mirror of
https://github.com/godotengine/godot
synced 2024-11-02 14:03:02 +00:00
411 lines
14 KiB
C++
411 lines
14 KiB
C++
/*************************************************************************/
|
|
/* translation_loader_po.cpp */
|
|
/*************************************************************************/
|
|
/* This file is part of: */
|
|
/* GODOT ENGINE */
|
|
/* https://godotengine.org */
|
|
/*************************************************************************/
|
|
/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */
|
|
/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */
|
|
/* */
|
|
/* Permission is hereby granted, free of charge, to any person obtaining */
|
|
/* a copy of this software and associated documentation files (the */
|
|
/* "Software"), to deal in the Software without restriction, including */
|
|
/* without limitation the rights to use, copy, modify, merge, publish, */
|
|
/* distribute, sublicense, and/or sell copies of the Software, and to */
|
|
/* permit persons to whom the Software is furnished to do so, subject to */
|
|
/* the following conditions: */
|
|
/* */
|
|
/* The above copyright notice and this permission notice shall be */
|
|
/* included in all copies or substantial portions of the Software. */
|
|
/* */
|
|
/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */
|
|
/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */
|
|
/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
|
|
/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */
|
|
/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */
|
|
/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */
|
|
/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */
|
|
/*************************************************************************/
|
|
|
|
#include "translation_loader_po.h"
|
|
|
|
#include "core/io/file_access.h"
|
|
#include "core/string/translation.h"
|
|
#include "core/string/translation_po.h"
|
|
|
|
RES TranslationLoaderPO::load_translation(FileAccess *f, Error *r_error) {
|
|
if (r_error) {
|
|
*r_error = ERR_FILE_CORRUPT;
|
|
}
|
|
|
|
const String path = f->get_path();
|
|
Ref<TranslationPO> translation = Ref<TranslationPO>(memnew(TranslationPO));
|
|
String config;
|
|
|
|
uint32_t magic = f->get_32();
|
|
if (magic == 0x950412de) {
|
|
// Load binary MO file.
|
|
|
|
uint16_t version_maj = f->get_16();
|
|
uint16_t version_min = f->get_16();
|
|
if (version_maj > 1) {
|
|
ERR_FAIL_V_MSG(RES(), vformat("Unsupported MO file %s, version %d.%d.", path, version_maj, version_min));
|
|
}
|
|
|
|
uint32_t num_strings = f->get_32();
|
|
uint32_t id_table_offset = f->get_32();
|
|
uint32_t trans_table_offset = f->get_32();
|
|
|
|
// Read string tables.
|
|
for (uint32_t i = 0; i < num_strings; i++) {
|
|
String msg_id;
|
|
String msg_id_plural;
|
|
String msg_context;
|
|
|
|
// Read id strings and context.
|
|
{
|
|
Vector<uint8_t> data;
|
|
f->seek(id_table_offset + i * 8);
|
|
uint32_t str_start = 0;
|
|
uint32_t str_len = f->get_32();
|
|
uint32_t str_offset = f->get_32();
|
|
|
|
data.resize(str_len + 1);
|
|
f->seek(str_offset);
|
|
f->get_buffer(data.ptrw(), str_len);
|
|
data.write[str_len] = 0;
|
|
|
|
bool is_plural = false;
|
|
for (uint32_t j = 0; j < str_len + 1; j++) {
|
|
if (data[j] == 0x04) {
|
|
msg_context.parse_utf8((const char *)data.ptr(), j);
|
|
str_start = j + 1;
|
|
}
|
|
if (data[j] == 0x00) {
|
|
if (is_plural) {
|
|
msg_id_plural.parse_utf8((const char *)(data.ptr() + str_start), j - str_start);
|
|
} else {
|
|
msg_id.parse_utf8((const char *)(data.ptr() + str_start), j - str_start);
|
|
is_plural = true;
|
|
}
|
|
str_start = j + 1;
|
|
}
|
|
}
|
|
}
|
|
|
|
// Read translated strings.
|
|
{
|
|
Vector<uint8_t> data;
|
|
f->seek(trans_table_offset + i * 8);
|
|
uint32_t str_len = f->get_32();
|
|
uint32_t str_offset = f->get_32();
|
|
|
|
data.resize(str_len + 1);
|
|
f->seek(str_offset);
|
|
f->get_buffer(data.ptrw(), str_len);
|
|
data.write[str_len] = 0;
|
|
|
|
if (msg_id.is_empty()) {
|
|
config = String::utf8((const char *)data.ptr(), str_len);
|
|
// Record plural rule.
