godot/core/string/ustring.h
2024-05-08 12:48:01 +02:00

644 lines
25 KiB
C++

/**************************************************************************/
/* ustring.h */
/**************************************************************************/
/* This file is part of: */
/* GODOT ENGINE */
/* https://godotengine.org */
/**************************************************************************/
/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */
/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */
/* */
/* 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. */
/**************************************************************************/
#ifndef USTRING_GODOT_H
#define USTRING_GODOT_H
// Note: _GODOT suffix added to header guard to avoid conflict with ICU header.
#include "core/string/char_utils.h"
#include "core/templates/cowdata.h"
#include "core/templates/vector.h"
#include "core/typedefs.h"
#include "core/variant/array.h"
/*************************************************************************/
/* CharProxy */
/*************************************************************************/
template <typename T>
class CharProxy {
friend class Char16String;
friend class CharString;
friend class String;
const int _index;
CowData<T> &_cowdata;
static const T _null = 0;
_FORCE_INLINE_ CharProxy(const int &p_index, CowData<T> &p_cowdata) :
_index(p_index),
_cowdata(p_cowdata) {}
public:
_FORCE_INLINE_ CharProxy(const CharProxy<T> &p_other) :
_index(p_other._index),
_cowdata(p_other._cowdata) {}
_FORCE_INLINE_ operator T() const {
if (unlikely(_index == _cowdata.size())) {
return _null;
}
return _cowdata.get(_index);
}
_FORCE_INLINE_ const T *operator&() const {
return _cowdata.ptr() + _index;
}
_FORCE_INLINE_ void operator=(const T &p_other) const {
_cowdata.set(_index, p_other);
}
_FORCE_INLINE_ void operator=(const CharProxy<T> &p_other) const {
_cowdata.set(_index, p_other.operator T());
}
};
/*************************************************************************/
/* Char16String */
/*************************************************************************/
class Char16String {
CowData<char16_t> _cowdata;
static const char16_t _null;
public:
_FORCE_INLINE_ char16_t *ptrw() { return _cowdata.ptrw(); }
_FORCE_INLINE_ const char16_t *ptr() const { return _cowdata.ptr(); }
_FORCE_INLINE_ int size() const { return _cowdata.size(); }
Error resize(int p_size) { return _cowdata.resize(p_size); }
_FORCE_INLINE_ char16_t get(int p_index) const { return _cowdata.get(p_index); }
_FORCE_INLINE_ void set(int p_index, const char16_t &p_elem) { _cowdata.set(p_index, p_elem); }
_FORCE_INLINE_ const char16_t &operator[](int p_index) const {
if (unlikely(p_index == _cowdata.size())) {
return _null;
}
return _cowdata.get(p_index);
}
_FORCE_INLINE_ CharProxy<char16_t> operator[](int p_index) { return CharProxy<char16_t>(p_index, _cowdata); }
_FORCE_INLINE_ Char16String() {}
_FORCE_INLINE_ Char16String(const Char16String &p_str) { _cowdata._ref(p_str._cowdata); }
_FORCE_INLINE_ void operator=(const Char16String &p_str) { _cowdata._ref(p_str._cowdata); }
_FORCE_INLINE_ Char16String(const char16_t *p_cstr) { copy_from(p_cstr); }
void operator=(const char16_t *p_cstr);
bool operator<(const Char16String &p_right) const;
Char16String &operator+=(char16_t p_char);
int length() const { return size() ? size() - 1 : 0; }
const char16_t *get_data() const;
operator const char16_t *() const { return get_data(); };
protected:
void copy_from(const char16_t *p_cstr);
};
/*************************************************************************/
/* CharString */
/*************************************************************************/
class CharString {
CowData<char> _cowdata;
static const char _null;
public:
_FORCE_INLINE_ char *ptrw() { return _cowdata.ptrw(); }
_FORCE_INLINE_ const char *ptr() const { return _cowdata.ptr(); }
_FORCE_INLINE_ int size() const { return _cowdata.size(); }
Error resize(int p_size) { return _cowdata.resize(p_size); }
_FORCE_INLINE_ char get(int p_index) const { return _cowdata.get(p_index); }
_FORCE_INLINE_ void set(int p_index, const char &p_elem) { _cowdata.set(p_index, p_elem); }
_FORCE_INLINE_ const char &operator[](int p_index) const {
if (unlikely(p_index == _cowdata.size())) {
return _null;
}
