From 41f53c88dff7c35231d1e004e420a87d807f2a90 Mon Sep 17 00:00:00 2001
From: Rob Watson <rob@netflux.io>
Date: Tue, 5 Jul 2022 18:26:47 +0200
Subject: [PATCH] podWatcher: prefer bufio.Reader to Scanner

This avoids "token too large" errors when scanning large lines, probably
at the cost of some memory usage and/or allocations.
---
 logs/pod_watcher.go  | 23 ++++++++++++++++-------
 logs/watcher_test.go |  1 +
 2 files changed, 17 insertions(+), 7 deletions(-)

diff --git a/logs/pod_watcher.go b/logs/pod_watcher.go
index 75fa1c5..596b975 100644
--- a/logs/pod_watcher.go
+++ b/logs/pod_watcher.go
@@ -225,13 +225,22 @@ func copyLogStream(ctx context.Context, wg *sync.WaitGroup, client KubernetesCli
 	errch := make(chan error, 1)
 	go func() {
 		defer close(errch)
-		scanner := bufio.NewScanner(logReader)
-		for scanner.Scan() {
-			logsLines <- "[" + stream.podName + "] " + scanner.Text() + nl
-		}
-		if err := scanner.Err(); err != nil {
-			errch <- newLogError(fmt.Errorf("error scanning logs: %v", err), stream)
-			return
+		r := bufio.NewReader(logReader)
+		for {
+			lineBytes, err := r.ReadBytes('\n')
+			line := string(lineBytes)
+			if len(line) > 0 && !strings.HasSuffix(line, "\n") {
+				line += "\n"
+			}
+			if len(line) > 0 {
+				logsLines <- "[" + stream.podName + "] " + line
+			}
+			if err != nil {
+				if err != io.EOF {
+					errch <- newLogError(fmt.Errorf("error reading logs: %v", err), stream)
+				}
+				return
+			}
 		}
 	}()
 
diff --git a/logs/watcher_test.go b/logs/watcher_test.go
index ad2492d..0f1177e 100644
--- a/logs/watcher_test.go
+++ b/logs/watcher_test.go
@@ -164,6 +164,7 @@ func TestWatcherWithPodWatcher(t *testing.T) {
 	assert.ElementsMatch(t, []string{"[foo] fake logs", "[bar] fake logs"}, lines)
 }
 
+// bufToLines splits the output buffer removing newline characters.
 func bufToLines(buf *bytes.Buffer) []string {
 	return strings.Split(strings.TrimSpace(buf.String()), "\n")
 }