GDScript: Enable compression on export

Besides the regular option to export GDScript as binary tokens, this
also includes a compression option on top of it. The binary format
needs to encode some information which generally makes it bigger than
the source text. This option reduces that difference by using Zstandard
compression on the buffer.

editor/export/editor_export.cpp

@@ -270,7 +270,7 @@ void EditorExport::load_config() {
 		preset->set_include_filter(config->get_value(section, "include_filter"));
 		preset->set_exclude_filter(config->get_value(section, "exclude_filter"));
 		preset->set_export_path(config->get_value(section, "export_path", ""));
-		preset->set_script_export_mode(config->get_value(section, "script_export_mode", EditorExportPreset::MODE_SCRIPT_BINARY_TOKENS));
+		preset->set_script_export_mode(config->get_value(section, "script_export_mode", EditorExportPreset::MODE_SCRIPT_BINARY_TOKENS_COMPRESSED));
 
 		if (config->has_section_key(section, "encrypt_pck")) {
 			preset->set_enc_pck(config->get_value(section, "encrypt_pck"));

editor/export/editor_export_preset.h

@@ -57,6 +57,7 @@ public:
 	enum ScriptExportMode {
 		MODE_SCRIPT_TEXT,
 		MODE_SCRIPT_BINARY_TOKENS,
+		MODE_SCRIPT_BINARY_TOKENS_COMPRESSED,
 	};
 
 private:
@@ -89,7 +90,7 @@ private:
 	bool enc_directory = false;
 
 	String script_key;
-	int script_mode = MODE_SCRIPT_BINARY_TOKENS;
+	int script_mode = MODE_SCRIPT_BINARY_TOKENS_COMPRESSED;
 
 protected:
 	bool _set(const StringName &p_name, const Variant &p_value);

editor/export/project_export.cpp

@@ -1398,7 +1398,9 @@ ProjectExportDialog::ProjectExportDialog() {
 	script_vb->add_margin_child(TTR("GDScript Export Mode:"), script_mode);
 	script_mode->add_item(TTR("Text (easier debugging)"), (int)EditorExportPreset::MODE_SCRIPT_TEXT);
 	script_mode->add_item(TTR("Binary tokens (faster loading)"), (int)EditorExportPreset::MODE_SCRIPT_BINARY_TOKENS);
+	script_mode->add_item(TTR("Compressed binary tokens (smaller files)"), (int)EditorExportPreset::MODE_SCRIPT_BINARY_TOKENS_COMPRESSED);
 	script_mode->connect("item_selected", callable_mp(this, &ProjectExportDialog::_script_export_mode_changed));
+
 	sections->add_child(script_vb);
 
 	sections->connect("tab_changed", callable_mp(this, &ProjectExportDialog::_tab_changed));

modules/gdscript/gdscript.cpp

@@ -1066,7 +1066,7 @@ const Vector<uint8_t> &GDScript::get_binary_tokens_source() const {
 
 Vector<uint8_t> GDScript::get_as_binary_tokens() const {
 	GDScriptTokenizerBuffer tokenizer;
-	return tokenizer.parse_code_string(source);
+	return tokenizer.parse_code_string(source, GDScriptTokenizerBuffer::COMPRESS_NONE);
 }
 
 const HashMap<StringName, GDScriptFunction *> &GDScript::debug_get_member_functions() const {

modules/gdscript/gdscript_parser.cpp

@@ -365,6 +365,7 @@ Error GDScriptParser::parse(const String &p_source_code, const String &p_script_
 	pop_multiline();
 
 	memdelete(text_tokenizer);
+	tokenizer = nullptr;
 
 #ifdef DEBUG_ENABLED
 	if (multiline_stack.size() > 0) {
@@ -384,6 +385,7 @@ Error GDScriptParser::parse_binary(const Vector<uint8_t> &p_binary, const String
 	Error err = buffer_tokenizer->set_code_buffer(p_binary);
 
 	if (err) {
+		memdelete(buffer_tokenizer);
 		return err;
 	}
 
@@ -404,6 +406,7 @@ Error GDScriptParser::parse_binary(const Vector<uint8_t> &p_binary, const String
 	pop_multiline();
 
 	memdelete(buffer_tokenizer);
+	tokenizer = nullptr;
 
 	if (errors.is_empty()) {
 		return OK;

modules/gdscript/gdscript_tokenizer.cpp

@@ -284,7 +284,7 @@ void GDScriptTokenizerText::push_expression_indented_block() {
 }
 
 void GDScriptTokenizerText::pop_expression_indented_block() {
-	ERR_FAIL_COND(indent_stack_stack.size() == 0);
+	ERR_FAIL_COND(indent_stack_stack.is_empty());
 	indent_stack = indent_stack_stack.back()->get();
 	indent_stack_stack.pop_back();
 }

modules/gdscript/gdscript_tokenizer_buffer.cpp

@@ -30,6 +30,7 @@
 
 #include "gdscript_tokenizer_buffer.h"
 
