From 2bfb407cef88281c59765e2a34031db78f8fd363 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Rapha=C3=ABl=20Bournhonesque?= <raphael@bournhonesque.eu>
Date: Fri, 14 Apr 2023 15:22:39 +0700
Subject: [PATCH] fix: improve Google Cloud OCR processing

- gzip all JSON OCRs when saving OCR file on disk
- add new `created_at` field to save the timestamp of OCR generation
---
 lib/ProductOpener/Images.pm                   | 20 +++++----
 lib/ProductOpener/Test.pm                     | 44 +++++++++++++++++++
 scripts/run_cloud_vision_ocr.pl               |  2 +-
 stop_words.txt                                |  1 +
 .../run_cloud_vision_ocr/ocr_data.json        |  3 --
 tests/integration/run_cloud_vision_ocr.t      |  6 +--
 .../send_image_to_cloud_vision/ocr_data.json  |  3 --
 .../ocr_data_2.json                           |  3 --
 .../ocr_data_3.json                           |  3 --
 tests/unit/send_image_to_cloud_vision.t       | 31 +++++++------
 10 files changed, 78 insertions(+), 38 deletions(-)
 delete mode 100644 tests/integration/expected_test_results/run_cloud_vision_ocr/ocr_data.json
 delete mode 100644 tests/unit/expected_test_results/send_image_to_cloud_vision/ocr_data.json
 delete mode 100644 tests/unit/expected_test_results/send_image_to_cloud_vision/ocr_data_2.json
 delete mode 100644 tests/unit/expected_test_results/send_image_to_cloud_vision/ocr_data_3.json

diff --git a/lib/ProductOpener/Images.pm b/lib/ProductOpener/Images.pm
index 7d9dcb9f31c80..54d4f9e4e2b0f 100644
--- a/lib/ProductOpener/Images.pm
+++ b/lib/ProductOpener/Images.pm
@@ -140,6 +140,7 @@ use ProductOpener::URL qw/:all/;
 use ProductOpener::Users qw/:all/;
 use ProductOpener::Text qw/:all/;
 
+use IO::Compress::Gzip qw(gzip $GzipError);
 use Log::Any qw($log);
 use Encode;
 use JSON::PP;
@@ -2002,7 +2003,7 @@ sub extract_text_from_image ($product_ref, $id, $field, $ocr_engine, $results_re
 	}
 	elsif ($ocr_engine eq 'google_cloud_vision') {
 
-		my $json_file = "$www_root/images/products/$path/$filename.json";
+		my $json_file = "$www_root/images/products/$path/$filename.json.gz";
 		open(my $gv_logs, ">>:encoding(UTF-8)", "$data_root/logs/cloud_vision.log");
 		my $cloudvision_ref = send_image_to_cloud_vision($image, $json_file, \@CLOUD_VISION_FEATURES_TEXT, $gv_logs);
 		close $gv_logs;
@@ -2047,7 +2048,7 @@ Call to Google Cloud vision API
 
 =head4 $image_path - str path to image
 
-=head4 $json_file - str path to the file where we will store OCR result as JSON
+=head4 $json_file - str path to the file where we will store OCR result as gzipped JSON
 
 =head4 $features_ref - hash reference - the "features" parameter of Google Cloud Vision
 
@@ -2108,14 +2109,17 @@ sub send_image_to_cloud_vision ($image_path, $json_file, $features_ref, $gv_logs
 
 		$cloudvision_ref = decode_json($json_response);
 
-		$log->info("saving google cloud vision json response to file", {path => $json_file}) if $log->is_info();
+		# Adding creation timestamp, to know when the OCR has been generated
+		$cloudvision_ref->{created_at} = time();
 
-		# UTF-8 issue , see https://stackoverflow.com/questions/4572007/perl-lwpuseragent-mishandling-utf-8-response
-		$json_response = decode("utf8", $json_response);
+		$log->info("saving google cloud vision json response to file", {path => $json_file}) if $log->is_info();
 