return _cowdata.get(p_index);
}
_FORCE_INLINE_ CharProxy<char> operator[](int p_index) { return CharProxy<char>(p_index, _cowdata); }
_FORCE_INLINE_ CharString() {}
_FORCE_INLINE_ CharString(const CharString &p_str) { _cowdata._ref(p_str._cowdata); }
_FORCE_INLINE_ void operator=(const CharString &p_str) { _cowdata._ref(p_str._cowdata); }
_FORCE_INLINE_ CharString(const char *p_cstr) { copy_from(p_cstr); }
void operator=(const char *p_cstr);
bool operator<(const CharString &p_right) const;
bool operator==(const CharString &p_right) const;
CharString &operator+=(char p_char);
int length() const { return size() ? size() - 1 : 0; }
const char *get_data() const;
operator const char *() const { return get_data(); };
protected:
void copy_from(const char *p_cstr);
};
/*************************************************************************/
/* String */
/*************************************************************************/
struct StrRange {
const char32_t *c_str;
int len;
StrRange(const char32_t *p_c_str = nullptr, int p_len = 0) {
c_str = p_c_str;
len = p_len;
}
};
class String {
CowData<char32_t> _cowdata;
static const char32_t _null;
static const char32_t _replacement_char;
void copy_from(const char *p_cstr);
void copy_from(const char *p_cstr, const int p_clip_to);
void copy_from(const wchar_t *p_cstr);
void copy_from(const wchar_t *p_cstr, const int p_clip_to);
void copy_from(const char32_t *p_cstr);
void copy_from(const char32_t *p_cstr, const int p_clip_to);
void copy_from(const char32_t &p_char);
void copy_from_unchecked(const char32_t *p_char, const int p_length);
bool _base_is_subsequence_of(const String &p_string, bool case_insensitive) const;
int _count(const String &p_string, int p_from, int p_to, bool p_case_insensitive) const;
int _count(const char *p_string, int p_from, int p_to, bool p_case_insensitive) const;
String _camelcase_to_underscore() const;
public:
enum {
npos = -1 ///<for "some" compatibility with std::string (npos is a huge value in std::string)
};
_FORCE_INLINE_ char32_t *ptrw() { return _cowdata.ptrw(); }
_FORCE_INLINE_ const char32_t *ptr() const { return _cowdata.ptr(); }
void remove_at(int p_index) { _cowdata.remove_at(p_index); }
_FORCE_INLINE_ void clear() { resize(0); }
_FORCE_INLINE_ char32_t get(int p_index) const { return _cowdata.get(p_index); }
_FORCE_INLINE_ void set(int p_index, const char32_t &p_elem) { _cowdata.set(p_index, p_elem); }
_FORCE_INLINE_ int size() const { return _cowdata.size(); }
Error resize(int p_size) { return _cowdata.resize(p_size); }
_FORCE_INLINE_ const char32_t &operator[](int p_index) const {
if (unlikely(p_index == _cowdata.size())) {
return _null;
}
return _cowdata.get(p_index);
}
_FORCE_INLINE_ CharProxy<char32_t> operator[](int p_index) { return CharProxy<char32_t>(p_index, _cowdata); }
bool operator==(const String &p_str) const;
bool operator!=(const String &p_str) const;
String operator+(const String &p_str) const;
String operator+(char32_t p_char) const;
String &operator+=(const String &);
String &operator+=(char32_t p_char);
String &operator+=(const char *p_str);
String &operator+=(const wchar_t *p_str);
String &operator+=(const char32_t *p_str);
/* Compatibility Operators */
void operator=(const char *p_str);
void operator=(const wchar_t *p_str);
void operator=(const char32_t *p_str);
bool operator==(const char *p_str) const;
bool operator==(const wchar_t *p_str) const;
bool operator==(const char32_t *p_str) const;
bool operator==(const StrRange &p_str_range) const;
bool operator!=(const char *p_str) const;
bool operator!=(const wchar_t *p_str) const;
bool operator!=(const char32_t *p_str) const;
bool operator<(const char32_t *p_str) const;
bool operator<(const char *p_str) const;
bool operator<(const wchar_t *p_str) const;
bool operator<(const String &p_str) const;
bool operator<=(const String &p_str) const;
bool operator>(const String &p_str) const;
bool operator>=(const String &p_str) const;
signed char casecmp_to(const String &p_str) const;
signed char nocasecmp_to(const String &p_str) const;
signed char naturalcasecmp_to(const String &p_str) const;
signed char naturalnocasecmp_to(const String &p_str) const;
// Special sorting for file names. Names starting with `_` are put before all others except those starting with `.`, otherwise natural comparison is used.
