#!/usr/bin/env python3
"""
This script automates the process of building and deploying the website.
It handles tasks such as:

- Running orgo to generate site content.
- Rewriting the footer's orgo version to match the orgo that built the site.
- Rewriting image URLs for the onion service (production only).
- Optionally deploying the built site to a remote server.
- Starting a local development server for previewing changes.

Usage:
    Set BUILD=true to build, DEPLOY=true to deploy or serve.
    Set ENV=prod for production builds; anything else builds for development.

    Builds are incremental. Production writes .build/ and development .build-dev/;
    delete either one for a clean build.

Dependencies:
    - Python 3
    - orgo, the static site generator (reads content/orgo.toml)
    - rsync for deployment

Author:
    Christian Cleberg <hello@cleberg.net>
"""

import os
import re
import subprocess
import sys
from pathlib import Path


def run(cmd, error, echo=False):
    """Run cmd quietly, exiting with its stderr if it fails.

    echo prints stdout on success. Only the dry run needs it: its output is the
    whole reason to run it, and the default of swallowing stdout would hide it.
    """
    result = subprocess.run(cmd, capture_output=True, text=True, check=False)
    if result.returncode != 0:
        print(error, file=sys.stderr)
        print(result.stderr, file=sys.stderr)
        sys.exit(1)
    if echo:
        print(result.stdout, end="")


def run_ruff():
    print("Running ruff...")
    for cmd in [["ruff", "check", "--fix"], ["ruff", "format"]]:
        run(cmd, f"ruff error ({' '.join(cmd)}):")


def rewrite_img_urls(build_dir=".build"):
    """
    Rewrite absolute img.cleberg.net URLs to root-relative /img/ paths so the
    onion serves images from its own origin instead of fetching them off-onion.
    Production only: dev builds keep the absolute URLs so local previews still
    load images from the live image host.

    Requires the server to serve /var/www/img/ at /img/ on the cleberg.net vhost
    (e.g. `ln -s /var/www/img /var/www/cleberg.net/img`).

    The pattern tolerates a stray number of slashes after the scheme
    (https:/img, https:///img, ...) so a typo in the org source can't silently
    slip through un-rewritten and ship a broken cross-origin URL.
    """
    pattern = re.compile(r"https:/+img\.cleberg\.net/")
    count = 0
    for html in Path(build_dir).rglob("*.html"):
        text = html.read_text(encoding="utf-8")
        new_text, n = pattern.subn("/img/", text)
        if n:
            count += n
            html.write_text(new_text, encoding="utf-8")
    print(f"Rewrote {count} img.cleberg.net references to /img/")


# The footer's "built with orgo <version>" note. Matching the whole phrase rewrites a
# page carrying an old version as readily as one still carrying the template's
# placeholder, and leaves prose that ends a paragraph with a link to orgo alone.
ORGO_NOTE = re.compile(
    r'(built with <a href="https://gitbay\.org/krz/orgo">orgo</a> )[^<]*(</p>)'
)


def orgo_version():
    """The version reported by the orgo on PATH, e.g. "0.22.0"."""
    result = subprocess.run(
        ["orgo", "--version"], capture_output=True, text=True, check=False
    )
    if result.returncode != 0:
        print("Could not read the orgo version:", file=sys.stderr)
        print(result.stderr, file=sys.stderr)
        sys.exit(1)
    # `orgo --version` prints "orgo X.Y.Z". Anything else means the output format
    # changed, and guessing at it would ship a wrong version to every page.
    parts = result.stdout.split()
    if len(parts) != 2 or parts[0] != "orgo":
        print(f"Unexpected `orgo --version` output: {result.stdout!r}", file=sys.stderr)
        sys.exit(1)
    return parts[1]


def rewrite_orgo_version(build_dir):
    """Point the footer's orgo version at the orgo that just built the site.

    Runs over every page rather than only the re-rendered ones, because the build cache
    keys on a cache *format* version and not on orgo's release version: after a version
    bump that leaves the format alone, the pages carried over from the previous build
    would otherwise keep printing the old one.