-		if (open(my $OUT, ">:encoding(UTF-8)", $json_file)) {
-			print($OUT $json_response);
-			close($OUT);
+		if (open(my $OUT, ">:raw", $json_file)) {
+			my $gzip_handle = IO::Compress::Gzip->new($OUT)
+				or die "Cannot create gzip filehandle: $GzipError\n";
+			my $encoded_json = encode_json($cloudvision_ref);
+			$gzip_handle->print($encoded_json);
+			$gzip_handle->close;
 
 			print($gv_logs "--> cloud vision success for $image_path\n");
 		}
diff --git a/lib/ProductOpener/Test.pm b/lib/ProductOpener/Test.pm
index ffc688b751f4f..b9e15ebd910d5 100644
--- a/lib/ProductOpener/Test.pm
+++ b/lib/ProductOpener/Test.pm
@@ -51,6 +51,8 @@ BEGIN {
 		&remove_all_orgs
 		&check_not_production
 		&wait_for
+		&read_gzip_file
+		&check_ocr_result
 	);    # symbols to export on request
 	%EXPORT_TAGS = (all => [@EXPORT_OK]);
 }
@@ -66,15 +68,57 @@ use ProductOpener::Store "store";
 use Carp qw/confess/;
 use Data::DeepAccess qw(deep_exists deep_get deep_set);
 use Getopt::Long;
+use IO::Uncompress::AnyInflate qw(anyinflate $AnyInflateError);
 use Test::More;
 use JSON "decode_json";
 use File::Basename "fileparse";
 use File::Path qw/make_path remove_tree/;
 use File::Copy;
 use Path::Tiny qw/path/;
+use Scalar::Util qw(looks_like_number);
 
 use Log::Any qw($log);
 
+=head2 read_gzip_file($filepath)
+
+Read gzipped file and return binary content
+
+=head3 Parameters
+
+=head4 String $filepath
+The path of the gzipped file.
+
+=cut
+
+sub read_gzip_file ($filepath) {
+	my $input = IO::File->new($filepath) or die "Cannot open '$filepath'\n";
+	my $buffer;
+	anyinflate $input => \$buffer or die "anyinflate failed: $AnyInflateError\n";
+	return $buffer;
+}
+
+=head2 check_ocr_result($ocr_result)
+
+Check that OCR result returned by Google Cloud Vision is as expected:
+  - a single [response] object in `responses` field
+  - `created_at` integer field
+
+=head3 Parameters
+
+=head4 String $ocr_result
+String of OCR result JSON as returned by Google Cloud Vision.
+
+=cut
+
+sub check_ocr_result ($ocr_result) {
+	ok(defined $ocr_result->{responses}, "OCR result contains the 'responses' field");
+	my @responses = $ocr_result->{responses};
+	my $created_at = $ocr_result->{created_at};
+	is(scalar @responses, 1, "OCR result contains a single response");
+	ok((defined $created_at and looks_like_number($created_at)), "OCR result `created_at` field is valid, $created_at");
+	return;
+}
+
 =head2 init_expected_results($filepath)
 
 Handles test options around expected_results initialization
diff --git a/scripts/run_cloud_vision_ocr.pl b/scripts/run_cloud_vision_ocr.pl
index 68376a3a952f4..1c5bb7800d1a1 100755
--- a/scripts/run_cloud_vision_ocr.pl
+++ b/scripts/run_cloud_vision_ocr.pl
@@ -69,7 +69,7 @@ ($file)
 
 	my $json_file = $destination;
 	$json_file =~ s/\.([^\.]+)$//;
-	$json_file .= ".json";
+	$json_file .= ".json.gz";
 
