#!/bin/sh
set -eu

export LANG=POSIX.UTF-8


normalize() {
	xargs realpath --relative-to=. | sort
}

## A language's front page is excluded: it is a listing now, and
## naming it here as well puts it in $(sources.adoc) twice.
pages() {
	find src/pages/*/*.adoc -type f ! -name 'index.adoc'     | normalize
}

## The front page of each language, which lists what is newest
## across that language's collections rather than holding articles
## of its own.
homes() {
	find src/pages/*/index.adoc -type f                      | normalize
}

articles() {
	find src/collections/*/*/*/ -type f -name '*.adoc'       | normalize
}

slides() {
	find src/slides/*/*/ -type f -name '*.eslaides'          | normalize
}

indexes() {
	find src/collections/*/*/index.adoc                      | normalize
}

categories() {
	grep -l '^:type: categories$' src/collections/*/*/*.adoc | normalize
}

media() {
	# src/content/music is excluded: those .ogg files are score
	# renderings, not published media, and the old site never
	# offered them over BitTorrent.
	find src/content/ -path 'src/content/music' -prune -or \
		\( \( -type f -and \! -type l \) -and \( \
			-name '*.flac' -or \
			-name '*.ogg'  -or \
			-name '*.webm'     \
		\) -print \)
}

music() {
	find src/content/music/*.ly                              | normalize
}

tarballs() {
	find src/content/ \( -type f -and \! -type l \) -and \( \
			-name '*.tar.gz'   \
		\)
}

extras() {
	media
	tarballs
}

listings() {
	indexes
	categories
	homes
}

## Which pages are one another's translations.  po4a already says
## so, set by set, so read it there rather than keep a second list
## that can quietly disagree with the first.
po4a_sets() {
	awk '
		/^\[type: asciidoc\]/ { set = "" }
		/^\[type: asciidoc\]/ || cont {
			line = $0
			sub(/^\[type: asciidoc\][ \t]*/, "", line)
			cont = (line ~ /\\$/)
			sub(/[ \t]*\\$/, "", line)
			n = split(line, f, /[ \t]+/)
			for (i = 1; i <= n; i++) {
				if (f[i] == "" || f[i] ~ /^add_/) continue
				sub(/^[a-z]+:/, "", f[i])
				set = set (set == "" ? "" : " ") f[i]
			}
			if (!cont && set != "") { print set; set = "" }
		}
	' po/po4a.cfg
}

files() {
	pages
	articles
	listings
}


pages                                      | varlist      'pages.adoc'
homes                                      | varlist      'homes.adoc'
articles                                   | varlist   'articles.adoc'
slides                                     | varlist 'slides.eslaides'
categories                                 | varlist 'categories.adoc'
indexes                                    | varlist    'indexes.adoc'
indexes | sed 's|/index\.adoc$|/feed.xml|' | varlist 'feeds.xml'
find src/content/img/ -name '*.svg' | sort | varlist 'images.svg'
media                                      | varlist 'sources.media'
tarballs                                   | varlist 'sources.tarballs'
extras                                     | varlist 'sources.extras'
find po/*.po po/*.pot                      | varlist 'sources.po'
music                                      | varlist 'music.ly'
find src/content/music/*.ogg               | varlist 'music.ogg'
find src/content/music/*.ly.include         | varlist 'music.include'

# lilypond emits the score alongside the MIDI, from one run.
music | sed 's/^\(.*\)\.ly$/\1.pdf:\t\1.midi/'


files    | sed 's/^\(.*\)\.adoc$/\1.htmlbody\t\1.snippets\t\1.json:\t\1.adoc/'
files    | sed 's/^\(.*\)\.adoc$/\1.html:\t\1.json\t\1.htmlbody/'
printf '\n'

files    | sed 's/^\(.*\)\.adoc$/\1.updatedat-check:\t\1.json/'
files    | sed 's/^\(.*\)\.adoc$/\1.links-internal-check:\t\1.links/'
files    | sed 's/^\(.*\)\.adoc$/\1.caslinks:\t\1.links/'
printf '\n'

articles | sed 's/^\(.*\)\.adoc$/\1.feedentry:\t\1.json\t\1.htmlbody/'
articles | sed 's/^\(.*\)\.adoc$/\1.sortdata:\t\1.json/'
articles | sed 's/^\(.*\)\.adoc$/\1.categorydata:\t\1.json/'
printf '\n'