signed char filecasecmp_to(const String &p_str) const;
signed char filenocasecmp_to(const String &p_str) const;
const char32_t *get_data() const;
/* standard size stuff */
_FORCE_INLINE_ int length() const {
int s = size();
return s ? (s - 1) : 0; // length does not include zero
}
bool is_valid_string() const;
/* debug, error messages */
void print_unicode_error(const String &p_message, bool p_critical = false) const;
/* complex helpers */
String substr(int p_from, int p_chars = -1) const;
int find(const String &p_str, int p_from = 0) const; ///< return <0 if failed
int find(const char *p_str, int p_from = 0) const; ///< return <0 if failed
int find_char(const char32_t &p_char, int p_from = 0) const; ///< return <0 if failed
int findn(const String &p_str, int p_from = 0) const; ///< return <0 if failed, case insensitive
int findn(const char *p_str, int p_from = 0) const; ///< return <0 if failed
int rfind(const String &p_str, int p_from = -1) const; ///< return <0 if failed
int rfind(const char *p_str, int p_from = -1) const; ///< return <0 if failed
int rfindn(const String &p_str, int p_from = -1) const; ///< return <0 if failed, case insensitive
int rfindn(const char *p_str, int p_from = -1) const; ///< return <0 if failed
int findmk(const Vector<String> &p_keys, int p_from = 0, int *r_key = nullptr) const; ///< return <0 if failed
bool match(const String &p_wildcard) const;
bool matchn(const String &p_wildcard) const;
bool begins_with(const String &p_string) const;
bool begins_with(const char *p_string) const;
bool ends_with(const String &p_string) const;
bool ends_with(const char *p_string) const;
bool is_enclosed_in(const String &p_string) const;
bool is_subsequence_of(const String &p_string) const;
bool is_subsequence_ofn(const String &p_string) const;
bool is_quoted() const;
Vector<String> bigrams() const;
float similarity(const String &p_string) const;
String format(const Variant &values, const String &placeholder = "{_}") const;
String replace_first(const String &p_key, const String &p_with) const;
String replace_first(const char *p_key, const char *p_with) const;
String replace(const String &p_key, const String &p_with) const;
String replace(const char *p_key, const char *p_with) const;
String replacen(const String &p_key, const String &p_with) const;
String replacen(const char *p_key, const char *p_with) const;
String repeat(int p_count) const;
String reverse() const;
String insert(int p_at_pos, const String &p_string) const;
String erase(int p_pos, int p_chars = 1) const;
String pad_decimals(int p_digits) const;
String pad_zeros(int p_digits) const;
String trim_prefix(const String &p_prefix) const;
String trim_prefix(const char *p_prefix) const;
String trim_suffix(const String &p_suffix) const;
String trim_suffix(const char *p_suffix) const;
String lpad(int min_length, const String &character = " ") const;
String rpad(int min_length, const String &character = " ") const;
String sprintf(const Array &values, bool *error) const;
String quote(const String &quotechar = "\"") const;
String unquote() const;
static String num(double p_num, int p_decimals = -1);
static String num_scientific(double p_num);
static String num_real(double p_num, bool p_trailing = true);
static String num_int64(int64_t p_num, int base = 10, bool capitalize_hex = false);
static String num_uint64(uint64_t p_num, int base = 10, bool capitalize_hex = false);
static String chr(char32_t p_char);
static String md5(const uint8_t *p_md5);
static String hex_encode_buffer(const uint8_t *p_buffer, int p_len);
Vector<uint8_t> hex_decode() const;
bool is_numeric() const;
double to_float() const;
int64_t hex_to_int() const;
int64_t bin_to_int() const;
int64_t to_int() const;
static int64_t to_int(const char *p_str, int p_len = -1);