|
|
int p_start = config.find("Plural-Forms");
|
|
if (p_start != -1) {
|
|
int p_end = config.find("\n", p_start);
|
|
translation->set_plural_rule(config.substr(p_start, p_end - p_start));
|
|
}
|
|
} else {
|
|
uint32_t str_start = 0;
|
|
Vector<String> plural_msg;
|
|
for (uint32_t j = 0; j < str_len + 1; j++) {
|
|
if (data[j] == 0x00) {
|
|
if (msg_id_plural.is_empty()) {
|
|
translation->add_message(msg_id, String::utf8((const char *)(data.ptr() + str_start), j - str_start), msg_context);
|
|
} else {
|
|
plural_msg.push_back(String::utf8((const char *)(data.ptr() + str_start), j - str_start));
|
|
}
|
|
str_start = j + 1;
|
|
}
|
|
}
|
|
if (!plural_msg.is_empty()) {
|
|
translation->add_plural_message(msg_id, plural_msg, msg_context);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
memdelete(f);
|
|
} else {
|
|
// Try to load as text PO file.
|
|
f->seek(0);
|
|
|
|
enum Status {
|
|
STATUS_NONE,
|
|
STATUS_READING_ID,
|
|
STATUS_READING_STRING,
|
|
STATUS_READING_CONTEXT,
|
|
STATUS_READING_PLURAL,
|
|
};
|
|
|
|
Status status = STATUS_NONE;
|
|
|
|
String msg_id;
|
|
String msg_str;
|
|
String msg_context;
|
|
Vector<String> msgs_plural;
|
|
|
|
if (r_error) {
|
|
*r_error = ERR_FILE_CORRUPT;
|
|
}
|
|
|
|
int line = 1;
|
|
int plural_forms = 0;
|
|
int plural_index = -1;
|
|
bool entered_context = false;
|
|
bool skip_this = false;
|
|
bool skip_next = false;
|
|
bool is_eof = false;
|
|
|
|
while (!is_eof) {
|
|
String l = f->get_line().strip_edges();
|
|
is_eof = f->eof_reached();
|
|
|
|
// If we reached last line and it's not a content line, break, otherwise let processing that last loop
|
|
if (is_eof && l.is_empty()) {
|
|
if (status == STATUS_READING_ID || status == STATUS_READING_CONTEXT || (status == STATUS_READING_PLURAL && plural_index != plural_forms - 1)) {
|
|
memdelete(f);
|
|
ERR_FAIL_V_MSG(RES(), "Unexpected EOF while reading PO file at: " + path + ":" + itos(line));
|
|
} else {
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (l.begins_with("msgctxt")) {
|
|
if (status != STATUS_READING_STRING && status != STATUS_READING_PLURAL) {
|
|
memdelete(f);
|
|
ERR_FAIL_V_MSG(RES(), "Unexpected 'msgctxt', was expecting 'msgid_plural' or 'msgstr' before 'msgctxt' while parsing: " + path + ":" + itos(line));
|
|
}
|
|
|
|
// In PO file, "msgctxt" appears before "msgid". If we encounter a "msgctxt", we add what we have read
|
|
// and set "entered_context" to true to prevent adding twice.
|
|
if (!skip_this && !msg_id.is_empty()) {
|
|
if (status == STATUS_READING_STRING) {
|
|
translation->add_message(msg_id, msg_str, msg_context);
|
|
} else if (status == STATUS_READING_PLURAL) {
|
|
if (plural_index != plural_forms - 1) {
|
|
memdelete(f);
|
|
ERR_FAIL_V_MSG(RES(), "Number of 'msgstr[]' doesn't match with number of plural forms: " + path + ":" + itos(line));
|
|
}
|
|
translation->add_plural_message(msg_id, msgs_plural, msg_context);
|
|
}
|
|
}
|
|
msg_context = "";
|
|
l = l.substr(7, l.length()).strip_edges();
|
|
status = STATUS_READING_CONTEXT;
|
|
entered_context = true;
|
|
}
|
|
|
|
if (l.begins_with("msgid_plural")) {
|
|
if (plural_forms == 0) {
|
|
memdelete(f);
|
|
ERR_FAIL_V_MSG(RES(), "PO file uses 'msgid_plural' but 'Plural-Forms' is invalid or missing in header: " + path + ":" + itos(line));
|
|
} else if (status != STATUS_READING_ID) {
|
|
memdelete(f);
|
|
ERR_FAIL_V_MSG(RES(), "Unexpected 'msgid_plural', was expecting 'msgid' before 'msgid_plural' while parsing: " + path + ":" + itos(line));
|
|
}
|
|
// We don't record the message in "msgid_plural" itself as tr_n(), TTRN(), RTRN() interfaces provide the plural string already.