    The whole point of this is that the note cannot go stale, so a footer that no longer
    matches is a failure and not a no-op — matching nothing anywhere means the note was
    removed or its markup changed, and a silent pass would ship the template's
    placeholder to every page.
    """
    version = orgo_version()
    count = 0
    for html in Path(build_dir).rglob("*.html"):
        text = html.read_text(encoding="utf-8")
        new_text, n = ORGO_NOTE.subn(rf"\g<1>{version}\g<2>", text)
        if n and new_text != text:
            html.write_text(new_text, encoding="utf-8")
        count += n
    if count == 0:
        print(
            "No 'built with orgo' note found in the build — the footer in "
            "content/templates/base.html no longer matches ORGO_NOTE",
            file=sys.stderr,
        )
        sys.exit(1)
    print(f"Set the footer orgo version to {version} on {count} pages")


def run_orgo_build(build_dir):
    """
    Build the site with orgo.

    orgo reads content/orgo.toml, which holds the routes, templates and collections: the
    blog index groups itself by year, the tags page is a collection, the recent-posts
    list comes from the blog collection, and the sitemap is written once base_url is set.

    The output directory is left in place between builds rather than wiped, because the
    .orgo-cache.json inside it is what makes a build incremental: orgo re-renders only
    the pages whose content, config or templates changed, re-emits any page whose output
    is missing, and deletes the outputs of pages that have since been removed. Wiping the
    directory would throw that away and force a full render every time. For a clean build
    from scratch, delete the output directory by hand.

    What is left around it is what orgo does not do: rewriting image URLs for the onion.
    """
    print("Building with orgo...")
    result = subprocess.run(
        ["orgo", "build", "content", "-o", str(build_dir), "--strict"],
        stdout=subprocess.PIPE,
        stderr=subprocess.STDOUT,
        text=True,
        check=False,
    )
    print(result.stdout, end="")
    if result.returncode != 0:
        print("orgo build failed", file=sys.stderr)
        sys.exit(1)


def deploy_to_server(build_dir, server, dry_run=False):
    """Push the built site to the server, or show what pushing it would do.

    The deploy deletes remote files the build no longer produces, so "what would
    this remove" is a question worth being able to ask before answering it
    irreversibly. DRY_RUN=true asks it: rsync connects and compares, then reports
    instead of transferring.
    """
    remote_path = f"{server}:/var/www/cleberg.net/"
    print(f"{'Would deploy' if dry_run else 'Deploying'} {build_dir}/ → {remote_path}")
    cmd = [
        "rsync",
        "-rt",
        # -t, or rsync gives every transferred file the time it arrived, no two
        # deploys agree on a timestamp, and the next run re-sends the whole site
        # because mtimes never match. With it, only what changed goes over.
        "--delete-before",
        # The build cache lives in the output directory because it describes it, but it
        # is not part of the site. Excluding it also stops --delete removing it locally.
        "--exclude",
        ".orgo-cache.json",
    ]
    if dry_run:
        # --itemize-changes because --dry-run alone prints almost nothing: the
        # point is to name every file that would be sent or deleted.
        cmd += ["--dry-run", "--itemize-changes"]
    cmd += [f"{build_dir}/", remote_path]
    run(cmd, "Error during rsync deployment:", echo=dry_run)


def start_dev_server(build_dir):
    print(f"Starting development HTTP server from {build_dir}/ on port 8000")
    os.chdir(build_dir)
    # This will run until interrupted (Ctrl+C)
    try:
        subprocess.run([sys.executable, "-m", "http.server", "8000"], check=True)
    except KeyboardInterrupt:
        print("\nDevelopment server stopped.")
    except subprocess.CalledProcessError as e:
        print(f"Error starting development server: {e}", file=sys.stderr)
        sys.exit(1)


def main():
    prod = os.environ.get("ENV", "").casefold() == "prod"
    if not prod:
        run_ruff()

    # One output directory per environment, because the two differ after orgo has run:
    # production rewrites image URLs in place and development does not. Sharing a
    # directory would let an incremental build reuse a page rendered for the other one —
    # a dev preview showing /img/ paths that only resolve on the server. Production keeps
    # .build/ because that is the directory the deploy and the CI manifest name.
    build_dir = Path(".build" if prod else ".build-dev")

    print(f"Environment: {'Production' if prod else 'Development'}")

    if os.environ.get("BUILD", "").casefold() == "true":
        run_orgo_build(build_dir)
        rewrite_orgo_version(build_dir)
        # The onion needs same-origin images; dev previews keep the absolute URLs.
        # Runs over every page, not just the re-rendered ones, so a page carried over
        # from an earlier build is rewritten too.
        if prod:
            rewrite_img_urls(build_dir)

    if os.environ.get("DEPLOY", "").casefold() == "true":
        if prod:
            dry_run = os.environ.get("DRY_RUN", "").casefold() == "true"
            print(
                "Dry run — the server will not be modified"
                if dry_run
                else "Deploying to production..."
            )
            deploy_to_server(build_dir, "homelab", dry_run)
        else:
            start_dev_server(build_dir)


if __name__ == "__main__":
    main()