+#include "core/io/compression.h"
 #include "core/io/marshalls.h"
 
 #define TOKENIZER_VERSION 100
@@ -139,19 +140,31 @@ GDScriptTokenizer::Token GDScriptTokenizerBuffer::_binary_to_token(const uint8_t
 
 Error GDScriptTokenizerBuffer::set_code_buffer(const Vector<uint8_t> &p_buffer) {
 	const uint8_t *buf = p_buffer.ptr();
-	int total_len = p_buffer.size();
+	ERR_FAIL_COND_V(p_buffer.size() < 12 || p_buffer[0] != 'G' || p_buffer[1] != 'D' || p_buffer[2] != 'S' || p_buffer[3] != 'C', ERR_INVALID_DATA);
-	ERR_FAIL_COND_V(p_buffer.size() < 24 || p_buffer[0] != 'G' || p_buffer[1] != 'D' || p_buffer[2] != 'S' || p_buffer[3] != 'C', ERR_INVALID_DATA);
 
 	int version = decode_uint32(&buf[4]);
 	ERR_FAIL_COND_V_MSG(version > TOKENIZER_VERSION, ERR_INVALID_DATA, "Binary GDScript is too recent! Please use a newer engine version.");
 
-	uint32_t identifier_count = decode_uint32(&buf[8]);
+	int decompressed_size = decode_uint32(&buf[8]);
-	uint32_t constant_count = decode_uint32(&buf[12]);
-	uint32_t token_line_count = decode_uint32(&buf[16]);
-	uint32_t token_count = decode_uint32(&buf[20]);
 
-	const uint8_t *b = &buf[24];
+	Vector<uint8_t> contents;
-	total_len -= 24;
+	if (decompressed_size == 0) {
+		contents = p_buffer.slice(12);
+	} else {
+		contents.resize(decompressed_size);
+		int result = Compression::decompress(contents.ptrw(), contents.size(), &buf[12], p_buffer.size() - 12, Compression::MODE_ZSTD);
+		ERR_FAIL_COND_V_MSG(result != decompressed_size, ERR_INVALID_DATA, "Error decompressing GDScript tokenizer buffer.");
+	}
+
+	int total_len = contents.size();
+	buf = contents.ptr();
+	uint32_t identifier_count = decode_uint32(&buf[0]);
+	uint32_t constant_count = decode_uint32(&buf[4]);
+	uint32_t token_line_count = decode_uint32(&buf[8]);
+	uint32_t token_count = decode_uint32(&buf[16]);
+
+	const uint8_t *b = &buf[20];
+	total_len -= 20;
 
 	identifiers.resize(identifier_count);
 	for (uint32_t i = 0; i < identifier_count; i++) {
@@ -226,9 +239,7 @@ Error GDScriptTokenizerBuffer::set_code_buffer(const Vector<uint8_t> &p_buffer)
 	return OK;
 }
 
-Vector<uint8_t> GDScriptTokenizerBuffer::parse_code_string(const String &p_code) {
+Vector<uint8_t> GDScriptTokenizerBuffer::parse_code_string(const String &p_code, CompressMode p_compress_mode) {
-	Vector<uint8_t> buf;
-
 	HashMap<StringName, uint32_t> identifier_map;
 	HashMap<Variant, uint32_t, VariantHasher, VariantComparator> constant_map;
 	Vector<uint8_t> token_buffer;
@@ -280,28 +291,23 @@ Vector<uint8_t> GDScriptTokenizerBuffer::parse_code_string(const String &p_code)
 		}
 	}
 
