split database into parts

Each category - sound/, networking/, et c. - now has its own little
database in .dbs/.  So the database only recompiles based on categories,
rather than a complete recompile every time a single file changes.

Makefile

@@ -10,19 +10,28 @@ articles != find * -type f -name "*.md"
 
 categories != ls -d */
 
-db.rec: $(articles)
+databases = $(patsubst %/, .dbs/%.rec, $(categories))
+
+default += $(databases)
+
+$(databases): .dbs/%.rec: %/ | .dbs/
+	@mkdir -p $(@D)
+	@for entry in $(shell find $< -type f -name "*.md") ; do \
+		sed -n '2,/^---$$/ {/^---$$/d; p}' "$$entry" |\
+		sed -e 's/\[ //'  -e 's/ \]//' |\
+		tr -d '"' ;\
+		printf "wordcount: %s\n" "$$(wc -w < $$entry)" ;\
+		printf "file: %s\n\n" "$$entry" ;\
+	done >> $@
+	for entry in $(shell find $< -type f -name "*.md"); do \
+		recset $@ -e "file = '$${entry}'" -f wordcount --set-add="$$(wc -w < $${entry})" ;\
+	done
+
+db.rec: $(databases)
 	printf '%s\n' '%rec: guide' > $@
 	printf '%s\n' '%type: wordcount int' >> $@
 	printf '%s\n\n' '%sort: title' >> $@
-	for x in $^ ; do \
-		sed -n '2,/^---$$/ {/^---$$/d; p}' "$$x" |\
-		sed -e 's/\[ //'  -e 's/ \]//' |\
-		tr -d '"' ;\
-		printf "file: %s\n\n" "$$x" ;\
-	done >> $@
-	for entry in $^; do \
-		recset $@ -e "file = '$${entry}'" -f wordcount --set-add="$$(wc -w < $${entry})" ;\
-	done
+	cat $^ >> $@
 	recsel $@ -e "requires != ''" -CR title,requires |\
 	while read title requires; do \
 		IFS=', ' && for provider in $$requires; do \