godot/core/ustring.cpp
2017-02-08 10:40:35 +09:00

4177 lines
74 KiB
C++

/*************************************************************************/
/* ustring.cpp */
/*************************************************************************/
/* This file is part of: */
/* GODOT ENGINE */
/* http://www.godotengine.org */
/*************************************************************************/
/* Copyright (c) 2007-2017 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. */
/*************************************************************************/
#include <wchar.h>
#include "ustring.h"
#include "os/memory.h"
#include "print_string.h"
#include "math_funcs.h"
#include "io/md5.h"
#include "io/sha256.h"
#include "ucaps.h"
#include "color.h"
#include "variant.h"
#define MAX_DIGITS 6
#define UPPERCASE(m_c) (((m_c)>='a' && (m_c)<='z')?((m_c)-('a'-'A')):(m_c))
#define LOWERCASE(m_c) (((m_c)>='A' && (m_c)<='Z')?((m_c)+('a'-'A')):(m_c))
#ifndef NO_USE_STDLIB
#include <stdlib.h>
#include <stdio.h>
#endif
#if defined(MINGW_ENABLED) || defined(_MSC_VER)
#define snprintf _snprintf
#endif
/** STRING **/
const char *CharString::get_data() const {
if (size())
return &operator[](0);
else
return "";
}
void String::copy_from(const char *p_cstr) {
int len=0;
const char *ptr=p_cstr;
while (*(ptr++)!=0)
len++;
if (len==0) {
resize(0);
return;
}
resize(len+1); // include 0
CharType *dst = this->ptr();
for (int i=0;i<len+1;i++) {
dst[i]=p_cstr[i];
}
}
void String::copy_from(const CharType* p_cstr, int p_clip_to) {
int len=0;
const CharType *ptr=p_cstr;
while (*(ptr++)!=0)
len++;
if (p_clip_to>=0 && len>p_clip_to)
len=p_clip_to;
if (len==0) {
resize(0);
return;
}
resize(len+1);
set(len,0);
CharType *dst = &operator[](0);
for (int i=0;i<len;i++) {
dst[i]=p_cstr[i];
}
}
void String::copy_from(const CharType& p_char) {
resize(2);
set(0,p_char);
set(1,0);
}
bool String::operator==(const String& p_str) const {
if (length()!=p_str.length())
return false;
if (empty())
return true;
int l=length();
const CharType *src = c_str();
const CharType *dst = p_str.c_str();
/* Compare char by char */
for (int i=0;i<l;i++) {
if (src[i]!=dst[i])
return false;
}
return true;
}
bool String::operator!=(const String& p_str) const {
return !(*this==p_str);
}
String String::operator+(const String& p_str) const {
String res=*this;
res+=p_str;
return res;
}
/*
String String::operator+(CharType p_chr) const {
String res=*this;
res+=p_chr;
return res;
}
*/
String& String::operator+=(const String &p_str) {
if (empty()) {
*this=p_str;
return *this;
}
if (p_str.empty())
return *this;
int from=length();
resize( length() + p_str.size() );
const CharType *src = p_str.c_str();
CharType *dst = &operator[](0);
set( length(), 0 );
for (int i=0;i<p_str.length();i++)
dst[from+i]=src[i];
return *this;
}
String& String::operator+=(const CharType *p_str) {
*this+=String(p_str);
return *this;
}
String& String::operator+=(CharType p_char) {
resize( size() ? size() + 1 : 2);
set( length(), 0 );
set( length() -1, p_char );
return *this;
}
String& String::operator+=(const char * p_str) {
if (!p_str || p_str[0]==0)
return *this;
int src_len=0;
const char *ptr=p_str;
while (*(ptr++)!=0)
src_len++;
int from=length();
resize( from + src_len + 1 );
CharType *dst = &operator[](0);
set( length(), 0 );
for (int i=0;i<src_len;i++)
dst[from+i]=p_str[i];
return *this;
}
void String::operator=(const char *p_str) {
copy_from(p_str);
}
void String::operator=(const CharType *p_str) {
copy_from(p_str);
}
bool String::operator==(const StrRange &p_range) const {
int len=p_range.len;
if (length()!=len)
return false;
if (empty())
return true;
const CharType *c_str=p_range.c_str;
const CharType *dst = &operator[](0);
/* Compare char by char */
for (int i=0;i<len;i++) {
if (c_str[i]!=dst[i])
return false;
}
return true;
}
bool String::operator==(const char *p_str) const {
int len=0;
const char *aux=p_str;
while (*(aux++)!=0)
len++;
if (length()!=len)
return false;
if (empty())
return true;
int l=length();
const CharType *dst = c_str();
/* Compare char by char */
for (int i=0;i<l;i++) {
if (p_str[i]!=dst[i])
return false;
}
return true;
}
bool String::operator==(const CharType *p_str) const {
int len=0;
const CharType *aux=p_str;
while (*(aux++)!=0)
len++;
if (length()!=len)
return false;
if (empty())
return true;
int l=length();
const CharType *dst = c_str();
/* Compare char by char */
for (int i=0;i<l;i++) {
if (p_str[i]!=dst[i])
return false;
}
return true;
}
bool String::operator!=(const char *p_str) const {
return (! ( *this==p_str ) );
}
bool String::operator!=(const CharType *p_str) const {
return (! ( *this==p_str ) );
}
bool String::operator<(const CharType *p_str) const {
if (empty() && p_str[0]==0)
return false;
if (empty())
return true;
const CharType *this_str=c_str();
while (true) {
if (*p_str==0 && *this_str==0)
return false; //this can't be equal, sadly
else if (*this_str==0)
return true; //if this is empty, and the other one is not, then we're less.. I think?
else if (*p_str==0)
return false; //otherwise the other one is smaller..
else if (*this_str < *p_str ) //more than
return true;
else if (*this_str > *p_str ) //less than
return false;
this_str++;
p_str++;
}
return false; //should never reach here anyway
}
bool String::operator<=(String p_str) const {
return (*this<p_str) || (*this==p_str);
}
bool String::operator<(const char *p_str) const {
if (empty() && p_str[0]==0)
return false;
if (empty())
return true;
const CharType *this_str=c_str();
while (true) {
if (*p_str==0 && *this_str==0)
return false; //this can't be equal, sadly
else if (*this_str==0)
return true; //if this is empty, and the other one is not, then we're less.. I think?