static int64_t to_int(const wchar_t *p_str, int p_len = -1);
static int64_t to_int(const char32_t *p_str, int p_len = -1, bool p_clamp = false);
static double to_float(const char *p_str);
static double to_float(const wchar_t *p_str, const wchar_t **r_end = nullptr);
static double to_float(const char32_t *p_str, const char32_t **r_end = nullptr);
static uint32_t num_characters(int64_t p_int);
String capitalize() const;
String to_camel_case() const;
String to_pascal_case() const;
String to_snake_case() const;
String get_with_code_lines() const;
int get_slice_count(const String &p_splitter) const;
int get_slice_count(const char *p_splitter) const;
String get_slice(const String &p_splitter, int p_slice) const;
String get_slice(const char *p_splitter, int p_slice) const;
String get_slicec(char32_t p_splitter, int p_slice) const;
Vector<String> split(const String &p_splitter = "", bool p_allow_empty = true, int p_maxsplit = 0) const;
Vector<String> split(const char *p_splitter = "", bool p_allow_empty = true, int p_maxsplit = 0) const;
Vector<String> rsplit(const String &p_splitter = "", bool p_allow_empty = true, int p_maxsplit = 0) const;
Vector<String> rsplit(const char *p_splitter = "", bool p_allow_empty = true, int p_maxsplit = 0) const;
Vector<String> split_spaces() const;
Vector<double> split_floats(const String &p_splitter, bool p_allow_empty = true) const;
Vector<float> split_floats_mk(const Vector<String> &p_splitters, bool p_allow_empty = true) const;
Vector<int> split_ints(const String &p_splitter, bool p_allow_empty = true) const;
Vector<int> split_ints_mk(const Vector<String> &p_splitters, bool p_allow_empty = true) const;
String join(const Vector<String> &parts) const;
static char32_t char_uppercase(char32_t p_char);
static char32_t char_lowercase(char32_t p_char);
String to_upper() const;
String to_lower() const;
int count(const String &p_string, int p_from = 0, int p_to = 0) const;
int count(const char *p_string, int p_from = 0, int p_to = 0) const;
int countn(const String &p_string, int p_from = 0, int p_to = 0) const;
int countn(const char *p_string, int p_from = 0, int p_to = 0) const;
String left(int p_len) const;
String right(int p_len) const;
String indent(const String &p_prefix) const;
String dedent() const;
String strip_edges(bool left = true, bool right = true) const;
String strip_escapes() const;
String lstrip(const String &p_chars) const;
String rstrip(const String &p_chars) const;
String get_extension() const;
String get_basename() const;
String path_join(const String &p_file) const;
char32_t unicode_at(int p_idx) const;
CharString ascii(bool p_allow_extended = false) const;
CharString utf8() const;
Error parse_utf8(const char *p_utf8, int p_len = -1, bool p_skip_cr = false);
static String utf8(const char *p_utf8, int p_len = -1);
Char16String utf16() const;
Error parse_utf16(const char16_t *p_utf16, int p_len = -1, bool p_default_little_endian = true);
static String utf16(const char16_t *p_utf16, int p_len = -1);
static uint32_t hash(const char32_t *p_cstr, int p_len); /* hash the string */
static uint32_t hash(const char32_t *p_cstr); /* hash the string */
static uint32_t hash(const wchar_t *p_cstr, int p_len); /* hash the string */
static uint32_t hash(const wchar_t *p_cstr); /* hash the string */
static uint32_t hash(const char *p_cstr, int p_len); /* hash the string */
static uint32_t hash(const char *p_cstr); /* hash the string */
uint32_t hash() const; /* hash the string */
uint64_t hash64() const; /* hash the string */
String md5_text() const;
String sha1_text() const;
String sha256_text() const;
Vector<uint8_t> md5_buffer() const;
Vector<uint8_t> sha1_buffer() const;
Vector<uint8_t> sha256_buffer() const;