|
|
// We just have to reset variables related to plurals for "msgstr[]" later on.
|
|
l = l.substr(12, l.length()).strip_edges();
|
|
plural_index = -1;
|
|
msgs_plural.clear();
|
|
msgs_plural.resize(plural_forms);
|
|
status = STATUS_READING_PLURAL;
|
|
} else if (l.begins_with("msgid")) {
|
|
if (status == STATUS_READING_ID) {
|
|
memdelete(f);
|
|
ERR_FAIL_V_MSG(RES(), "Unexpected 'msgid', was expecting 'msgstr' while parsing: " + path + ":" + itos(line));
|
|
}
|
|
|
|
if (!msg_id.is_empty()) {
|
|
if (!skip_this && !entered_context) {
|
|
if (status == STATUS_READING_STRING) {
|
|
translation->add_message(msg_id, msg_str, msg_context);
|
|
} else if (status == STATUS_READING_PLURAL) {
|
|
if (plural_index != plural_forms - 1) {
|
|
memdelete(f);
|
|
ERR_FAIL_V_MSG(RES(), "Number of 'msgstr[]' doesn't match with number of plural forms: " + path + ":" + itos(line));
|
|
}
|
|
translation->add_plural_message(msg_id, msgs_plural, msg_context);
|
|
}
|
|
}
|
|
} else if (config.is_empty()) {
|
|
config = msg_str;
|
|
// Record plural rule.
|
|
int p_start = config.find("Plural-Forms");
|
|
if (p_start != -1) {
|
|
int p_end = config.find("\n", p_start);
|
|
translation->set_plural_rule(config.substr(p_start, p_end - p_start));
|
|
plural_forms = translation->get_plural_forms();
|
|
}
|
|
}
|
|
|
|
l = l.substr(5, l.length()).strip_edges();
|
|
status = STATUS_READING_ID;
|
|
// If we did not encounter msgctxt, we reset context to empty to reset it.
|
|
if (!entered_context) {
|
|
msg_context = "";
|
|
}
|
|
msg_id = "";
|
|
msg_str = "";
|
|
skip_this = skip_next;
|
|
skip_next = false;
|
|
entered_context = false;
|
|
}
|
|
|
|
if (l.begins_with("msgstr[")) {
|
|
if (status != STATUS_READING_PLURAL) {
|
|
memdelete(f);
|
|
ERR_FAIL_V_MSG(RES(), "Unexpected 'msgstr[]', was expecting 'msgid_plural' before 'msgstr[]' while parsing: " + path + ":" + itos(line));
|
|
}
|
|
plural_index++; // Increment to add to the next slot in vector msgs_plural.
|
|
l = l.substr(9, l.length()).strip_edges();
|
|
} else if (l.begins_with("msgstr")) {
|
|
if (status != STATUS_READING_ID) {
|
|
memdelete(f);
|
|
ERR_FAIL_V_MSG(RES(), "Unexpected 'msgstr', was expecting 'msgid' before 'msgstr' while parsing: " + path + ":" + itos(line));
|
|
}
|
|
|
|
l = l.substr(6, l.length()).strip_edges();
|
|
status = STATUS_READING_STRING;
|
|
}
|
|
|
|
if (l.is_empty() || l.begins_with("#")) {
|
|
if (l.contains("fuzzy")) {
|
|
skip_next = true;
|
|
}
|
|
line++;
|
|
continue; // Nothing to read or comment.
|
|
}
|
|
|
|
if (!l.begins_with("\"") || status == STATUS_NONE) {
|
|
memdelete(f);
|
|
ERR_FAIL_V_MSG(RES(), "Invalid line '" + l + "' while parsing: " + path + ":" + itos(line));
|
|
}
|
|
|
|
l = l.substr(1, l.length());
|
|
// Find final quote, ignoring escaped ones (\").
|
|
// The escape_next logic is necessary to properly parse things like \\"
|
|
// where the backslash is the one being escaped, not the quote.