else if (*p_str==0)
return false; //otherwise the other one is smaller..
else if (*this_str < *p_str ) //more than
return true;
else if (*this_str > *p_str ) //less than
return false;
this_str++;
p_str++;
}
return false; //should never reach here anyway
}
bool String::operator<(String p_str) const {
return operator<(p_str.c_str());
}
signed char String::nocasecmp_to(const String &p_str) const {
if (empty() && p_str.empty())
return 0;
if (empty())
return -1;
if (p_str.empty())
return 1;
const CharType *that_str=p_str.c_str();
const CharType *this_str=c_str();
while (true) {
if (*that_str==0 && *this_str==0)
return 0; //we're equal
else if (*this_str==0)
return -1; //if this is empty, and the other one is not, then we're less.. I think?
else if (*that_str==0)
return 1; //otherwise the other one is smaller..
else if (_find_upper(*this_str) < _find_upper(*that_str) ) //more than
return -1;
else if (_find_upper(*this_str) > _find_upper(*that_str) ) //less than
return 1;
this_str++;
that_str++;
}
return 0; //should never reach anyway
}
signed char String::casecmp_to(const String &p_str) const {
if (empty() && p_str.empty())
return 0;
if (empty())
return -1;
if (p_str.empty())
return 1;
const CharType *that_str=p_str.c_str();
const CharType *this_str=c_str();
while (true) {
if (*that_str==0 && *this_str==0)
return 0; //we're equal
else if (*this_str==0)
return -1; //if this is empty, and the other one is not, then we're less.. I think?
else if (*that_str==0)
return 1; //otherwise the other one is smaller..
else if (*this_str < *that_str ) //more than
return -1;
else if (*this_str > *that_str ) //less than
return 1;
this_str++;
that_str++;
}
return 0; //should never reach anyway
}
void String::erase(int p_pos, int p_chars) {
*this=left(p_pos)+substr( p_pos + p_chars , length() - ( (p_pos+p_chars) ) );
}
String String::capitalize() const {
String aux=this->replace("_"," ").to_lower();
String cap;
for (int i=0;i<aux.get_slice_count(" ");i++) {
String slice=aux.get_slicec(' ',i);
if (slice.length()>0) {
slice[0]=_find_upper(slice[0]);
if (i>0)
cap+=" ";
cap+=slice;
}
}
return cap;
}
String String::camelcase_to_underscore(bool lowercase) const {
const CharType * cstr = c_str();
String new_string;
const char A = 'A', Z = 'Z';
const char a = 'a', z = 'z';
int start_index = 0;
for ( size_t i = 1; i < this->size(); i++ ) {
bool is_upper = cstr[i] >= A && cstr[i] <= Z;
bool are_next_2_lower = false;
bool was_precedent_upper = cstr[i-1] >= A && cstr[i-1] <= Z;
if (i+2 < this->size()) {
are_next_2_lower = cstr[i+1] >= a && cstr[i+1] <= z && cstr[i+2] >= a && cstr[i+2] <= z;
}
bool should_split = ((is_upper && !was_precedent_upper) || (was_precedent_upper && is_upper && are_next_2_lower));
if (should_split) {
new_string += this->substr(start_index, i - start_index) + "_";
start_index = i;
}
}
new_string += this->substr(start_index, this->size() - start_index);
return lowercase ? new_string.to_lower() : new_string;
}
int String::get_slice_count(String p_splitter) const{
if (empty())
return 0;
if (p_splitter.empty())
return 0;
int pos=0;
int slices=1;
while ( (pos=find(p_splitter,pos))>=0) {
slices++;
pos+=p_splitter.length();
}
return slices;
}
String String::get_slice(String p_splitter, int p_slice) const {
if (empty() || p_splitter.empty())
return "";
int pos=0;
int prev_pos=0;
// int slices=1;
if (p_slice<0)
return "";
if (find(p_splitter)==-1)
return *this;
int i=0;
while(true) {
pos=find(p_splitter,pos);
if (pos==-1)
pos=length(); //reached end
int from=prev_pos;
// int to=pos;
if (p_slice==i) {
return substr( from, pos-from );
}
if (pos==length()) //reached end and no find
break;
pos+=p_splitter.length();
prev_pos=pos;
i++;
}
return ""; //no find!
}
String String::get_slicec(CharType p_splitter, int p_slice) const {
if (empty())
return String();
if (p_slice<0)
return String();
const CharType *c=this->ptr();
int i=0;
int prev=0;
int count=0;
while(true) {
if (c[i]==0 || c[i]==p_splitter) {
if (p_slice==count) {
return substr(prev,i-prev);
} else {
count++;
prev=i+1;
}
}
i++;
}
return String(); //no find!
