From 4f429d6b86f16c1baafc5a29caf919e19c223935 Mon Sep 17 00:00:00 2001
From: Philipp Strube <philipp.strube@container-solutions.com>
Date: Wed, 6 Mar 2019 19:13:38 +0100
Subject: [PATCH] Reduce time required for cloning remote bases

This commit changes git/cloner.go from cloning the whole history
and then checking out the desired ref to a implementation that
only downloads the history for the desired ref.

It does so by first initializing an empty repository, setting the
source repository as a remote, fetching just the desired ref and
then hard resetting the empty local repo to that ref.

This reduces the time it takes to build the multibases example
as a remote base at ref v2.0.3 from an avg of 8s with the
current implementation to an avg of 2s out of 10 runs each, by
drastically decreasing the data transferred.

The improvement should increase as repositories grow.
---
 pkg/git/cloner.go | 53 ++++++++++++++++++++++++++++++++++++++++-------
 1 file changed, 45 insertions(+), 8 deletions(-)

diff --git a/pkg/git/cloner.go b/pkg/git/cloner.go
index 465fdb1d1..cbf176448 100644
--- a/pkg/git/cloner.go
+++ b/pkg/git/cloner.go
@@ -41,24 +41,61 @@ func ClonerUsingGitExec(repoSpec *RepoSpec) error {
 	}
 	cmd := exec.Command(
 		gitProgram,
-		"clone",
-		repoSpec.CloneSpec(),
+		"init",
 		repoSpec.cloneDir.String())
 	var out bytes.Buffer
 	cmd.Stdout = &out
 	err = cmd.Run()
 	if err != nil {
-		return errors.Wrapf(err, "trouble cloning %s", repoSpec.raw)
+		return errors.Wrapf(
+			err,
+			"trouble initializing empty git repo in %s",
+			repoSpec.cloneDir.String())
 	}
-	if repoSpec.ref == "" {
-		return nil
-	}
-	cmd = exec.Command(gitProgram, "checkout", repoSpec.ref)
+
+	cmd = exec.Command(
+		gitProgram,
+		"remote",
+		"add",
+		"origin",
+		repoSpec.CloneSpec())
+	cmd.Stdout = &out
 	cmd.Dir = repoSpec.cloneDir.String()
 	err = cmd.Run()
 	if err != nil {
 		return errors.Wrapf(
-			err, "trouble checking out href %s", repoSpec.ref)
+			err,
+			"trouble adding remote %s",
+			repoSpec.CloneSpec())
+	}
+
+	if repoSpec.ref == "" {
+		return nil
+	}
+	cmd = exec.Command(
+		gitProgram,
+		"fetch",
+		"--depth=1",
+		"origin",
+		repoSpec.ref)
+	cmd.Stdout = &out
+	cmd.Dir = repoSpec.cloneDir.String()
+	err = cmd.Run()
+	if err != nil {
+		return errors.Wrapf(err, "trouble fetching %s", repoSpec.ref)
+	}
+
+	cmd = exec.Command(
+		gitProgram,
+		"reset",
+		"--hard",
+		"FETCH_HEAD")
+	cmd.Stdout = &out
+	cmd.Dir = repoSpec.cloneDir.String()
+	err = cmd.Run()
+	if err != nil {
+		return errors.Wrapf(
+			err, "trouble hard resetting empty repository to %s", repoSpec.ref)
 	}
 	return nil
 }