 headline. Also,
        # if a page title was set via front matter, use that even though a h1
        # might be given or the page name was specified in nav in mkdocs.yml
        if not section.title:
            section.title = [str(page.meta.get("title", page.title))]

        # Compute title and text
        title = "".join(section.title).strip()
        text  = "".join(section.text).strip()

        # Segment Chinese characters if jieba is available
        if jieba:
            title = self._segment_chinese(title)
            text  = self._segment_chinese(text)

        # Create entry for section
        entry = {
            "location": url,
            "title": title,
            "text": text
        }

        # Set document tags
        tags = page.meta.get("tags")
        if isinstance(tags, list):
            entry["tags"] = []
            for name in tags:
                if name and isinstance(name, (str, int, float, bool)):
                    entry["tags"].append(name)

        # Set document boost
        search = page.meta.get("search", {})
        if "boost" in search:
            entry["boost"] = search["boost"]

        # Add entry to index
        self.entries.append(entry)

    # Generate search index
    def generate_search_index(self, prev):
        config = {
            key: self.config[key]
                for key in ["lang", "separator", "pipeline"]
        }

        # Hack: if we're running under dirty reload, the search index will only
        # include the entries for the current page. However, MkDocs > 1.4 allows
        # us to persist plugin state across rebuilds, which is exactly what we
        # do by passing the previously built index to this method. Thus, we just
        # remove the previous entries for the current page, and append the new
        # entries to the end of the index, as order doesn't matter.
        if prev and self.entries:
            path = self.entries[0]["location"]

            # Since we're sure that we're running under dirty reload, the list
            # of entries will only contain sections for a single page. Thus, we
            # use the first entry to remove all entries from the previous run
            # that belong to the current page. The rationale behind this is that
            # authors might add or remove section headers, so we need to make
            # sure that sections are synchronized correctly.
            entries = [
                entry for entry in prev.entries
                    if not entry["location"].startswith(path)
            ]

            # Merge previous with current entries
            self.entries = entries + self.entries

        # Otherwise just set previous entries
        if prev and not self.entries:
            self.entries = prev.entries

        # Return search index as JSON
        data = { "config": config, "docs": self.entries }
        return json.dumps(
            data,
            separators = (",", ":"),
            default = str
        )

    # -------------------------------------------------------------------------

    # Retrieve item for anchor
    def _find_toc_by_id(self, toc, id):
        for toc_item in toc:
            if toc_item.id == id:
                return toc_item

            # Recurse into children of item
            toc_item = self._find_toc_by_id(toc_item.children, id)
            if toc_item is not None:
                return toc_item

        # No item found
        return None

    # Find and segment Chinese characters in string
    def _segment_chinese(self, data):
        expr = re.compile(r"(\p{IsHan}+)", re.UNICODE)

        # Replace callback
        def replace(match):
            value = match.group(0)

            # Replace occurrence in original string with segmented version and
            # surround with zero-width whitespace for efficient indexing
            return "".join([
                "\u200b",
                "\u200b".join(jieba.cut(value.encode("utf-8"))),
                "\u200b",
            ])

        # Return string with segmented occurrences
        return expr.sub(replace, data).strip("\u200b")

# -----------------------------------------------------------------------------

# HTML element
class Element:
    """
    An element with attributes, essentially a small wrapper object for the
    parser to access attributes in other callbacks than handle_starttag.
    """

    # Initialize HTML element
    def __init__(self, tag, attrs = None):
        self.tag   = tag
        self.attrs = attrs or {}

    # String representation
    def __repr__(self):
        return self.tag

    # Support comparison (compare by tag only)
    def __eq__(self, other):
        if other is Element:
            return self.tag == other.tag
        else:
            return self.tag == other

    # Support set operations
    def __hash__(self):
        return hash(self.tag)

    # Check whether the element should be excluded
    def is_excluded(self):
        return "data-search-exclude" in self.attrs

# -----------------------------------------------------------------------------

# HTML section
class Section:
    """
    A block of text with markup, preceded by a title (with markup), i.e., a
    headline with a certain level (h1-h6). Internally used by the parser.
    """

    # Initialize HTML section
    def __init__(self, el, depth = 0):
        self.el = el
        self.depth = depth

        # Initialize section data
        self.text  = []
        self.title = []
        self.id = None

    # String representation
    def __repr__(self):
        if self.id:
            return "#".join([self.el.tag, self.id])
        else:
            return self.el.tag

    # Check whether the section should be excluded
    def is_excluded(self):
        return self.el.is_excluded()

# -----------------------------------------------------------------------------

# HTML parser
class Parser(HTMLParser):
    """
    This parser divides the given string of HTML into a list of sections, each
    of which are preceded by a h1-h6 level heading. A white- and blacklist of
    tags dictates which tags should be preserved as part of the index, and
    which should be ignored in their entirety.
    """