listings | sed 's/^\(.*\)\.adoc$/\1.htmlheader\t\1.htmlfooter:\t\1.json/'
listings | sed 's/^\(.*\)\.adoc$/\1.htmllisting:\t\1.json/'
listings | sed 's/^\(.*\)\.adoc$/\1.html:\t\1.htmlheader\t\1.htmlfooter/'
listings | sed 's/^\(.*\)\.adoc$/\1.html:\t\1.htmllisting\t\1.htmlbody/'
printf '\n'

media    | sed 's/^\(.*\)$/\1.torrent:\t\1/'
printf '\n'

slides   | sed 's/^\(.*\)\.eslaides$/\1.ps:\t\1.eslaides/'
slides   | sed 's/^\(.*\)\.eslaides$/\1.pdf:\t\1.ps/'
printf '\n'



collectionentries() {
	langlink="$1"
	colllink="$2"
	lang="$(basename "$langlink")"
	c="$(printf '%s' "$colllink" | normalize)"
	plural="$(cat src/names/categories/"$lang".txt)"

	printf '\n\n'
	name="$(basename "$c")"
	art=articles."$lang"
	# A collection with no articles yet -- every non-English one --
	# has no YYYY/ directories for the glob to match.
	find "$c"/*/ -type f -name '*.adoc' 2>/dev/null |
		varlist "$art.$name.adoc"
	echo "$art.$name.sortdata     = \$($art.$name.adoc:.adoc=.sortdata)"
	echo "$art.$name.indexentry   = \$($art.$name.adoc:.adoc=.indexentry)"
	echo "$art.$name.feedentry    = \$($art.$name.adoc:.adoc=.feedentry)"
	echo "$art.$name.categorydata = \$($art.$name.adoc:.adoc=.categorydata)"
	echo "$art.$name.homeentry    = \$($art.$name.adoc:.adoc=.homeentry)"

	printf '%s/sortdata.txt:\tdeps.mk\n' "$c"
	printf '\tprintf %s $(%s.%s.sortdata) > $@\n\n' "'%s\n'" "$art" "$name"

	listings="
		feed.xml
		index.htmllisting
		$plural.htmllisting
		$plural.txt
	"
	for lst in $listings; do
		printf '%s/%s:\t%s/sortdata.txt\n'   "$c" "$lst" "$c"
		printf '%s/%s:\t$(%s.%s.sortdata)\n' "$c" "$lst" "$art" "$name"
	done

	printf      '%s/index.htmllisting\t'   "$c"
	printf '%s/%s.htmllisting:\t'          "$c" "$plural"
	printf '$(%s.%s.indexentry)\n'         "$art" "$name"

	printf '%s/%s.txt:\t'                  "$c" "$plural"
	printf '$(%s.%s.categorydata)\n'       "$art" "$name"

	printf '%s/%s.txt\t'                   "$c" "$plural"
	printf '%s/feed.xml:\t'                "$c"
	printf '$(%s.%s.feedentry)\n'          "$art" "$name"

	printf '%s/%s.htmllisting\t'           "$c" "$plural"
	printf '%s/%s.xml:\t'                  "$c" "$plural"
	printf '%s/%s.txt\n'                   "$c" "$plural"
}


for langlink in src/collections/*; do
	for colllink in "$langlink"/*; do
		collectionentries "$langlink" "$colllink"
	done
done

## A front page lists each collection, so it is rebuilt when any
## of that language's collections gains an entry.
printf '\n'
for langlink in src/collections/*; do
	lang="$(basename "$langlink")"
	deps=
	for colllink in "$langlink"/*; do
		[ -d "$colllink" ] || continue
		name="$(basename "$colllink")"
		deps="$deps src/content/$lang/$name/sortdata.txt"
		deps="$deps \$(articles.$lang.$name.homeentry)"
	done
	[ -n "$deps" ] || continue
	printf 'src/content/%s/index.htmllisting:%s\n\n' "$lang" "$deps"
done


## The map mkwb conf reads to link a page to its translations,
## written out here so it cannot drift from po4a.cfg.
printf '\ntranslations.map = src/translations.map\n\n'
printf 'src/translations.map: deps.mk\n'
printf '\tprintf %s \\\n' "'%s\\n'"
po4a_sets | sed "s/^/\t\t'/; s/\$/' \\\\/"
printf '\t\t> $@\n\n'


## The canonical rules, inlined so no include is needed
## at make time.
echo
mkwb rules