_FORCE_INLINE_ bool is_empty() const { return length() == 0; }
_FORCE_INLINE_ bool contains(const char *p_str) const { return find(p_str) != -1; }
_FORCE_INLINE_ bool contains(const String &p_str) const { return find(p_str) != -1; }
_FORCE_INLINE_ bool containsn(const char *p_str) const { return findn(p_str) != -1; }
_FORCE_INLINE_ bool containsn(const String &p_str) const { return findn(p_str) != -1; }
// path functions
bool is_absolute_path() const;
bool is_relative_path() const;
bool is_resource_file() const;
String path_to(const String &p_path) const;
String path_to_file(const String &p_path) const;
String get_base_dir() const;
String get_file() const;
static String humanize_size(uint64_t p_size);
String simplify_path() const;
bool is_network_share_path() const;
String xml_escape(bool p_escape_quotes = false) const;
String xml_unescape() const;
String uri_encode() const;
String uri_decode() const;
String c_escape() const;
String c_escape_multiline() const;
String c_unescape() const;
String json_escape() const;
Error parse_url(String &r_scheme, String &r_host, int &r_port, String &r_path) const;
String property_name_encode() const;
// node functions
static String get_invalid_node_name_characters(bool p_allow_internal = false);
String validate_node_name() const;
String validate_identifier() const;
String validate_filename() const;
bool is_valid_identifier() const;
bool is_valid_int() const;
bool is_valid_float() const;
bool is_valid_hex_number(bool p_with_prefix) const;
bool is_valid_html_color() const;
bool is_valid_ip_address() const;
bool is_valid_filename() const;
/**
* The constructors must not depend on other overloads
*/
_FORCE_INLINE_ String() {}
_FORCE_INLINE_ String(const String &p_str) { _cowdata._ref(p_str._cowdata); }
_FORCE_INLINE_ void operator=(const String &p_str) { _cowdata._ref(p_str._cowdata); }
Vector<uint8_t> to_ascii_buffer() const;
Vector<uint8_t> to_utf8_buffer() const;
Vector<uint8_t> to_utf16_buffer() const;
Vector<uint8_t> to_utf32_buffer() const;
Vector<uint8_t> to_wchar_buffer() const;
String(const char *p_str);
String(const wchar_t *p_str);
String(const char32_t *p_str);
String(const char *p_str, int p_clip_to_len);
String(const wchar_t *p_str, int p_clip_to_len);
String(const char32_t *p_str, int p_clip_to_len);
String(const StrRange &p_range);
};
bool operator==(const char *p_chr, const String &p_str);
bool operator==(const wchar_t *p_chr, const String &p_str);
bool operator!=(const char *p_chr, const String &p_str);
bool operator!=(const wchar_t *p_chr, const String &p_str);
String operator+(const char *p_chr, const String &p_str);
String operator+(const wchar_t *p_chr, const String &p_str);
String operator+(char32_t p_chr, const String &p_str);
String itos(int64_t p_val);
String uitos(uint64_t p_val);
String rtos(double p_val);
String rtoss(double p_val); //scientific version
struct NoCaseComparator {
bool operator()(const String &p_a, const String &p_b) const {
return p_a.nocasecmp_to(p_b) < 0;
}
};
struct NaturalNoCaseComparator {
bool operator()(const String &p_a, const String &p_b) const {
return p_a.naturalnocasecmp_to(p_b) < 0;
}
};
struct FileNoCaseComparator {
bool operator()(const String &p_a, const String &p_b) const {
return p_a.filenocasecmp_to(p_b) < 0;
}
};
template <typename L, typename R>
_FORCE_INLINE_ bool is_str_less(const L *l_ptr, const R *r_ptr) {
while (true) {
const char32_t l = *l_ptr;
const char32_t r = *r_ptr;
if (l == 0 && r == 0) {
return false;
} else if (l == 0) {
return true;
} else if (r == 0) {
return false;
} else if (l < r) {
return true;
} else if (l > r) {
return false;
}
l_ptr++;
r_ptr++;
}
}
/* end of namespace */
// Tool translate (TTR and variants) for the editor UI,
// and doc translate for the class reference (DTR).