}
Vector<String> String::split_spaces() const {
Vector<String> ret;
int from=0;
int i=0;
int len = length();
bool inside=false;
while(true) {
bool empty=operator[](i)<33;
if (i==0)
inside=!empty;
if (!empty && !inside) {
inside=true;
from=i;
}
if (empty && inside) {
ret.push_back(substr(from,i-from));
inside=false;
}
if (i==len)
break;
i++;
}
return ret;
}
Vector<String> String::split(const String &p_splitter,bool p_allow_empty) const {
Vector<String> ret;
int from=0;
int len = length();
while(true) {
int end=find(p_splitter,from);
if (end<0)
end=len;
if (p_allow_empty || (end>from))
ret.push_back(substr(from,end-from));
if (end==len)
break;
from = end+p_splitter.length();
}
return ret;
}
Vector<float> String::split_floats(const String &p_splitter,bool p_allow_empty) const {
Vector<float> ret;
int from=0;
int len = length();
while(true) {
int end=find(p_splitter,from);
if (end<0)
end=len;
if (p_allow_empty || (end>from))
ret.push_back(String::to_double(&c_str()[from]));
if (end==len)
break;
from = end+p_splitter.length();
}
return ret;
}
Vector<float> String::split_floats_mk(const Vector<String> &p_splitters,bool p_allow_empty) const {
Vector<float> ret;
int from=0;
int len = length();
while(true) {
int idx;
int end=findmk(p_splitters,from,&idx);
int spl_len=1;
if (end<0) {
end=len;
} else {
spl_len=p_splitters[idx].length();
}
if (p_allow_empty || (end>from)) {
ret.push_back(String::to_double(&c_str()[from]));
}
if (end==len)
break;
from = end+spl_len;
}
return ret;
}
Vector<int> String::split_ints(const String &p_splitter,bool p_allow_empty) const {
Vector<int> ret;
int from=0;
int len = length();
while(true) {
int end=find(p_splitter,from);
if (end<0)
end=len;
if (p_allow_empty || (end>from))
ret.push_back(String::to_int(&c_str()[from],end-from));
if (end==len)
break;
from = end+p_splitter.length();
}
return ret;
}
Vector<int> String::split_ints_mk(const Vector<String> &p_splitters,bool p_allow_empty) const {
Vector<int> ret;
int from=0;
int len = length();
while(true) {
int idx;
int end=findmk(p_splitters,from,&idx);
int spl_len=1;
if (end<0) {
end=len;
} else {
spl_len=p_splitters[idx].length();
}
if (p_allow_empty || (end>from))
ret.push_back(String::to_int(&c_str()[from],end-from));
if (end==len)
break;
from = end+spl_len;
}
return ret;
}
CharType String::char_uppercase(CharType p_char) {
return _find_upper(p_char);
}
CharType String::char_lowercase(CharType p_char){
return _find_lower(p_char);
}
String String::to_upper() const {
String upper=*this;
for(int i=0;i<upper.size();i++) {
upper[i]=_find_upper(upper[i]);
}
return upper;
}
String String::to_lower() const {
String upper=*this;
for(int i=0;i<upper.size();i++) {
upper[i]=_find_lower(upper[i]);
}
return upper;
}
int String::length() const {
int s=size();
return s?(s-1):0; // length does not include zero
}
const CharType * String::c_str() const {
static const CharType zero=0;
return size()?&operator[](0):&zero;
}
String String::md5(const uint8_t *p_md5) {
return String::hex_encode_buffer(p_md5, 16);
}
String String::hex_encode_buffer(const uint8_t *p_buffer, int p_len) {
static const char hex[16]={'0','1','2','3','4','5','6','7','8','9','a','b','c','d','e','f'};
String ret;
char v[2]={0,0};
for(int i=0;i<p_len;i++) {
v[0]=hex[p_buffer[i]>>4];
ret+=v;
v[0]=hex[p_buffer[i]&0xF];
ret+=v;
}
return ret;
}
String String::chr(CharType p_char) {
CharType c[2]={p_char,0};
return String(c);
}
String String::num(double p_num,int p_decimals) {
#ifndef NO_USE_STDLIB
if (p_decimals>12)
p_decimals=12;
char fmt[7];
fmt[0]='%';
fmt[1]='.';
if (p_decimals<0) {
fmt[1]='l';
fmt[2]='f';
fmt[3]=0;
} else if(p_decimals<10) {
fmt[2]='0'+p_decimals;
fmt[3]='l';
fmt[4]='f';
fmt[5]=0;
} else {
fmt[2]='0'+(p_decimals/10);
fmt[3]='0'+(p_decimals%10);
fmt[4]='l';
fmt[5]='f';
fmt[6]=0;
}
char buf[256];
#if defined(__GNUC__) || defined(_MSC_VER)
snprintf(buf,256,fmt,p_num);
#else
sprintf(buf,fmt,p_num);
#endif
buf[255]=0;
//destroy trailing zeroes
{
bool period=false;
int z=0;
while(buf[z]) {
if (buf[z]=='.')
period=true;
z++;
}
if (period) {
z--;
while(z>0) {
if (buf[z]=='0') {
buf[z]=0;
} else if (buf[z]=='.') {
buf[z]=0;
break;
} else {
break;
}
z--;
}
}
}
return buf;
#else
String s;
String sd;
/* integer part */
bool neg=p_num<0;
p_num=ABS(p_num);
int intn=(int)p_num;
/* decimal part */
if (p_decimals>0 || (p_decimals==-1 && (int)p_num!=p_num)) {
double dec=p_num-(float)((int)p_num);
int digit=0;
if (p_decimals>MAX_DIGITS)
p_decimals=MAX_DIGITS;
int dec_int=0;
int dec_max=0;
while (true) {
dec*=10.0;
dec_int=dec_int*10+(int)dec%10;
dec_max=dec_max*10+9;
digit++;
if (p_decimals==-1) {
if (digit==MAX_DIGITS) //no point in going to infinite
break;
if ((dec-(float)((int)dec))<1e-6)
break;
}
if (digit==p_decimals)
break;
}
dec*=10;
int last=(int)dec%10;
if (last>5) {
if (dec_int==dec_max) {
dec_int=0;
intn++;
} else {
dec_int++;
}
}
String decimal;
for (int i=0;i<digit;i++) {
char num[2]={0,0};
num[0]='0'+dec_int%10;
decimal=num+decimal;
dec_int/=10;
}
sd='.'+decimal;
}
if (intn==0)
s="0";
else {
while(intn) {
CharType num='0'+(intn%10);
intn/=10;
s=num+s;
}
}
s=s+sd;
if (neg)
s="-"+s;
return s;
#endif
}
String String::num_int64(int64_t p_num, int base, bool capitalize_hex) {
bool sign=p_num<0;
int64_t num=ABS(p_num);
int64_t n=num;
int chars=0;
do {
n/=base;
chars++;
} while(n);
if (sign)
chars++;
String s;
s.resize(chars+1);