    # Initialize HTML parser
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)

        # Tags to skip
        self.skip = set([
            "object",                  # Objects
            "script",                  # Scripts
            "style"                    # Styles
        ])

        # Tags to keep
        self.keep = set([
            "p",                       # Paragraphs
            "code", "pre",             # Code blocks
            "li", "ol", "ul",          # Lists
            "sub", "sup"               # Sub- and superscripts
        ])

        # Current context and section
        self.context = []
        self.section = None

        # All parsed sections
        self.data = []

    # Called at the start of every HTML tag
    def handle_starttag(self, tag, attrs):
        attrs = dict(attrs)

        # Ignore self-closing tags
        el = Element(tag, attrs)
        if not tag in void:
            self.context.append(el)
        else:
            return

        # Handle heading
        if tag in ([f"h{x}" for x in range(1, 7)]):
            depth = len(self.context)
            if "id" in attrs:

                # Ensure top-level section
                if tag != "h1" and not self.data:
                    self.section = Section(Element("hx"), depth)
                    self.data.append(self.section)

                # Set identifier, if not first section
                self.section = Section(el, depth)
                if self.data:
                    self.section.id = attrs["id"]

                # Append section to list
                self.data.append(self.section)

        # Handle preface - ensure top-level section
        if not self.section:
            self.section = Section(Element("hx"))
            self.data.append(self.section)

        # Handle special cases to skip
        for key, value in attrs.items():

            # Skip block if explicitly excluded from search
            if key == "data-search-exclude":
                self.skip.add(el)
                return

            # Skip line numbers - see https://bit.ly/3GvubZx
            if key == "class" and value == "linenodiv":
                self.skip.add(el)
                return

        # Render opening tag if kept
        if not self.skip.intersection(self.context) and tag in self.keep:

            # Check whether we're inside the section title
            data = self.section.text
            if self.section.el in self.context:
                data = self.section.title

            # Append to section title or text
            data.append(f"<{tag}>")

    # Called at the end of every HTML tag
    def handle_endtag(self, tag):
        if not self.context or self.context[-1] != tag:
            return

        # Check whether we're exiting the current context, which happens when
        # a headline is nested in another element. In that case, we close the
        # current section, continuing to append data to the previous section,
        # which could also be a nested section – see https://bit.ly/3IxxIJZ
        if self.section.depth > len(self.context):
            for section in reversed(self.data):
                if section.depth <= len(self.context):

                    # Set depth to infinity in order to denote that the current
                    # section is exited and must never be considered again.
                    self.section.depth = float("inf")
                    self.section = section
                    break

        # Remove element from skip list
        el = self.context.pop()
        if el in self.skip:
            if el.tag not in ["script", "style", "object"]:
                self.skip.remove(el)
            return

        # Render closing tag if kept
        if not self.skip.intersection(self.context) and tag in self.keep:

            # Check whether we're inside the section title
            data = self.section.text
            if self.section.el in self.context:
                data = self.section.title

            # Search for corresponding opening tag
            index = data.index(f"<{tag}>")
            for i in range(index + 1, len(data)):
                if not data[i].isspace():
                    index = len(data)
                    break

            # Remove element if empty (or only whitespace)
            if len(data) > index:
                while len(data) > index:
                    data.pop()

            # Append to section title or text
            else:
                data.append(f"</{tag}>")

    # Called for the text contents of each tag
    def handle_data(self, data):
        if self.skip.intersection(self.context):
            return

        # Collapse whitespace in non-pre contexts
        if not "pre" in self.context:
            if not data.isspace():
                data = data.replace("\n", " ")
            else:
                data = " "

        # Handle preface - ensure top-level section
        if not self.section:
            self.section = Section(Element("hx"))
            self.data.append(self.section)

        # Handle section headline
        if self.section.el in self.context:
            permalink = False
            for el in self.context:
                if el.tag == "a" and el.attrs.get("class") == "headerlink":
                    permalink = True

            # Ignore permalinks
            if not permalink:
                self.section.title.append(
                    escape(data, quote = False)
                )

        # Collapse adjacent whitespace
        elif data.isspace():
            if not self.section.text or not self.section.text[-1].isspace():
                self.section.text.append(data)
            elif "pre" in self.context:
                self.section.text.append(data)

        # Handle everything else
        else:
            self.section.text.append(
                escape(data, quote = False)
            )

# -----------------------------------------------------------------------------
# Data
# -----------------------------------------------------------------------------

# Set up logging
log = logging.getLogger("mkdocs.material.search")

# Tags that are self-closing
void = set([
    "area",                            # Image map areas
    "base",                            # Document base
    "br",                              # Line breaks
    "col",                             # Table columns
    "embed",                           # External content
    "hr",                              # Horizontal rules
    "img",                             # Images
    "input",                           # Input fields
    "link",                            # Links
    "meta",                            # Metadata
    "param",                           # External parameters
    "source",                          # Image source sets
    "track",                           # Text track
    "wbr"                              # Line break opportunities
])