sfeed_update: revert new merge logic

this requires more work without breaking the order in sfeed_html and other
tools (top to bottom: new to oldest), vs sfeed_plain in tail mode: oldest to
newest.

There will also be improvements to the merge logic to reduce many writes in
the future.

2 files changed, 4 insertions, 9 deletions

diff --git a/README b/README
index a76641f..b5bbd03 100644
--- a/README
+++ b/README
@@ -74,7 +74,6 @@ Optional dependencies
 - make(1) (for Makefile).
 - POSIX sh(1),
   used by sfeed_update(1) and sfeed_opml_export(1).
-- awk(1), used by sfeed_update(1).
 - curl(1) binary: http://curl.haxx.se/ ,
   used by sfeed_update(1), can be replaced with any tool like wget(1),
   OpenBSD ftp(1).
diff --git a/sfeed_update b/sfeed_update
index 2977bdc..5e265a7 100755
--- a/sfeed_update
+++ b/sfeed_update
@@ -31,14 +31,10 @@ loadconfig() {
 # merge raw files.
 # merge(oldfile, newfile)
 merge() {
-	# unique check by id, title, link.
-	# print only new entries in newfile.
-	# order new items by timestamp (asc).
-	(sed 's@^@O	@' "$1"
-	sed 's@^@N	@' "$2") | \
-		LC_ALL=C awk '!x[$7 "	" $3 "	" $4]++ && $1 == "N"' 2>/dev/null | \
-		cut -f 2- | \
-		sort -t '	' -k1n,1
+	# unique sort by id, title, link.
+	# order by timestamp (desc).
+	(sort -t '	' -u -k6,6 -k2,2 -k3,3 "$1" "$2" 2>/dev/null) |
+	sort -t '	' -k1rn,1
 }
 
 # fetch a feed via HTTP/HTTPS etc.