 	print $LOG "file: $file destination: $destination code: $code image_url: $image_url json_file: $json_file\n";
 	open(my $gv_logs, ">>:encoding(UTF-8)", "$data_root/logs/cloud_vision.log");
diff --git a/stop_words.txt b/stop_words.txt
index f6d991ceb7aad..18fa94991cd0e 100644
--- a/stop_words.txt
+++ b/stop_words.txt
@@ -217,3 +217,4 @@ weigher
 weighers
 www
 xml
+gzipped
\ No newline at end of file
diff --git a/tests/integration/expected_test_results/run_cloud_vision_ocr/ocr_data.json b/tests/integration/expected_test_results/run_cloud_vision_ocr/ocr_data.json
deleted file mode 100644
index 1f8d1d822dd56..0000000000000
--- a/tests/integration/expected_test_results/run_cloud_vision_ocr/ocr_data.json
+++ /dev/null
@@ -1,3 +0,0 @@
-{
-   "ocr" : "success"
-}
diff --git a/tests/integration/run_cloud_vision_ocr.t b/tests/integration/run_cloud_vision_ocr.t
index 35b732caf9ab4..8fa22cfd5e2bc 100644
--- a/tests/integration/run_cloud_vision_ocr.t
+++ b/tests/integration/run_cloud_vision_ocr.t
@@ -39,7 +39,7 @@ dircopy("$sample_products_images_path/$product_code_path", $image_dir);
 fcopy($input_image_path, "$image_dir/2.jpg");
 # fake responses for OCR and robtoff
 my @responses = (
-	HTTP::Response->new("200", "OK", HTTP::Headers->new(), '{"ocr": "success"}'),
+	HTTP::Response->new("200", "OK", HTTP::Headers->new(), '{"responses": [{}]}'),
 	HTTP::Response->new("200", "OK", HTTP::Headers->new(), '{"robotoff": "success"}'),
 );
 my $dump_path = File::Temp->newdir();
@@ -56,10 +56,10 @@ is(scalar @requests, 2, "Two request issued");
 my $ocr_request = retrieve("$dump_path/req-0.sto");
 my $request_json_body = decode_json($ocr_request->content());
 compare_to_expected_results($request_json_body, "$expected_result_dir/ocr_request_body.json", $update_expected_results);
-my $ocr_content = read_file("$image_dir/2.json");
+my $ocr_content = read_gzip_file("$image_dir/2.json.gz");
 ok($ocr_content, "OCR file is not empty");
 my $ocr_data = decode_json($ocr_content);
-compare_to_expected_results($ocr_data, "$expected_result_dir/ocr_data.json", $update_expected_results);
+check_ocr_result($ocr_data);
 my $robotoff_request = retrieve("$dump_path/req-1.sto");
 # we have url encoded parameters, and order might change --> convert to hash
 my $request_content = url_params_mixed($robotoff_request->content());
diff --git a/tests/unit/expected_test_results/send_image_to_cloud_vision/ocr_data.json b/tests/unit/expected_test_results/send_image_to_cloud_vision/ocr_data.json
deleted file mode 100644
index 07046c2aa314e..0000000000000
--- a/tests/unit/expected_test_results/send_image_to_cloud_vision/ocr_data.json
+++ /dev/null
@@ -1,3 +0,0 @@
-{
-   "foo" : "blah"
-}
diff --git a/tests/unit/expected_test_results/send_image_to_cloud_vision/ocr_data_2.json b/tests/unit/expected_test_results/send_image_to_cloud_vision/ocr_data_2.json
deleted file mode 100644
index 8dd659ba8030f..0000000000000
--- a/tests/unit/expected_test_results/send_image_to_cloud_vision/ocr_data_2.json
+++ /dev/null
@@ -1,3 +0,0 @@
-{
-   "foo" : "bar"
-}
diff --git a/tests/unit/expected_test_results/send_image_to_cloud_vision/ocr_data_3.json b/tests/unit/expected_test_results/send_image_to_cloud_vision/ocr_data_3.json
deleted file mode 100644
index 8dd659ba8030f..0000000000000
--- a/tests/unit/expected_test_results/send_image_to_cloud_vision/ocr_data_3.json
+++ /dev/null
@@ -1,3 +0,0 @@
-{
-   "foo" : "bar"
-}
diff --git a/tests/unit/send_image_to_cloud_vision.t b/tests/unit/send_image_to_cloud_vision.t
index 4f483c823d176..45be363b60ccb 100644
--- a/tests/unit/send_image_to_cloud_vision.t
+++ b/tests/unit/send_image_to_cloud_vision.t
@@ -14,6 +14,9 @@ use ProductOpener::Images qw/:all/;
 
 my ($test_id, $test_dir, $expected_result_dir, $update_expected_results) = (init_expected_results(__FILE__));
 
+# Default OCR response, containing a single response element
+my $ocr_default_response = '{"responses": [{}]}';
+
 my @ua_requests = ();
 # put responses for call to requests here, we will pop first
 my @ua_responses = ();
@@ -35,9 +38,9 @@ my $image_path = dirname(__FILE__) . "/inputs/small-img.jpg";
 
 	# normal test
 	open(my $gv_logs, ">:encoding(UTF-8)", $gv_logs_path);
-	my $json_path = $tmp_dir . "/small-img.json";
+	my $json_path = $tmp_dir . "/small-img.json.gz";
 	# expected response
-	my $response = HTTP::Response->new("200", "OK", HTTP::Headers->new(), '{"foo": "blah"}');
+	my $response = HTTP::Response->new("200", "OK", HTTP::Headers->new(), $ocr_default_response);
 	push @ua_responses, $response;
 	send_image_to_cloud_vision($image_path, $json_path, \@CLOUD_VISION_FEATURES_FULL, $gv_logs);
 	close($gv_logs);
@@ -45,16 +48,16 @@ my $image_path = dirname(__FILE__) . "/inputs/small-img.jpg";
 	my $issued_request = shift @ua_requests;
 	my $request_json_body = decode_json($issued_request->content());
 	compare_to_expected_results($request_json_body, "$expected_result_dir/request_body.json", $update_expected_results);
-	my $ocr_content = read_file($json_path);
+	my $ocr_content = read_gzip_file($json_path);
 	ok($ocr_content, "normal test - OCR file is not empty");
 	my $ocr_data = decode_json($ocr_content);
-	compare_to_expected_results($ocr_data, "$expected_result_dir/ocr_data.json", $update_expected_results);
+	check_ocr_result($ocr_data);
 	my $logs = read_file($gv_logs_path);
 	like($logs, qr/cloud vision success/, "normal test - cloud vision success in logs");
 