CharType *c=s.ptr();
c[chars]=0;
n=num;
do {
int mod = n%base;
if (mod >= 10) {
char a = (capitalize_hex ? 'A' : 'a');
c[--chars]=a+(mod - 10);
} else {
c[--chars]='0'+mod;
}
n/=base;
} while(n);
if (sign)
c[0]='-';
return s;
}
String String::num_real(double p_num) {
String s;
String sd;
/* integer part */
bool neg=p_num<0;
p_num=ABS(p_num);
int intn=(int)p_num;
/* decimal part */
if ((int)p_num!=p_num) {
double dec=p_num-(float)((int)p_num);
int digit=0;
int decimals=MAX_DIGITS;
int dec_int=0;
int dec_max=0;
while (true) {
dec*=10.0;
dec_int=dec_int*10+(int)dec%10;
dec_max=dec_max*10+9;
digit++;
if ((dec-(float)((int)dec))<1e-6)
break;
if (digit==decimals)
break;
}
dec*=10;
int last=(int)dec%10;
if (last>5) {
if (dec_int==dec_max) {
dec_int=0;
intn++;
} else {
dec_int++;
}
}
String decimal;
for (int i=0;i<digit;i++) {
char num[2]={0,0};
num[0]='0'+dec_int%10;
decimal=num+decimal;
dec_int/=10;
}
sd='.'+decimal;
} else {
sd=".0";
}
if (intn==0)
s="0";
else {
while(intn) {
CharType num='0'+(intn%10);
intn/=10;
s=num+s;
}
}
s=s+sd;
if (neg)
s="-"+s;
return s;
}
String String::num_scientific(double p_num) {
#ifndef NO_USE_STDLIB
char buf[256];
#if defined(__GNUC__) || defined(_MSC_VER)
snprintf(buf,256,"%lg",p_num);
#else
sprintf(buf,"%.16lg",p_num);
#endif
buf[255]=0;
return buf;
#else
return String::num(p_num);
#endif
}
CharString String::ascii(bool p_allow_extended) const {
if (!length())
return CharString();
CharString cs;
cs.resize(size());
for (int i=0;i<size();i++)
cs[i]=operator[](i);
return cs;
}
String String::utf8(const char* p_utf8,int p_len) {
String ret;
ret.parse_utf8(p_utf8,p_len);
return ret;
};
#if 0
_FORCE_INLINE static int parse_utf8_char(const char *p_utf8,unsigned int *p_ucs4,int p_left) { //return len
int len=0;
/* Determine the number of characters in sequence */
if ((*p_utf8 & 0x80)==0)
len=1;
else if ((*p_utf8 & 0xE0)==0xC0)
len=2;
else if ((*p_utf8 & 0xF0)==0xE0)
len=3;
else if ((*p_utf8 & 0xF8)==0xF0)
len=4;
else if ((*p_utf8 & 0xFC)==0xF8)
len=5;
else if ((*p_utf8 & 0xFE)==0xFC)
len=6;
else
return -1; //invalid UTF8
if (len>p_left)
return -1; //not enough space
if (len==2 && (*p_utf8&0x1E)==0) {
//printf("overlong rejected\n");
return -1; //reject overlong
}
/* Convert the first character */
unsigned int unichar=0;
if (len == 1)
unichar=*p_utf8;
else {
unichar=(0xFF >> (len +1)) & *p_utf8;;
for (int i=1;i<len;i++) {
if ((p_utf8[i] & 0xC0) != 0x80) {
//printf("invalid utf8\n");
return -1; //invalid utf8
}
if (unichar==0 && i==2 && ((p_utf8[i] & 0x7F) >> (7 - len)) == 0) {
//printf("no overlong\n");
return -1; //no overlong
}
unichar = (unichar << 6) | (p_utf8[i] & 0x3F);
}
}
*p_ucs4=unichar;
return len;
}
#endif
bool String::parse_utf8(const char* p_utf8,int p_len) {
#define _UNICERROR(m_err) print_line("unicode error: "+String(m_err));
String aux;
int cstr_size=0;
int str_size=0;
/* HANDLE BOM (Byte Order Mark) */
if (p_len<0 || p_len>=3) {
bool has_bom = uint8_t(p_utf8[0])==0xEF && uint8_t(p_utf8[1])==0xBB && uint8_t(p_utf8[2])==0xBF;
if (has_bom) {
//just skip it
if (p_len>=0)
p_len-=3;
p_utf8+=3;
}
}
{
const char *ptrtmp=p_utf8;
const char *ptrtmp_limit=&p_utf8[p_len];
int skip=0;
while (ptrtmp!=ptrtmp_limit && *ptrtmp) {
if (skip==0) {
uint8_t c = *ptrtmp;
/* Determine the number of characters in sequence */
if ((c & 0x80)==0)
skip=0;
else if ((c & 0xE0)==0xC0)
skip=1;
else if ((c & 0xF0)==0xE0)
skip=2;
else if ((c & 0xF8)==0xF0)
skip=3;
else if ((c & 0xFC)==0xF8)
skip=4;
else if ((c & 0xFE)==0xFC)
skip=5;
else {
_UNICERROR("invalid skip");
return true; //invalid utf8
}
if (skip==1 && (c&0x1E)==0) {
//printf("overlong rejected\n");
_UNICERROR("overlong rejected");
return true; //reject overlong
}
str_size++;
} else {
--skip;
}
cstr_size++;
ptrtmp++;
}
if (skip) {
_UNICERROR("no space left");
return true; //not enough spac
}
}
if (str_size==0) {
clear();
return false;
}
resize(str_size+1);
CharType *dst = &operator[](0);
dst[str_size]=0;
while (cstr_size) {
int len=0;
/* Determine the number of characters in sequence */
if ((*p_utf8 & 0x80)==0)
len=1;
else if ((*p_utf8 & 0xE0)==0xC0)
len=2;
else if ((*p_utf8 & 0xF0)==0xE0)
len=3;
else if ((*p_utf8 & 0xF8)==0xF0)
len=4;
else if ((*p_utf8 & 0xFC)==0xF8)
len=5;
else if ((*p_utf8 & 0xFE)==0xFC)
len=6;
else {
_UNICERROR("invalid len");
return true; //invalid UTF8
}
if (len>cstr_size) {
_UNICERROR("no space left");
return true; //not enough space
}
if (len==2 && (*p_utf8&0x1E)==0) {
//printf("overlong rejected\n");
_UNICERROR("no space left");
return true; //reject overlong
}
/* Convert the first character */
uint32_t unichar=0;
if (len == 1)
unichar=*p_utf8;
else {
unichar=(0xFF >> (len +1)) & *p_utf8;;
for (int i=1;i<len;i++) {
if ((p_utf8[i] & 0xC0) != 0x80) {
_UNICERROR("invalid utf8");
return true; //invalid utf8
}
if (unichar==0 && i==2 && ((p_utf8[i] & 0x7F) >> (7 - len)) == 0) {
_UNICERROR("invalid utf8 overlong");
return true; //no overlong
}
unichar = (unichar << 6) | (p_utf8[i] & 0x3F);
}
}
// printf("char %i, len %i\n",unichar,len);
if (sizeof(wchar_t)==2 && unichar>0xFFFF) {
unichar=' '; //too long for windows
}
*(dst++) = unichar;
cstr_size-=len;
p_utf8 += len;
}
return false;
}
CharString String::utf8() const {
int l = length();
if (!l)
return CharString();
const CharType *d=&operator[](0);
int fl=0;
for (int i=0;i<l;i++) {
uint32_t c=d[i];
if (c <= 0x7f) // 7 bits.
