slugify() {
	printf '%s\n' "$1" |
		tr '[:upper:]' '[:lower:]' |
		sed -e 's/[^a-z0-9][^a-z0-9]*/-/g' \
		    -e 's/^-//' \
		    -e 's/-$//' \
		    -e 's/^$/section/'
}

heading_text() {
	tag=$1
	printf '%s\n' "$2" |
	awk -v tag="$tag" '
		function decode(s) {
			gsub(/"|"/, "\"", s)
			gsub(/'|'/, "\047", s)
			gsub(/&lt;|&#60;/, "<", s)
			gsub(/&gt;|&#62;/, ">", s)
			gsub(/&amp;|&#38;/, "\\&", s)
			return s
		}
		{
			line = $0
			start = index(line, "<" tag)
			end = index(line, "</" tag ">")
			if (!start || !end || end < start)
				exit
			head = substr(line, start, end - start)
			for (i = 1; i <= length(head); i++)
				if (substr(head, i, 1) == ">") {
					close_pos = i
					break
				}
			if (!close_pos)
				exit
			text = substr(head, close_pos + 1)
			if (text !~ /</)
				print decode(text)
			exit
		}
	'
}

read_sections() {
	source=$1
	slug=$2
	heading_meta=$tmp/headings-$slug.tsv
	section_ids=$tmp/section-ids-$slug.txt
	subsection_meta=$tmp/subsections-$slug.tsv
	subsection_parents=$tmp/subsection-parents-$slug.txt
	: > "$heading_meta"
	: > "$section_ids"
	: > "$subsection_meta"
	: > "$subsection_parents"
	current_section=

	while IFS= read -r line || [ -n "$line" ]; do
		case $line in
			*'<h2'*)
				heading_tag=h2
				closing='</h2>'
				;;
			*'<h3'*)
				heading_tag=h3
				closing='</h3>'
				;;
			*) continue ;;
		esac
		case $line in
			*"$closing"*) ;;
			*) die "Indexable $heading_tag must be plain one-line text: $source" ;;
		esac

		heading_title=$(heading_text "$heading_tag" "$line")
		[ -n "$heading_title" ] ||
			die "Indexable $heading_tag must contain plain one-line text: $source"
		case $heading_title in
			*"$tab"*) die "Heading title may not contain tabs: $source" ;;
		esac

		heading_id=$(attr id "$line")
		if [ -n "$heading_id" ]; then
			safe_name "$heading_id" ||
				die "Unsafe $heading_tag id in $source: $heading_id"
		else
			case $heading_tag in
				h2) anchor_base=$slug-$(slugify "$heading_title") ;;
				h3)
					[ -n "$current_section" ] ||
						die "Indexable h3 must follow an h2: $source"
					anchor_base=$current_section-$(slugify "$heading_title")
					;;
			esac
			heading_id=$(unique_auto_anchor "$anchor_base")
		fi
		reserve_anchor "$heading_id" "$source"

		case $heading_tag in
			h2)
				current_section=$heading_id
				parent_id=
				printf '%s\n' "$heading_id" >> "$section_ids"
				;;
			h3)
				[ -n "$current_section" ] ||
					die "Indexable h3 must follow an h2: $source"
				parent_id=$current_section
				printf '%s\t%s\n' "$parent_id" "$heading_id" >> "$subsection_meta"
				if ! grep -F -x -q -e "$parent_id" "$subsection_parents"; then
					printf '%s\n' "$parent_id" >> "$subsection_parents"
				fi
				subsection_count=$((subsection_count + 1))
				;;
		esac
		printf '%s\t%s\t%s\t%s\n' \
			"$heading_tag" "$heading_id" "$heading_title" "$parent_id" >> "$heading_meta"
	done < "$source"
}