From cc7854171493b70f355cf8e3ae0fc2a488966de8 Mon Sep 17 00:00:00 2001
From: Jakob Ackermann <jakob.ackermann@overleaf.com>
Date: Wed, 3 Nov 2021 12:04:29 +0100
Subject: [PATCH] Merge pull request #5670 from overleaf/jpa-cdn-pre-compress

[web] cdn_upload: compress assets prior to uploading them to GCS

GitOrigin-RevId: a9b0970beb124d20bd2ffe21d30a674ffafd6258
---
 services/web/bin/cdn_upload               | 41 +++++----------------
 services/web/bin/cdn_upload_batch         | 44 +++++++++++++++++++++++
 services/web/bin/compress_assets          | 14 ++++++++
 services/web/bin/compress_batch_of_assets | 16 +++++++++
 4 files changed, 83 insertions(+), 32 deletions(-)
 create mode 100755 services/web/bin/cdn_upload_batch
 create mode 100755 services/web/bin/compress_assets
 create mode 100755 services/web/bin/compress_batch_of_assets

diff --git a/services/web/bin/cdn_upload b/services/web/bin/cdn_upload
index cdd3df2f50..00b9eaa48d 100755
--- a/services/web/bin/cdn_upload
+++ b/services/web/bin/cdn_upload
@@ -1,47 +1,22 @@
 #!/bin/bash
 set -e
 
-function upload_with_content_type() {
-  content_type=$1
-  bucket=$2
-  shift 2
-  content_type_options=""
-  if [[ "$content_type" != "-" ]]; then
-    content_type_options="-h Content-Type:${content_type};charset=utf-8"
-  fi
-
-  # DOCS for gsutil -- it does not have long command line flags!
-  ## global flags
-  # -h NAME:VALUE   add header, can occur multiples times
-  # -m              upload with multiple threads
-  ## rsync flags
-  # -r              traverse into directories recursively
-  # -x              Python regex for excluding files from the sync
-  gsutil \
-    -h "Cache-Control:public, max-age=31536000" \
-    ${content_type_options} \
-    -m \
-    rsync \
-    -r \
-    "$@" \
-    "/tmp/public/" \
-    "${bucket}/public/"
-}
 function upload_into_bucket() {
   bucket=$1
 
   # stylesheets
-  upload_with_content_type 'text/css' "$bucket" \
-    -x '.+(?<!\.css)$'
+  bin/cdn_upload_batch 'text/css' "$bucket" '.css' \
+    -x '.+(?<!\.css)$' &
 
   # javascript files
-  upload_with_content_type 'application/javascript' "$bucket" \
-    -x '.+(?<!\.js)$'
+  bin/cdn_upload_batch 'application/javascript' "$bucket" '.js' \
+    -x '.+(?<!\.js)$' &
 
   # the rest
-  upload_with_content_type '-' "$bucket" \
-    -x '.+\.(css|js)$'
+  bin/cdn_upload_batch '-' "$bucket" '-' \
+    -x '.+\.(css|js)$' &
 
+  wait
 }
 
 # Upload to staging CDN if branch is either 'master' or 'staging-master' or main variants
@@ -51,6 +26,8 @@ if [[ "$BRANCH_NAME" == "master" || "$BRANCH_NAME" == "staging-master" || "$BRAN
   # delete source maps
   find /tmp/public -name '*.js.map' -delete
 
+  bin/compress_assets
+
   upload_into_bucket $CDN_STAG
   # Only upload to production CDN if branch is
   if [[ "$BRANCH_NAME" == "master" || "$BRANCH_NAME" == "main" ]]; then
diff --git a/services/web/bin/cdn_upload_batch b/services/web/bin/cdn_upload_batch
new file mode 100755
index 0000000000..2ed2f3358b
--- /dev/null
+++ b/services/web/bin/cdn_upload_batch
@@ -0,0 +1,44 @@
+#!/bin/bash
+set -e
+
+content_type=$1
+bucket=$2
+text_extension=$3
+shift 3
+content_type_options=""
+if [[ "$content_type" != "-" ]]; then
+  content_type_options="-h Content-Type:${content_type};charset=utf-8"
+fi
+
+# DOCS for gsutil -- it does not have long command line flags!
+## global flags
+# -h NAME:VALUE   add header, can occur multiples times
+# -m              upload with multiple threads
+## rsync flags
+# -c              use checksums for determining changed files (mtime is not stable)
+# -r              traverse into directories recursively
+# -x              Python regex for excluding files from the sync
+if [[ "$text_extension" == "-" || $(find /tmp/public -type f -name "*$text_extension" | wc -l) != "0" ]]; then
+  # Potentially skip upload of non-compressed .js/.css files.
+  gsutil \
+    -h "Cache-Control:public, max-age=31536000" \
+    ${content_type_options} \
+    -m \
+    rsync \
+    -c \
+    -r \
+    "$@" \
+    "/tmp/public/" \
+    "${bucket}/public/"
+fi
+gsutil \
+  -h "Cache-Control:public, max-age=31536000" \
+  -h "Content-Encoding:gzip" \
+  ${content_type_options} \
+  -m \
+  rsync \
+  -c \
+  -r \
+  "$@" \
+  "/tmp/compressed/public/" \
+  "${bucket}/public/"
diff --git a/services/web/bin/compress_assets b/services/web/bin/compress_assets
new file mode 100755
index 0000000000..e50f0a3389
--- /dev/null
+++ b/services/web/bin/compress_assets
@@ -0,0 +1,14 @@
+#!/usr/bin/env bash
+set -e -o pipefail
+
+SCRIPT_PATH=$(realpath "${BASH_SOURCE[0]}")
+SCRIPT_DIR=$(dirname "$SCRIPT_PATH")
+
+pushd /tmp
+
+find public/ -type d | sed 's!^!compressed/!' | xargs mkdir --parents
+
+find public/ -type f -print0 \
+| xargs --null --max-args 20 --max-procs 10 "$SCRIPT_DIR/compress_batch_of_assets"
+
+popd
diff --git a/services/web/bin/compress_batch_of_assets b/services/web/bin/compress_batch_of_assets
new file mode 100755
index 0000000000..4a650da4b8
--- /dev/null
+++ b/services/web/bin/compress_batch_of_assets
@@ -0,0 +1,16 @@
+#!/usr/bin/env bash
+set -e
+
+for file in "$@"; do
+  file_gzipped="compressed/$file"
+
+  gzip -9 --no-name --stdout "$file" > "$file_gzipped"
+
+  before=$(stat -c%s "$file")
+  after=$(stat -c%s "$file_gzipped")
+  if [[ "$after" -ge "$before" ]]; then
+    rm "$file_gzipped"
+  else
+    rm "$file"
+  fi
+done