fl+=1;
else if (c <= 0x7ff) { // 11 bits
fl+=2;
} else if (c <= 0xffff) { // 16 bits
fl+=3;
} else if (c <= 0x001fffff) { // 21 bits
fl+=4;
} else if (c <= 0x03ffffff) { // 26 bits
fl+=5;
} else if (c <= 0x7fffffff) { // 31 bits
fl+=6;
}
}
CharString utf8s;
if (fl==0) {
return utf8s;
}
utf8s.resize(fl+1);
uint8_t *cdst=(uint8_t*)utf8s.get_data();
#define APPEND_CHAR( m_c ) *(cdst++) = m_c
for (int i=0;i<l;i++) {
uint32_t c=d[i];
if (c <= 0x7f) // 7 bits.
APPEND_CHAR( c );
else if (c <= 0x7ff) { // 11 bits
APPEND_CHAR( uint32_t(0xc0 | ((c >> 6) & 0x1f)) ); // Top 5 bits.
APPEND_CHAR( uint32_t(0x80 | (c & 0x3f)) ); // Bottom 6 bits.
} else if (c <= 0xffff) { // 16 bits
APPEND_CHAR( uint32_t(0xe0 | ((c >> 12) & 0x0f)) ); // Top 4 bits.
APPEND_CHAR( uint32_t(0x80 | ((c >> 6) & 0x3f)) ); // Middle 6 bits.
APPEND_CHAR( uint32_t(0x80 | (c & 0x3f)) ); // Bottom 6 bits.
} else if (c <= 0x001fffff) { // 21 bits
APPEND_CHAR( uint32_t(0xf0 | ((c >> 18) & 0x07)) ); // Top 3 bits.
APPEND_CHAR( uint32_t(0x80 | ((c >> 12) & 0x3f)) ); // Upper middle 6 bits.
APPEND_CHAR( uint32_t(0x80 | ((c >> 6) & 0x3f)) ); // Lower middle 6 bits.
APPEND_CHAR( uint32_t(0x80 | (c & 0x3f)) ); // Bottom 6 bits.
} else if (c <= 0x03ffffff) { // 26 bits
APPEND_CHAR( uint32_t(0xf8 | ((c >> 24) & 0x03)) ); // Top 2 bits.
APPEND_CHAR( uint32_t(0x80 | ((c >> 18) & 0x3f)) ); // Upper middle 6 bits.
APPEND_CHAR( uint32_t(0x80 | ((c >> 12) & 0x3f)) ); // middle 6 bits.
APPEND_CHAR( uint32_t(0x80 | ((c >> 6) & 0x3f)) ); // Lower middle 6 bits.
APPEND_CHAR( uint32_t(0x80 | (c & 0x3f)) ); // Bottom 6 bits.
} else if (c <= 0x7fffffff) { // 31 bits
APPEND_CHAR( uint32_t(0xfc | ((c >> 30) & 0x01)) ); // Top 1 bit.
APPEND_CHAR( uint32_t(0x80 | ((c >> 24) & 0x3f)) ); // Upper upper middle 6 bits.
APPEND_CHAR( uint32_t(0x80 | ((c >> 18) & 0x3f)) ); // Lower upper middle 6 bits.
APPEND_CHAR( uint32_t(0x80 | ((c >> 12) & 0x3f)) ); // Upper lower middle 6 bits.
APPEND_CHAR( uint32_t(0x80 | ((c >> 6) & 0x3f)) ); // Lower lower middle 6 bits.
APPEND_CHAR( uint32_t(0x80 | (c & 0x3f)) ); // Bottom 6 bits.
}
}
#undef APPEND_CHAR
*cdst=0; //trailing zero
return utf8s;
}
/*
String::String(CharType p_char) {
shared=NULL;
copy_from(p_char);
}
*/
String::String(const char *p_str) {
copy_from(p_str);
}
String::String(const CharType *p_str,int p_clip_to_len) {
copy_from(p_str,p_clip_to_len);
}
String::String(const StrRange& p_range) {
if (!p_range.c_str)
return;
copy_from(p_range.c_str,p_range.len);
}
int String::hex_to_int(bool p_with_prefix) const {
int l = length();
if (p_with_prefix && l<3)
return 0;
const CharType *s=ptr();
int sign = s[0]=='-' ? -1 : 1;
if (sign<0) {
s++;
l--;
if (p_with_prefix && l<2)
return 0;
}
if (p_with_prefix) {
if (s[0]!='0' || s[1]!='x')
return 0;
s+=2;
l-=2;
};
int hex=0;
while(*s) {
CharType c = LOWERCASE(*s);
int n;
if (c>='0' && c<='9') {
n=c-'0';
} else if (c>='a' && c<='f') {
n=(c-'a')+10;
} else {
return 0;
}
hex*=16;
hex+=n;
s++;
}
return hex*sign;
}
int String::to_int() const {
if (length()==0)
return 0;
int to=(find(".")>=0) ? find(".") : length() ;
int integer=0;
int sign=1;
for (int i=0;i<to;i++) {
CharType c=operator[](i);
if (c>='0' && c<='9') {
integer*=10;
integer+=c-'0';
} else if (integer==0 && c=='-') {
sign=-sign;
}
}
return integer*sign;
}
int64_t String::to_int64() const {
if (length()==0)
return 0;
int to=(find(".")>=0) ? find(".") : length() ;
int64_t integer=0;
int64_t sign=1;
for (int i=0;i<to;i++) {
CharType c=operator[](i);
if (c>='0' && c<='9') {
integer*=10;
integer+=c-'0';
} else if (integer==0 && c=='-') {
sign=-sign;
}
}
return integer*sign;
}
int String::to_int(const char* p_str,int p_len) {
int to=0;
if (p_len>=0)
to=p_len;
else {
while(p_str[to]!=0 && p_str[to]!='.')
