learning-website-framework1-5  Exercise 2: Resolve a Link From Any Site
===========================================================================

In the content, links are written as root-relative paths of the old single
site (for example /linux/shell-and-scripting/vim/). On a multi-site build, a
link must stay relative when the target is on the same site and become an
absolute URL when the target is on another site.

Save the full site map as sitemap.py (the same map as Learning Website:
Framework & Architecture 2, plus the one special prefix for the kanji
resources):

SITES = {
    "languages": ["france", "germany", "hungary", "japan", "culture"],
    "webdevelopment": ["web-development", "web-platforms", "web-servers"],
    "programming": ["programming", "software-development", "maths-for-programmers",
                    "retro-computing", "game-development", "android-development",
                    "ios-development", "databases", "blockchain-and-web3",
                    "developer-tools", "projects"],
    "systems": ["linux", "operating-systems", "networking", "cloud-and-devops",
                "security", "technical-support", "raspberry-pi", "windows"],
    "ai": ["ai", "data-science-and-ml"],
    "humanities": ["history", "politics", "philosophy", "art-history",
                   "classic-literature", "science-fiction", "sci-fi-project",
                   "music", "football", "ethics", "study-methodologies", "science"],
    "lifeskills": ["practical-life-skills", "cookery",
                   "freelancing-and-business-skills", "youtube"],
    "creative": ["creative-and-design-tools", "audio-video-production",
                 "office-and-productivity-software"],
}
SIDEBAR_ROUTES = {
    "ai": "ai", "drinks": "lifeskills", "football": "humanities",
    "history": "humanities", "linux": "systems",
    "programming": "programming", "web-development": "webdevelopment",
}
# URLs that are not under a content/ folder name (see Learning Website: Framework & Architecture 2)
SPECIAL_PREFIXES = {"resources/japanese": "languages"}

FOLDER_TO_SITE = {f: s for s, fs in SITES.items() for f in fs}

def site_for(path):
    path = path.strip("/")
    for prefix, site in SPECIAL_PREFIXES.items():
        if path == prefix or path.startswith(prefix + "/"):
            return site
    parts = path.split("/")
    if parts[0] == "sidebar":
        return SIDEBAR_ROUTES[parts[1]]
    return FOLDER_TO_SITE[parts[0]]

Save the resolver as links.py:

from urllib.parse import urlsplit, urlunsplit
from sitemap import site_for

BASE = {"production": "https://{site}.osztromok.com",
        "dev":        "http://{site}.localhost:4321"}

def link_for(current_site, href, env="production"):
    """Turn an old-style site-root path into the right link from current_site.

    - external links and in-page anchors are left alone
    - same-site targets stay relative (root-relative, trailing slash kept)
    - other-site targets become absolute URLs on that site
    """
    parts = urlsplit(href)
    if parts.scheme or parts.netloc or not href.startswith("/"):
        return href
    target = site_for(parts.path)
    if target == current_site:
        return href
    base = BASE[env].format(site=target)
    return urlunsplit((urlsplit(base).scheme, urlsplit(base).netloc,
                       parts.path, parts.query, parts.fragment))

if __name__ == "__main__":
    tests = [
        ("languages", "/japan/hiragana-1/"),
        ("languages", "/linux/shell-and-scripting/vim/"),
        ("systems",   "/linux/shell-and-scripting/vim/#modes"),
        ("languages", "/resources/japanese/kanji/kanji_%E4%B8%8A.html"),
        ("systems",   "/sidebar/linux/"),
        ("languages", "/sidebar/football/"),
        ("languages", "https://example.org/page/"),
        ("languages", "#top"),
        ("systems",   "../relative/page.html"),
    ]
    for cur, href in tests:
        print(f"{cur:<10} {href:<52} -> {link_for(cur, href)}")
    print()
    print("dev:", link_for("languages", "/linux/", env="dev"))

Run it:

  python links.py

Output (checked by running it):

languages  /japan/hiragana-1/                                   -> /japan/hiragana-1/
languages  /linux/shell-and-scripting/vim/                      -> https://systems.osztromok.com/linux/shell-and-scripting/vim/
systems    /linux/shell-and-scripting/vim/#modes                -> /linux/shell-and-scripting/vim/#modes
languages  /resources/japanese/kanji/kanji_%E4%B8%8A.html       -> /resources/japanese/kanji/kanji_%E4%B8%8A.html
systems    /sidebar/linux/                                      -> /sidebar/linux/
languages  /sidebar/football/                                   -> https://humanities.osztromok.com/sidebar/football/
languages  https://example.org/page/                            -> https://example.org/page/
languages  #top                                                 -> #top
systems    ../relative/page.html                                -> ../relative/page.html

dev: http://systems.localhost:4321/linux/

Design points
-------------
- External links, in-page anchors (#top) and relative paths (../x.html) are
  never touched. Only root-relative paths are resolved.
- The path, query and fragment are preserved exactly, including the trailing
  slash that the site's Astro config requires (trailingSlash: 'always').
- The base address per environment is in one table, so local development can
  use *.localhost while production uses osztromok.com.
- A path that belongs to no site raises a KeyError on purpose: a broken link
  should fail the build, not be shipped.
- The kanji resource pages are served from /resources/japanese/..., not from a
  content folder name, so they need an explicit entry in SPECIAL_PREFIXES.

WHY THIS WORKS AS AN ANSWER
---------------------------
Content authors can keep writing simple links, and the build decides what is
local and what is cross-site. If a folder moves to another site, one line of
the map changes and every link follows.