-	// Save header.
+	Vector<uint8_t> contents;
-	buf.resize(24);
+	contents.resize(20);
-	buf.write[0] = 'G';
+	encode_uint32(identifier_map.size(), &contents.write[0]);
-	buf.write[1] = 'D';
+	encode_uint32(constant_map.size(), &contents.write[4]);
-	buf.write[2] = 'S';
+	encode_uint32(token_lines.size(), &contents.write[8]);
-	buf.write[3] = 'C';
+	encode_uint32(token_counter, &contents.write[16]);
-	encode_uint32(TOKENIZER_VERSION, &buf.write[4]);
-	encode_uint32(identifier_map.size(), &buf.write[8]);
-	encode_uint32(constant_map.size(), &buf.write[12]);
-	encode_uint32(token_lines.size(), &buf.write[16]);
-	encode_uint32(token_counter, &buf.write[20]);
 
-	int buf_pos = 24;
+	int buf_pos = 20;
 
 	// Save identifiers.
 	for (const StringName &id : rev_identifier_map) {
 		String s = id.operator String();
 		int len = s.length();
 
-		buf.resize(buf_pos + (len + 1) * 4);
+		contents.resize(buf_pos + (len + 1) * 4);
 
-		encode_uint32(len, &buf.write[buf_pos]);
+		encode_uint32(len, &contents.write[buf_pos]);
 		buf_pos += 4;
 
 		for (int i = 0; i < len; i++) {
@@ -309,7 +315,7 @@ Vector<uint8_t> GDScriptTokenizerBuffer::parse_code_string(const String &p_code)
 			encode_uint32(s[i], tmp);
 
 			for (int b = 0; b < 4; b++) {
-				buf.write[buf_pos + b] = tmp[b] ^ 0xb6;
+				contents.write[buf_pos + b] = tmp[b] ^ 0xb6;
 			}
 
 			buf_pos += 4;
@@ -322,28 +328,58 @@ Vector<uint8_t> GDScriptTokenizerBuffer::parse_code_string(const String &p_code)
 		// Objects cannot be constant, never encode objects.
 		Error err = encode_variant(v, nullptr, len, false);
 		ERR_FAIL_COND_V_MSG(err != OK, Vector<uint8_t>(), "Error when trying to encode Variant.");
-		buf.resize(buf_pos + len);
+		contents.resize(buf_pos + len);
-		encode_variant(v, &buf.write[buf_pos], len, false);
+		encode_variant(v, &contents.write[buf_pos], len, false);
 		buf_pos += len;
 	}
 
 	// Save lines and columns.
-	buf.resize(buf_pos + token_lines.size() * 16);
+	contents.resize(buf_pos + token_lines.size() * 16);
 	for (const KeyValue<uint32_t, uint32_t> &e : token_lines) {
-		encode_uint32(e.key, &buf.write[buf_pos]);
+		encode_uint32(e.key, &contents.write[buf_pos]);
 		buf_pos += 4;
-		encode_uint32(e.value, &buf.write[buf_pos]);
+		encode_uint32(e.value, &contents.write[buf_pos]);
 		buf_pos += 4;
 	}
 	for (const KeyValue<uint32_t, uint32_t> &e : token_columns) {
-		encode_uint32(e.key, &buf.write[buf_pos]);
+		encode_uint32(e.key, &contents.write[buf_pos]);
 		buf_pos += 4;
-		encode_uint32(e.value, &buf.write[buf_pos]);
+		encode_uint32(e.value, &contents.write[buf_pos]);
 		buf_pos += 4;
 	}
 
 	// Store tokens.
-	buf.append_array(token_buffer);
+	contents.append_array(token_buffer);
+
+	Vector<uint8_t> buf;
+
+	// Save header.
+	buf.resize(12);
+	buf.write[0] = 'G';
+	buf.write[1] = 'D';
+	buf.write[2] = 'S';
+	buf.write[3] = 'C';
+	encode_uint32(TOKENIZER_VERSION, &buf.write[4]);
+
+	switch (p_compress_mode) {
+		case COMPRESS_NONE:
+			encode_uint32(0u, &buf.write[8]);
+			buf.append_array(contents);
+			break;
+
+		case COMPRESS_ZSTD: {
+			encode_uint32(contents.size(), &buf.write[8]);
+			Vector<uint8_t> compressed;
+			int max_size = Compression::get_max_compressed_buffer_size(contents.size(), Compression::MODE_ZSTD);
+			compressed.resize(max_size);
+
+			int compressed_size = Compression::compress(compressed.ptrw(), contents.ptr(), contents.size(), Compression::MODE_ZSTD);
+			ERR_FAIL_COND_V_MSG(compressed_size < 0, Vector<uint8_t>(), "Error compressing GDScript tokenizer buffer.");
+			compressed.resize(compressed_size);
+
+			buf.append_array(compressed);
+		} break;
+	}
 
