#!/bin/bash
# site-qa.sh - QA for the Grav Website repo and (with -l BASE) the live site.
#   ./site-qa.sh              repo checks: structure, frontmatter, link targets
#   ./site-qa.sh -l BASE      live checks: every known route returns 200, all
#                             internal links/assets resolve, pages have content
# Exit non-zero on any failure. Used by CI and the post-deploy sync job.
set -uo pipefail
cd "$(dirname "$0")/.."
FAIL=0
fail() { echo "FAIL: $*"; FAIL=$((FAIL+1)); }
route_of() { # pages-relative dir -> URL route (lowercase, strip NN. prefixes)
  echo "$1" | tr 'A-Z' 'a-z' | sed -E 's|/([0-9]+\.)|/|g; s|^([0-9]+\.)||'
}

repo_checks() {
  declare -A ROUTES
  while IFS= read -r f; do
    grep -q '^title:' "$f" || fail "no title frontmatter: $f"
    d=$(dirname "$f"); r=$(route_of "${d#pages/}")
    ROUTES["${r:-/}"]=1
  done < <(find pages -name '*.md')
  while IFS= read -r f; do
    d=$(dirname "$f")
    for link in $(grep -oE '\]\([^)#]+' "$f" | sed -E 's/^\]\(//'); do
      case "$link" in http*|mailto:*) continue ;; esac
      t=$(realpath -m --relative-to=. "$(dirname "$f")/$link" 2>/dev/null) || { fail "unresolvable link $link in $f"; continue; }
      t=${t#pages}; t=$(route_of "$t")
      [ -n "${ROUTES[${t:-/}]:-}" ] || fail "broken link '$link' in $f (no route '${t:-/}')"
    done
  done < <(find pages -name '*.md')
  echo "repo: ${#ROUTES[@]} routes checked"
}

live_checks() {
  local BASE=${1%/}
  local ROUTES=("/")
  while IFS= read -r f; do
    d=$(dirname "$f"); r=$(route_of "${d#pages/}")
    ROUTES+=("/${r#/}")
  done < <(find pages -name '*.md' | grep -v 'pages/01.home/home.md')
  local SITEMAP=/tmp/qa-sitemap.txt CHECKED=/tmp/qa-checked.txt
  : > $SITEMAP; : > $CHECKED
  for r in $(printf '%s\n' "${ROUTES[@]}" | sort -u); do
    code=$(curl -s -o /dev/null -w '%{http_code}' -m 20 --retry 1 "$BASE$r")
    if [ "$code" = "200" ]; then echo "$BASE$r" >> $SITEMAP; else fail "route $r -> $code"; fi
    sleep 0.2
  done
  for url in $(cat $SITEMAP); do
    body=$(curl -s -m 20 "$url"); sleep 0.2
    [ ${#body} -lt 500 ] && fail "page too small ($url, ${#body} bytes)"
    upath=${url#$BASE}; [ -z "$upath" ] && upath=/
    for link in $(echo "$body" | grep -oE 'href="[^"]*"' | sed 's/href="//; s/"$//' | sort -u); do
      case "$link" in
        http*|mailto:*|\#*) [[ "$link" == "$BASE"* ]] || continue ;;
        /*) link="$BASE$link" ;;
        *)  link="$BASE$(realpath -m "$upath/../$link" | sed 's|//\+|/|g')" ;;
      esac
      grep -qxF "$link" $CHECKED && continue
      echo "$link" >> $CHECKED
      case "$link" in
        *.css|*.js|*.png|*.jpg|*.jpeg|*.svg|*.ico|*.woff*|*.map|*.webp|*.txt|*.xml) ;;
        *) continue ;;
      esac
      code=$(curl -s -o /dev/null -w '%{http_code}' -m 20 "$link"); sleep 0.1
      [ "$code" != "200" ] && fail "asset $link -> $code"
    done
    for link in $(echo "$body" | grep -oE 'href="[^"]*"' | sed 's/href="//; s/"$//' | sort -u); do
      case "$link" in http*|mailto:*|\#*) continue ;; esac
      case "$link" in
        /*) L="$BASE$link" ;;
        *)  L="$BASE$(realpath -m "$upath/../$link" | sed 's|//\+|/|g')" ;;
      esac
      grep -qxF "$L" $CHECKED && continue
      echo "$L" >> $CHECKED
      case "$L" in
        *.css|*.js|*.png|*.jpg|*.jpeg|*.svg|*.ico|*.woff*|*.map|*.webp|*.txt|*.xml) continue ;;
      esac
      code=$(curl -s -o /dev/null -w '%{http_code}' -m 20 "$L"); sleep 0.1
      [ "$code" != "200" ] && fail "link $L -> $code"
    done
  done
  # canonical must be https (reverse-proxy scheme detection)
  can=$(curl -s -m 15 "$BASE/" | grep -oE '<link rel="canonical" href="[^"]*"' | head -1)
  case "$can" in *https://*) : ;; "") fail "no canonical link on home" ;; *) fail "canonical not https: $can" ;; esac
  echo "checked $(wc -l < $SITEMAP) routes, $(wc -l < $CHECKED) links/assets"
}

case "${1:-}" in
  -l) live_checks "$2" ;;
  *)  repo_checks ;;
esac
[ $FAIL -eq 0 ] && echo "QA: PASS" || { echo "QA: $FAIL failure(s)"; exit 1; }