to++;
}
int integer=0;
int sign=1;
for (int i=0;i<to;i++) {
char c=p_str[i];
if (c>='0' && c<='9') {
integer*=10;
integer+=c-'0';
} else if (c=='-' && integer==0) {
sign=-sign;
} else if (c!=' ')
break;
}
return integer*sign;
}
bool String::is_numeric() const {
if (length() == 0) {
return false;
};
int s = 0;
if (operator[](0) == '-') ++s;
bool dot = false;
for (int i=s; i<length(); i++) {
CharType c = operator[](i);
if (c == '.') {
if (dot) {
return false;
};
dot = true;
}
if (c < '0' || c > '9') {
return false;
};
};
return true; // TODO: Use the parser below for this instead
};
#define IS_DIGIT(m_d) ( (m_d)>='0' && (m_d)<='9' )
#define IS_HEX_DIGIT(m_d) ( ( (m_d)>='0' && (m_d)<='9' ) || ( (m_d)>='a' && (m_d)<='f' ) || ( (m_d)>='A' && (m_d)<='F' ) )
template<class C>
static double built_in_strtod(const C *string, /* A decimal ASCII floating-point number,
* optionally preceded by white space. Must
* have form "-I.FE-X", where I is the integer
* part of the mantissa, F is the fractional
* part of the mantissa, and X is the
* exponent. Either of the signs may be "+",
* "-", or omitted. Either I or F may be
* omitted, or both. The decimal point isn't
* necessary unless F is present. The "E" may
* actually be an "e". E and X may both be
* omitted (but not just one). */
C **endPtr=NULL) /* If non-NULL, store terminating Cacter's
* address here. */
{
static const int maxExponent = 511; /* Largest possible base 10 exponent. Any
* exponent larger than this will already
* produce underflow or overflow, so there's
* no need to worry about additional digits.
*/
static const double powersOf10[] = { /* Table giving binary powers of 10. Entry */
10., /* is 10^2^i. Used to convert decimal */
100., /* exponents into floating-point numbers. */
1.0e4,
1.0e8,
1.0e16,
1.0e32,
1.0e64,
1.0e128,
1.0e256
};
int sign, expSign = false;
double fraction, dblExp;
const double *d;
register const C *p;
register int c;
int exp = 0; /* Exponent read from "EX" field. */
int fracExp = 0; /* Exponent that derives from the fractional
* part. Under normal circumstatnces, it is
* the negative of the number of digits in F.
* However, if I is very long, the last digits
* of I get dropped (otherwise a long I with a
* large negative exponent could cause an
* unnecessary overflow on I alone). In this
* case, fracExp is incremented one for each
* dropped digit. */
int mantSize; /* Number of digits in mantissa. */
int decPt; /* Number of mantissa digits BEFORE decimal
* point. */
const C *pExp; /* Temporarily holds location of exponent in
* string. */
/*
* Strip off leading blanks and check for a sign.
*/
p = string;
while (*p == ' ' || *p=='\t' || *p=='\n') {
p += 1;
}
if (*p == '-') {
sign = true;
p += 1;
} else {
if (*p == '+') {
p += 1;
}
sign = false;
}
/*
* Count the number of digits in the mantissa (including the decimal
* point), and also locate the decimal point.
*/
decPt = -1;
for (mantSize = 0; ; mantSize += 1)
{
c = *p;
if (!IS_DIGIT(c)) {
if ((c != '.') || (decPt >= 0)) {
break;
}
decPt = mantSize;
}
p += 1;
}
/*
* Now suck up the digits in the mantissa. Use two integers to collect 9
* digits each (this is faster than using floating-point). If the mantissa
* has more than 18 digits, ignore the extras, since they can't affect the
* value anyway.
*/
pExp = p;
p -= mantSize;
if (decPt < 0) {
decPt = mantSize;
} else {
mantSize -= 1; /* One of the digits was the point. */
}
if (mantSize > 18) {
fracExp = decPt - 18;
mantSize = 18;
} else {
fracExp = decPt - mantSize;
}
if (mantSize == 0) {
fraction = 0.0;
p = string;
goto done;
} else {
int frac1, frac2;
frac1 = 0;
for ( ; mantSize > 9; mantSize -= 1) {
c = *p;
p += 1;
if (c == '.') {
c = *p;
p += 1;
}
frac1 = 10*frac1 + (c - '0');
}
frac2 = 0;
for (; mantSize > 0; mantSize -= 1) {
c = *p;
p += 1;
if (c == '.') {
c = *p;
p += 1;
}
frac2 = 10*frac2 + (c - '0');
}
fraction = (1.0e9 * frac1) + frac2;
}
/*
* Skim off the exponent.
*/
p = pExp;
if ((*p == 'E') || (*p == 'e')) {
p += 1;
if (*p == '-') {
expSign = true;
p += 1;
} else {
if (*p == '+') {
p += 1;
}
expSign = false;
}
if (!IS_DIGIT(CharType(*p))) {
p = pExp;
goto done;
}
while (IS_DIGIT(CharType(*p))) {
exp = exp * 10 + (*p - '0');
p += 1;
}
}
if (expSign) {
exp = fracExp - exp;
} else {
exp = fracExp + exp;
}
/*
* Generate a floating-point number that represents the exponent. Do this
* by processing the exponent one bit at a time to combine many powers of
* 2 of 10. Then combine the exponent with the fraction.
*/
if (exp < 0) {
expSign = true;
exp = -exp;
} else {
expSign = false;
}
if (exp > maxExponent) {
exp = maxExponent;
WARN_PRINT("Exponent too high");
}
dblExp = 1.0;
for (d = powersOf10; exp != 0; exp >>= 1, ++d) {
if (exp & 01) {
dblExp *= *d;
}
}
if (expSign) {
fraction /= dblExp;
} else {
fraction *= dblExp;
}
done:
if (endPtr != NULL) {
*endPtr = (C *) p;
}
if (sign) {
return -fraction;
}
return fraction;
}
#define READING_SIGN 0
#define READING_INT 1
#define READING_DEC 2
#define READING_EXP 3
#define READING_DONE 4
double String::to_double(const char* p_str) {
#ifndef NO_USE_STDLIB
return built_in_strtod<char>(p_str);
//return atof(p_str); DOES NOT WORK ON ANDROID(??)
