[PATCH] fix scalability problems with git-deltafy-script (fwd)

From: Nicolas Pitre <nico@cam.org>
Date: Jun 21, 2005, 14:18 UTC
Message-ID: <Pine.LNX.4.63.0506211008380.1667@localhost.localdomain>

[ Third resent... ]

Current version would spin forever and exhaust memory while attempting to sort all files from all revisions at once, until it dies before even doing any real work. This is especially noticeable when used on a big repository like the imported bkcvs repo for the Linux kernel.

This patch allows for batching the sort to put a bound on needed resources and making progress early, as well as including some small cleanups.

Signed-off-by: Nicolas Pitre <nico@cam.org>

diff --git a/git-deltafy-script b/git-deltafy-script
--- a/git-deltafy-script
+++ b/git-deltafy-script
@@ -1,6 +1,6 @@
 #!/bin/bash
 
-# Example script to deltafy an entire GIT repository based on the commit list.
+# Example script to deltify an entire GIT repository based on the commit list.
 # The most recent version of a file is the reference and previous versions
 # are made delta against the best earlier version available. And so on for
 # successive versions going back in time.  This way the increasing delta
@@ -25,37 +25,51 @@
 
 set -e
 
-depth=
-[ "$1" == "-d" ] && depth="--max-depth=$2" && shift 2
+max_depth=
+[ "$1" == "-d" ] && max_depth="--max-depth=$2" && shift 2
+
+overlap=30
+max_behind="--max-behind=$overlap"
 
 function process_list() {
 	if [ "$list" ]; then
 		echo "Processing $curr_file"
-		echo "$head $list" | xargs git-mkdelta $depth --max-behind=30 -v
+		echo "$list" | xargs git-mkdelta $max_depth $max_behind -v
 	fi
 }
 
+rev_list=""
 curr_file=""
 
 git-rev-list HEAD |
-git-diff-tree -r -t --stdin |
-awk '/^:/ { if ($5 == "M" || $5 == "N") print $4, $6;
-            if ($5 == "M") print $3, $6 }' |
-LC_ALL=C sort -s -k 2 | uniq |
-while read sha1 file; do
-	if [ "$file" == "$curr_file" ]; then
-		list="$list $sha1"
-	else
-		process_list
-		curr_file="$file"
-		list=""
-		head="$sha1"
-	fi
+while true; do
+	# Let's batch revisions into groups of 1000 to give it a chance to
+	# scale with repositories containing long revision lists.  We also
+	# overlap with the previous batch the size of mkdelta's look behind
+	# value in order to account for the processing discontinuity.
+	rev_list="$(echo -e -n "$rev_list" | tail --lines=$overlap)"
+	for i in $(seq 1000); do
+		read rev || break
+		rev_list="$rev_list$rev\n"
+	done
+	echo -e -n "$rev_list" |
+	git-diff-tree -r -t --stdin |
+	awk '/^:/ { if ($5 == "M") printf "%s %s\n%s %s\n", $4, $6, $3, $6 }' |
+	LC_ALL=C sort -s -k 2 | uniq |
+	while read sha1 file; do
+		if [ "$file" == "$curr_file" ]; then
+			list="$list $sha1"
+		else
+			process_list
+			curr_file="$file"
+			list="$sha1"
+		fi
+	done
+	[ "$rev" ] || break
 done
 process_list
 
 curr_file="root directory"
-head=""
 list="$(
 	git-rev-list HEAD |
 	while read commit; do