|
|
int end_pos = -1;
|
|
bool escape_next = false;
|
|
for (int i = 0; i < l.length(); i++) {
|
|
if (l[i] == '\\' && !escape_next) {
|
|
escape_next = true;
|
|
continue;
|
|
}
|
|
|
|
if (l[i] == '"' && !escape_next) {
|
|
end_pos = i;
|
|
break;
|
|
}
|
|
|
|
escape_next = false;
|
|
}
|
|
|
|
if (end_pos == -1) {
|
|
memdelete(f);
|
|
ERR_FAIL_V_MSG(RES(), "Expected '\"' at end of message while parsing: " + path + ":" + itos(line));
|
|
}
|
|
|
|
l = l.substr(0, end_pos);
|
|
l = l.c_unescape();
|
|
|
|
if (status == STATUS_READING_ID) {
|
|
msg_id += l;
|
|
} else if (status == STATUS_READING_STRING) {
|
|
msg_str += l;
|
|
} else if (status == STATUS_READING_CONTEXT) {
|
|
msg_context += l;
|
|
} else if (status == STATUS_READING_PLURAL && plural_index >= 0) {
|
|
if (plural_index >= plural_forms) {
|
|
memdelete(f);
|
|
ERR_FAIL_V_MSG(RES(), "Unexpected plural form while parsing: " + path + ":" + itos(line));
|
|
}
|
|
msgs_plural.write[plural_index] = msgs_plural[plural_index] + l;
|
|
}
|
|
|
|
line++;
|
|
}
|
|
|
|
memdelete(f);
|
|
|
|
// Add the last set of data from last iteration.
|
|
if (status == STATUS_READING_STRING) {
|
|
if (!msg_id.is_empty()) {
|
|
if (!skip_this) {
|
|
translation->add_message(msg_id, msg_str, msg_context);
|
|
}
|
|
} else if (config.is_empty()) {
|
|
config = msg_str;
|
|
}
|
|
} else if (status == STATUS_READING_PLURAL) {
|
|
if (!skip_this && !msg_id.is_empty()) {
|
|
if (plural_index != plural_forms - 1) {
|
|
memdelete(f);
|
|
ERR_FAIL_V_MSG(RES(), "Number of 'msgstr[]' doesn't match with number of plural forms: " + path + ":" + itos(line));
|
|
}
|
|
translation->add_plural_message(msg_id, msgs_plural, msg_context);
|
|
}
|
|
}
|
|
}
|
|
|
|
ERR_FAIL_COND_V_MSG(config.is_empty(), RES(), "No config found in file: " + path + ".");
|
|
|
|
Vector<String> configs = config.split("\n");
|
|
for (int i = 0; i < configs.size(); i++) {
|
|
String c = configs[i].strip_edges();
|
|
int p = c.find(":");
|
|
if (p == -1) {
|
|
continue;
|
|
}
|
|
String prop = c.substr(0, p).strip_edges();
|
|
String value = c.substr(p + 1, c.length()).strip_edges();
|
|
|
|
if (prop == "X-Language" || prop == "Language") {
|
|
translation->set_locale(value);
|
|
}
|
|
}
|
|
|
|
if (r_error) {
|
|
*r_error = OK;
|
|
}
|
|
|
|
return translation;
|
|
}
|
|
|
|
RES TranslationLoaderPO::load(const String &p_path, const String &p_original_path, Error *r_error, bool p_use_sub_threads, float *r_progress, CacheMode p_cache_mode) {
|
|
if (r_error) {
|
|
*r_error = ERR_CANT_OPEN;
|
|
}
|
|
|
|
FileAccess *f = FileAccess::open(p_path, FileAccess::READ);
|
|
ERR_FAIL_COND_V_MSG(!f, RES(), "Cannot open file '" + p_path + "'.");
|
|
|
|
return load_translation(f, r_error);
|
|
}
|
|
|
|
void TranslationLoaderPO::get_recognized_extensions(List<String> *p_extensions) const {
|
|
p_extensions->push_back("po");
|
|
p_extensions->push_back("mo");
|
|
}
|
|
|
|
bool TranslationLoaderPO::handles_type(const String &p_type) const {
|
|
return (p_type == "Translation");
|
|
}
|
|
|
|
String TranslationLoaderPO::get_resource_type(const String &p_path) const {
|
|
if (p_path.get_extension().to_lower() == "po" || p_path.get_extension().to_lower() == "mo") {
|
|
return "Translation";
|
|
}
|
|
return "";
|
|
}
|