#else
return built_in_strtod<char>(p_str);
#endif
#if 0
#if 0
return atof(p_str);
#else
if (!p_str[0])
return 0;
///@todo make more exact so saving and loading does not lose precision
double integer=0;
double decimal=0;
double decimal_mult=0.1;
double sign=1.0;
double exp=0;
double exp_sign=1.0;
int reading=READING_SIGN;
const char *str=p_str;
while(*str && reading!=READING_DONE) {
CharType c=*(str++);
switch(reading) {
case READING_SIGN: {
if (c>='0' && c<='9')
reading=READING_INT;
// let it fallthrough
else if (c=='-') {
sign=-1.0;
reading=READING_INT;
break;
} else if (c=='.') {
reading=READING_DEC;
break;
} else {
break;
}
}
case READING_INT: {
if (c>='0' && c<='9') {
integer*=10;
integer+=c-'0';
} else if (c=='.') {
reading=READING_DEC;
} else if (c=='e') {
reading=READING_EXP;
} else {
reading=READING_DONE;
}
} break;
case READING_DEC: {
if (c>='0' && c<='9') {
decimal+=(c-'0')*decimal_mult;
decimal_mult*=0.1;
} else if (c=='e') {
reading=READING_EXP;
} else {
reading=READING_DONE;
}
} break;
case READING_EXP: {
if (c>='0' && c<='9') {
exp*=10.0;
exp+=(c-'0');
} else if (c=='-' && exp==0) {
exp_sign=-1.0;
} else if (exp_sign>=0 && c=='+') {
//redundant...
exp_sign=1.0;
} else {
reading=READING_DONE;
}
} break;
}
}
return sign*(integer+decimal)*Math::pow(10,exp_sign*exp);
#endif
#endif
}
float String::to_float() const {
return to_double();
}
double String::to_double(const CharType* p_str, const CharType **r_end) {
return built_in_strtod<CharType>(p_str,(CharType**)r_end);
#if 0
#if 0
//ndef NO_USE_STDLIB
return wcstod(p_str,p_len<0?NULL:p_str+p_len);
#else
if (p_len==0 || !p_str[0])
return 0;
///@todo make more exact so saving and loading does not lose precision
double integer=0;
double decimal=0;
double decimal_mult=0.1;
double sign=1.0;
double exp=0;
double exp_sign=1.0;
int reading=READING_SIGN;
const CharType *str=p_str;
const CharType *limit=&p_str[p_len];
while(reading!=READING_DONE && str!=limit) {
CharType c=*(str++);
switch(reading) {
case READING_SIGN: {
if (c>='0' && c<='9')
reading=READING_INT;
// let it fallthrough
else if (c=='-') {
sign=-1.0;
reading=READING_INT;
break;
} else if (c=='.') {
reading=READING_DEC;
break;
} else if (c==0) {
reading=READING_DONE;
break;
} else {
break;
}
}
case READING_INT: {
if (c>='0' && c<='9') {
integer*=10;
integer+=c-'0';
} else if (c=='.') {
reading=READING_DEC;
} else if (c=='e' || c=='E') {
reading=READING_EXP;
} else {
reading=READING_DONE;
}
} break;
case READING_DEC: {
if (c>='0' && c<='9') {
decimal+=(c-'0')*decimal_mult;
decimal_mult*=0.1;
} else if (c=='e' || c=='E') {
reading=READING_EXP;
} else {
reading=READING_DONE;
}
} break;
case READING_EXP: {
if (c>='0' && c<='9') {
exp*=10.0;
exp+=(c-'0');
} else if (c=='-' && exp==0) {
exp_sign=-1.0;
} else if (exp_sign>=0 && c=='+') {
//redundant...
exp_sign=1.0;
} else {
reading=READING_DONE;
}
} break;
}
}
if (r_end)
*r_end=str-1;
return sign*(integer+decimal)*Math::pow(10,exp_sign*exp);
#endif
#endif
}
int64_t String::to_int(const CharType* p_str,int p_len) {
if (p_len==0 || !p_str[0])
return 0;
///@todo make more exact so saving and loading does not lose precision
int64_t integer=0;
int64_t sign=1;
int reading=READING_SIGN;
const CharType *str=p_str;
const CharType *limit=&p_str[p_len];
while(*str && reading!=READING_DONE && str!=limit) {
CharType c=*(str++);
switch(reading) {
case READING_SIGN: {
if (c>='0' && c<='9') {
reading=READING_INT;
// let it fallthrough
} else if (c=='-') {
sign=-1;
reading=READING_INT;
break;
} else if (c=='+') {
sign=1;
reading=READING_INT;
break;
} else {
break;
}
}
case READING_INT: {
if (c>='0' && c<='9') {
integer*=10;
integer+=c-'0';
} else {
reading=READING_DONE;
}
} break;
}
}
return sign*integer;
}
double String::to_double() const {
if (empty())
return 0;
#ifndef NO_USE_STDLIB
return built_in_strtod<CharType>(c_str());
//return wcstod(c_str(),NULL); DOES NOT WORK ON ANDROID :(
#else
return built_in_strtod<CharType>(c_str());
#endif
#if 0
#ifndef NO_USE_STDLIB
return atof(utf8().get_data());
#else
double integer=0;
double decimal=0;
double decimal_mult=0.1;
double sign=1.0;
double exp=0;
double exp_sign=1.0;
int reading=READING_SIGN;
const CharType *str=&operator[](0);
while(*str && reading!=READING_DONE) {
CharType c=*(str++);
switch(reading) {
case READING_SIGN: {
if (c>='0' && c<='9')
reading=READING_INT;
// let it fallthrough
else if (c=='-') {
sign=-1.0;
reading=READING_INT;
break;
} else if (c=='.') {
reading=READING_DEC;
break;
} else {
break;
}
}
case READING_INT: {
if (c>='0' && c<='9') {
integer*=10;
integer+=c-'0';
} else if (c=='.') {
reading=READING_DEC;
} else if (c=='e') {
reading=READING_EXP;
} else {
reading=READING_DONE;
}
} break;
case READING_DEC: {
if (c>='0' && c<='9') {
decimal+=(c-'0')*decimal_mult;
decimal_mult*=0.1;
} else if (c=='e') {
reading=READING_EXP;
} else {
reading=READING_DONE;
}
} break;
case READING_EXP: {
if (c>='0' && c<='9') {
exp*=10.0;
exp+=(c-'0');
} else if (c=='-' && exp==0) {
exp_sign=-1.0;
} else if (exp_sign>=0 && c=='+') {
//redundant...