 	# test new request updates
 	open($gv_logs, ">:encoding(UTF-8)", $gv_logs_path);
-	$response = HTTP::Response->new("200", "OK", HTTP::Headers->new(), '{"foo": "bar"}');
+	$response = HTTP::Response->new("200", "OK", HTTP::Headers->new(), $ocr_default_response);
 	push @ua_responses, $response;
 	send_image_to_cloud_vision($image_path, $json_path, \@CLOUD_VISION_FEATURES_FULL, $gv_logs);
 	close($gv_logs);
@@ -63,15 +66,15 @@ my $image_path = dirname(__FILE__) . "/inputs/small-img.jpg";
 	$request_json_body = decode_json($issued_request->content());
 	compare_to_expected_results($request_json_body, "$expected_result_dir/request_body_2.json",
 		$update_expected_results);
-	$ocr_content = read_file($json_path);
+	$ocr_content = read_gzip_file($json_path);
 	$ocr_data = decode_json($ocr_content);
-	compare_to_expected_results($ocr_data, "$expected_result_dir/ocr_data_2.json", $update_expected_results);
+	check_ocr_result($ocr_data);
 	$logs = read_file($gv_logs_path);
 	like($logs, qr/cloud vision success/, "test request update - cloud vision success in logs");
 
 	# test with different feature set \@CLOUD_VISION_FEATURES_TEXT
 	open($gv_logs, ">:encoding(UTF-8)", $gv_logs_path);
-	$response = HTTP::Response->new("200", "OK", HTTP::Headers->new(), '{"foo": "bar"}');
+	$response = HTTP::Response->new("200", "OK", HTTP::Headers->new(), $ocr_default_response);
 	push @ua_responses, $response;
 	send_image_to_cloud_vision($image_path, $json_path, \@CLOUD_VISION_FEATURES_TEXT, $gv_logs);
 	close($gv_logs);
@@ -80,19 +83,19 @@ my $image_path = dirname(__FILE__) . "/inputs/small-img.jpg";
 	$request_json_body = decode_json($issued_request->content());
 	compare_to_expected_results($request_json_body, "$expected_result_dir/request_body_3.json",
 		$update_expected_results);
-	$ocr_content = read_file($json_path);
+	$ocr_content = read_gzip_file($json_path);
 	$ocr_data = decode_json($ocr_content);
-	compare_to_expected_results($ocr_data, "$expected_result_dir/ocr_data_3.json", $update_expected_results);
+	check_ocr_result($ocr_data);
 	$logs = read_file($gv_logs_path);
 	like($logs, qr/cloud vision success/, "test request features text - cloud vision success in logs");
 
 	# test with bad json path
 	open($gv_logs, ">:encoding(UTF-8)", $gv_logs_path);
-	$response = HTTP::Response->new("200", "OK", HTTP::Headers->new(), '{"foo": "blah"}');
+	$response = HTTP::Response->new("200", "OK", HTTP::Headers->new(), $ocr_default_response);
 	push @ua_responses, $response;
 	send_image_to_cloud_vision(
 		$image_path,
-		"/var/lib/not-a-directory/not-writable.json",
+		"/var/lib/not-a-directory/not-writable.json.gz",
 		\@CLOUD_VISION_FEATURES_FULL, $gv_logs
 	);
 	close($gv_logs);
@@ -105,8 +108,8 @@ my $image_path = dirname(__FILE__) . "/inputs/small-img.jpg";
 
 	# test bad request
 	open($gv_logs, ">:encoding(UTF-8)", $gv_logs_path);
-	$json_path = $tmp_dir . "/small-img2.json";
-	$response = HTTP::Response->new("403", "Not authorized", HTTP::Headers->new(), '{"foo": "blah"}');
+	$json_path = $tmp_dir . "/small-img2.json.gz";
+	$response = HTTP::Response->new("403", "Not authorized", HTTP::Headers->new(), $ocr_default_response);
 	push @ua_responses, $response;
 	send_image_to_cloud_vision($image_path, $json_path, \@CLOUD_VISION_FEATURES_FULL, $gv_logs);
 	close($gv_logs);