Wrong URL for i18 multisite

[arbinada-com]

When using i18_multisite plugin the generation of search.toml uses wrong URLs.

For example

[[input.files]]
path = "pages/about_fr.html"
url = "/pages/about_fr.html"

should be

[[input.files]]
path = "pages/about_fr.html"
url = "/fr/pages/about_fr.html"

To solve:

• add SEARCH_I18_SUBSITE parameter to generator
• in site configuration add as follows

I18N_SUBSITES = {
    "fr": {
        "SITENAME": "Mon beau site",
        "SEARCH_I18_SUBSITE": "fr"
    }
}

The patch for search.py:

--- ./search.py.bak	2023-02-20 17:00:10.379008769 +0100
+++ ./search.py	2023-02-20 17:19:26.119005800 +0100
@@ -30,6 +30,7 @@
         self.tpages = settings.get("TEMPLATE_PAGES")
         self.search_mode = settings.get("SEARCH_MODE", "output")
         self.html_selector = settings.get("SEARCH_HTML_SELECTOR", "main")
+        self.i18_subsite = settings.get("SEARCH_I18_SUBSITE", None)
 
     def build_search_index(self, search_settings_path):
         if not which("stork"):
@@ -62,6 +63,7 @@
             pages += article.translations
 
         input_files = ""
+        url_i18_prefix = self.i18_subsite + "/" if self.i18_subsite is not None else ""
 
         # Generate list of articles and pages to index
         for page in pages:
@@ -72,7 +74,7 @@
             input_file = f"""
                 [[input.files]]
                 path = "{page_to_index}"
-                url = "/{page.url}"
+                url = "/{url_i18_prefix}{page.url}"
                 title = {dumps(striptags(page.title))}
             """
             input_files = "".join([input_files, input_file])
@@ -86,7 +88,7 @@
             input_file = f"""
                 [[input.files]]
                 path = "{tpage_to_index}"
-                url = "{self.tpages[tpage]}"
+                url = "{url_i18_prefix}{self.tpages[tpage]}"
                 title = ""
             """
             input_files = "".join([input_files, input_file])

[qforce85]

The above patch is outdated. Here's a fork from the search plugin version 1.1.0, with patch correctly applied:

"""Search plugin for Pelican to generate an index for static site searches.

Copyright (c) Justin Mayer
"""

import logging
from pathlib import Path
from shutil import which
import subprocess
from typing import Dict, List

from jinja2.filters import do_striptags as striptags
import rtoml

from pelican import signals

logger = logging.getLogger(__name__)


class SearchSettingsGenerator:
    """Generate site search settings."""

    def __init__(self, context, settings, path, theme, output_path, *null):
        self.output_path = output_path
        self.context = context
        self.content = settings.get("PATH")
        self.tpages = settings.get("TEMPLATE_PAGES")
        self.input_options = settings.get("STORK_INPUT_OPTIONS", {})
        self.output_options = settings.get("STORK_OUTPUT_OPTIONS")

        # Add i18n_multisite support
        self.i18_subsite = settings.get("SEARCH_I18_SUBSITE", None)

        # Set default values
        self.input_options.setdefault("html_selector", "main")
        self.input_options.setdefault("base_directory", self.output_path)

        # Handle deprecated settings
        if settings.get("SEARCH_HTML_SELECTOR") and not settings.get(
                "STORK_INPUT_OPTIONS"
        ):
            logger.warning(
                "The SEARCH_HTML_SELECTOR setting is deprecated "
                "and will be removed in a future version. "
                "Use the STORK_INPUT_OPTIONS setting instead."
            )
            self.input_options["html_selector"] = settings.get("SEARCH_HTML_SELECTOR")
        if settings.get("SEARCH_MODE") and not settings.get("STORK_INPUT_OPTIONS"):
            logger.warning(
                f"SEARCH_MODE = {settings.get('SEARCH_MODE')} is deprecated "
                "and will be removed in a future version. "
                "Use the STORK_INPUT_OPTIONS setting instead."
            )
            self.input_options["base_directory"] = self.output_path
            if settings.get("SEARCH_MODE") == "source":
                self.input_options["base_directory"] = self.content

    def generate_output(self, writer):
        search_settings_path = Path(self.output_path) / "search.toml"

        self.generate_stork_settings(search_settings_path)

        # Build the search index
        build_log = self.build_search_index(search_settings_path)
        build_log = "".join(["Search plugin reported ", build_log])
        logger.error(build_log) if "error" in build_log else logger.debug(build_log)

    def build_search_index(self, search_settings_path: Path):
        if not which("stork"):
            raise Exception("Stork must be installed and available on $PATH.")
        try:
            output = subprocess.run(
                [
                    "stork",
                    "build",
                    "--input",
                    str(search_settings_path),
                    "--output",
                    f"{self.output_path}/search-index.st",
                ],
                capture_output=True,
                encoding="utf-8",
                check=True,
            )
        except subprocess.CalledProcessError as e:
            error_output = "".join(["Search plugin reported ", e.stdout, e.stderr])
            raise Exception(error_output) from e

        return output.stdout

    def generate_stork_settings(self, search_settings_path: Path):
        self.input_options["files"] = self.get_input_files()

        search_settings = {"input": self.input_options}

        if self.output_options:
            search_settings["output"] = self.output_options

        # Write the search settings file to disk
        with search_settings_path.open("w", encoding="utf-8") as fd:
            rtoml.dump(obj=search_settings, file=fd)

    def _index_output(self) -> bool:
        return self.input_options["base_directory"] == self.output_path

    def _get_url_with_i18_prefix(self, url: str) -> str:
        """Generate URL with i18n prefix if configured."""
        if self.i18_subsite:
            # Remove leading slash if present to avoid double slashes
            clean_url = url.lstrip('/')
            return f"/{self.i18_subsite}/{clean_url}"
        else:
            # Ensure URL starts with /
            return f"/{url.lstrip('/')}"

    def get_input_files(
            self,
    ) -> List[Dict]:
        pages = self.context["pages"] + self.context["articles"]

        for article in self.context["articles"]:
            pages += article.translations

        input_files = []
        # Generate list of articles and pages to index
        for page in pages:
            page_to_index = (
                page.save_as if self._index_output() else page.relative_source_path
            )
            # Escape double-quotation marks in the title
            title = striptags(page.title).replace('"', '\\"')
            input_files.append(
                {
                    "path": page_to_index,
                    "url": self._get_url_with_i18_prefix(page.url),
                    "title": f"{title}",
                }
            )

        # Generate list of *template* pages to index (if any)
        for tpage in self.tpages:
            tpage_to_index = self.tpages[tpage] if self._index_output() else tpage
            input_files.append(
                {
                    "path": tpage_to_index,
                    "url": self._get_url_with_i18_prefix(self.tpages[tpage]),
                    "title": ""
                }
            )

        return input_files


def get_generators(generators):
    """Get the search settings generator."""
    return SearchSettingsGenerator


def register():
    """Register the plugin."""
    signals.get_generators.connect(get_generators)

Note that in pelicanconf.py the SEARCH_I18_SUBSITE attribute must not only be defined for the translation subsites, but also globally, as SEARCH_I18_SUBSITE = None. Without the latter, the i18n-subsites plugin will not pass the attribute to the search plugin.