#ifdef TOOLS_ENABLED
// Gets parsed.
String TTR(const String &p_text, const String &p_context = "");
String TTRN(const String &p_text, const String &p_text_plural, int p_n, const String &p_context = "");
String DTR(const String &p_text, const String &p_context = "");
String DTRN(const String &p_text, const String &p_text_plural, int p_n, const String &p_context = "");
// Use for C strings.
#define TTRC(m_value) (m_value)
// Use to avoid parsing (for use later with C strings).
#define TTRGET(m_value) TTR(m_value)
#else
#define TTRC(m_value) (m_value)
#define TTRGET(m_value) (m_value)
#endif
// Use this to mark property names for editor translation.
// Often for dynamic properties defined in _get_property_list().
// Property names defined directly inside EDITOR_DEF, GLOBAL_DEF, and ADD_PROPERTY macros don't need this.
#define PNAME(m_value) (m_value)
// Similar to PNAME, but to mark groups, i.e. properties with PROPERTY_USAGE_GROUP.
// Groups defined directly inside ADD_GROUP macros don't need this.
// The arguments are the same as ADD_GROUP. m_prefix is only used for extraction.
#define GNAME(m_value, m_prefix) (m_value)
// Runtime translate for the public node API.
String RTR(const String &p_text, const String &p_context = "");
String RTRN(const String &p_text, const String &p_text_plural, int p_n, const String &p_context = "");
/**
* "Extractable TRanslate". Used for strings that can appear inside an exported
* project (such as the ones in nodes like `FileDialog`), which are made possible
* to add in the POT generator. A translation context can optionally be specified
* to disambiguate between identical source strings in translations.
* When placeholders are desired, use vformat(ETR("Example: %s"), some_string)`.
* If a string mentions a quantity (and may therefore need a dynamic plural form),
* use `ETRN()` instead of `ETR()`.
*
* NOTE: This function is for string extraction only, and will just return the
* string it was given. The translation itself should be done internally by nodes
* with `atr()` instead.
*/
_FORCE_INLINE_ String ETR(const String &p_text, const String &p_context = "") {
return p_text;
}
/**
* "Extractable TRanslate for N items". Used for strings that can appear inside an
* exported project (such as the ones in nodes like `FileDialog`), which are made
* possible to add in the POT generator. A translation context can optionally be
* specified to disambiguate between identical source strings in translations.
* Use `ETR()` if the string doesn't need dynamic plural form. When placeholders
* are desired, use `vformat(ETRN("%d item", "%d items", some_integer), some_integer)`.
* The placeholder must be present in both strings to avoid run-time warnings in `vformat()`.
*
* NOTE: This function is for string extraction only, and will just return the
* string it was given. The translation itself should be done internally by nodes
* with `atr()` instead.
*/
_FORCE_INLINE_ String ETRN(const String &p_text, const String &p_text_plural, int p_n, const String &p_context = "") {
if (p_n == 1) {
return p_text;
}
return p_text_plural;
}
bool select_word(const String &p_s, int p_col, int &r_beg, int &r_end);
_FORCE_INLINE_ void sarray_add_str(Vector<String> &arr) {
}
_FORCE_INLINE_ void sarray_add_str(Vector<String> &arr, const String &p_str) {
arr.push_back(p_str);
}
template <typename... P>
_FORCE_INLINE_ void sarray_add_str(Vector<String> &arr, const String &p_str, P... p_args) {
arr.push_back(p_str);
sarray_add_str(arr, p_args...);
}
template <typename... P>
_FORCE_INLINE_ Vector<String> sarray(P... p_args) {
Vector<String> arr;
sarray_add_str(arr, p_args...);
return arr;
}
#endif // USTRING_GODOT_H