 	return buf;
 }
@@ -372,7 +408,7 @@ void GDScriptTokenizerBuffer::push_expression_indented_block() {
 }
 
 void GDScriptTokenizerBuffer::pop_expression_indented_block() {
-	ERR_FAIL_COND(indent_stack_stack.size() == 0);
+	ERR_FAIL_COND(indent_stack_stack.is_empty());
 	indent_stack = indent_stack_stack.back()->get();
 	indent_stack_stack.pop_back();
 }

modules/gdscript/gdscript_tokenizer_buffer.h

@@ -34,6 +34,12 @@
 #include "gdscript_tokenizer.h"
 
 class GDScriptTokenizerBuffer : public GDScriptTokenizer {
+public:
+	enum CompressMode {
+		COMPRESS_NONE,
+		COMPRESS_ZSTD,
+	};
+
 	enum {
 		TOKEN_BYTE_MASK = 0x80,
 		TOKEN_BITS = 8,
@@ -64,7 +70,7 @@ class GDScriptTokenizerBuffer : public GDScriptTokenizer {
 
 public:
 	Error set_code_buffer(const Vector<uint8_t> &p_buffer);
-	static Vector<uint8_t> parse_code_string(const String &p_code);
+	static Vector<uint8_t> parse_code_string(const String &p_code, CompressMode p_compress_mode);
 
 	virtual int get_cursor_line() const override;
 	virtual int get_cursor_column() const override;

modules/gdscript/register_types.cpp

@@ -84,7 +84,7 @@ class EditorExportGDScript : public EditorExportPlugin {
 
 public:
 	virtual void _export_file(const String &p_path, const String &p_type, const HashSet<String> &p_features) override {
-		int script_mode = EditorExportPreset::MODE_SCRIPT_BINARY_TOKENS;
+		int script_mode = EditorExportPreset::MODE_SCRIPT_BINARY_TOKENS_COMPRESSED;
 
 		const Ref<EditorExportPreset> &preset = get_export_preset();
 
@@ -103,7 +103,8 @@ public:
 
 		String source;
 		source.parse_utf8(reinterpret_cast<const char *>(file.ptr()), file.size());
-		file = GDScriptTokenizerBuffer::parse_code_string(source);
+		GDScriptTokenizerBuffer::CompressMode compress_mode = script_mode == EditorExportPreset::MODE_SCRIPT_BINARY_TOKENS_COMPRESSED ? GDScriptTokenizerBuffer::COMPRESS_ZSTD : GDScriptTokenizerBuffer::COMPRESS_NONE;
+		file = GDScriptTokenizerBuffer::parse_code_string(source, compress_mode);
 		if (file.is_empty()) {
 			return;
 		}

modules/gdscript/tests/gdscript_test_runner.cpp

@@ -538,7 +538,7 @@ GDScriptTest::TestResult GDScriptTest::execute_test_code(bool p_is_generating) {
 	} else {
 		String code = FileAccess::get_file_as_string(source_file, &err);
 		if (!err) {
-			Vector<uint8_t> buffer = GDScriptTokenizerBuffer::parse_code_string(code);
+			Vector<uint8_t> buffer = GDScriptTokenizerBuffer::parse_code_string(code, GDScriptTokenizerBuffer::COMPRESS_ZSTD);
 			script->set_binary_tokens_source(buffer);
 		}
 	}

modules/gdscript/tests/test_gdscript.cpp

@@ -111,7 +111,7 @@ static void test_tokenizer(const String &p_code, const Vector<String> &p_lines)
 static void test_tokenizer_buffer(const Vector<uint8_t> &p_buffer, const Vector<String> &p_lines);
 
 static void test_tokenizer_buffer(const String &p_code, const Vector<String> &p_lines) {
-	Vector<uint8_t> binary = GDScriptTokenizerBuffer::parse_code_string(p_code);
+	Vector<uint8_t> binary = GDScriptTokenizerBuffer::parse_code_string(p_code, GDScriptTokenizerBuffer::COMPRESS_NONE);
 	test_tokenizer_buffer(binary, p_lines);
 }