exp_sign=1.0;
} else {
reading=READING_DONE;
}
} break;
}
}
return sign*(integer+decimal)*Math::pow(10,exp_sign*exp);
#endif
#if 0
double ret=sign*(integer+decimal)*Math::pow(10,exp_sign*exp);
print_line(*this +" == "+rtos(ret));
return ret;
#endif
#endif
}
bool operator==(const char*p_chr, const String& p_str) {
return p_str==p_chr;
}
String operator+(const char*p_chr, const String& p_str) {
String tmp=p_chr;
tmp+=p_str;
return tmp;
}
String operator+(CharType p_chr, const String& p_str) {
return (String::chr(p_chr)+p_str);
}
uint32_t String::hash(const char* p_cstr) {
uint32_t hashv = 5381;
uint32_t c;
while ((c = *p_cstr++))
hashv = ((hashv << 5) + hashv) + c; /* hash * 33 + c */
return hashv;
}
uint32_t String::hash(const char* p_cstr,int p_len) {
uint32_t hashv = 5381;
for(int i=0;i<p_len;i++)
hashv = ((hashv << 5) + hashv) + p_cstr[i]; /* hash * 33 + c */
return hashv;
}
uint32_t String::hash(const CharType* p_cstr,int p_len) {
uint32_t hashv = 5381;
for(int i=0;i<p_len;i++)
hashv = ((hashv << 5) + hashv) + p_cstr[i]; /* hash * 33 + c */
return hashv;
}
uint32_t String::hash(const CharType* p_cstr) {
uint32_t hashv = 5381;
uint32_t c;
while ((c = *p_cstr++))
hashv = ((hashv << 5) + hashv) + c; /* hash * 33 + c */
return hashv;
}
uint32_t String::hash() const {
/* simple djb2 hashing */
const CharType * chr = c_str();
uint32_t hashv = 5381;
uint32_t c;
while ((c = *chr++))
hashv = ((hashv << 5) + hashv) + c; /* hash * 33 + c */
return hashv;
}
uint64_t String::hash64() const {
/* simple djb2 hashing */
const CharType * chr = c_str();
uint64_t hashv = 5381;
uint64_t c;
while ((c = *chr++))
hashv = ((hashv << 5) + hashv) + c; /* hash * 33 + c */
return hashv;
}
String String::md5_text() const {
CharString cs=utf8();
MD5_CTX ctx;
MD5Init(&ctx);
MD5Update(&ctx,(unsigned char*)cs.ptr(),cs.length());
MD5Final(&ctx);
return String::md5(ctx.digest);
}
String String::sha256_text() const {
CharString cs=utf8();
unsigned char hash[32];
sha256_context ctx;
sha256_init(&ctx);
sha256_hash(&ctx,(unsigned char*)cs.ptr(),cs.length());
sha256_done(&ctx, hash);
return String::hex_encode_buffer(hash, 32);
}
Vector<uint8_t> String::md5_buffer() const {
CharString cs=utf8();
MD5_CTX ctx;
MD5Init(&ctx);
MD5Update(&ctx,(unsigned char*)cs.ptr(),cs.length());
MD5Final(&ctx);
Vector<uint8_t> ret;
ret.resize(16);
for (int i=0; i<16; i++) {
ret[i] = ctx.digest[i];
};
return ret;
};
Vector<uint8_t> String::sha256_buffer() const {
CharString cs = utf8();
unsigned char hash[32];
sha256_context ctx;
sha256_init(&ctx);
sha256_hash(&ctx, (unsigned char*)cs.ptr(), cs.length());
sha256_done(&ctx, hash);
Vector<uint8_t> ret;
ret.resize(32);
for (int i = 0; i < 32; i++) {
ret[i] = hash[i];
}
return ret;
}
String String::insert(int p_at_pos,String p_string) const {
if (p_at_pos<0)
return *this;
if (p_at_pos>length())
p_at_pos=length();
String pre;
if (p_at_pos>0)
pre=substr( 0, p_at_pos );
String post;
if (p_at_pos<length())
post=substr( p_at_pos, length()-p_at_pos);
return pre+p_string+post;
}
String String::substr(int p_from,int p_chars) const{
if (empty() || p_from<0 || p_from>=length() || p_chars<=0)
return "";
if ( (p_from+p_chars)>length()) {
p_chars=length()-p_from;
}
return String(&c_str()[p_from],p_chars);
}
int String::find_last(String p_str) const {
int pos=-1;
int findfrom=0;
int findres=-1;
while((findres=find(p_str,findfrom))!=-1) {
pos=findres;
findfrom=pos+1;
}
return pos;
}
int String::find(String p_str,int p_from) const {
if (p_from<0)
return -1;
int src_len=p_str.length();
int len=length();
if(src_len==0 || len==0)
return -1; //wont find anything!
const CharType *src = c_str();
for (int i=p_from;i<=(len-src_len);i++) {
bool found=true;
for (int j=0;j<src_len;j++) {
int read_pos=i+j;
if (read_pos>=len) {
ERR_PRINT("read_pos>=len");
return -1;
};
if (src[read_pos]!=p_str[j]) {
found=false;
break;
}
}
if (found)
return i;
}
return -1;
}
int String::findmk(const Vector<String>& p_keys,int p_from,int *r_key) const {
if (p_from<0)
return -1;
if (p_keys.size()==0)
return -1;
//int src_len=p_str.length();
const String *keys=&p_keys[0];
int key_count=p_keys.size();
int len = length();
if(len==0)
return -1; //wont find anything!
const CharType *src = c_str();
for (int i=p_from;i<len;i++) {
bool found=true;
for(int k=0;k<key_count;k++) {
found=true;
if (r_key)
*r_key=k;
const CharType *cmp=keys[k].c_str();
int l=keys[k].length();
for (int j=0;j<l;j++) {
int read_pos=i+j;
if (read_pos>=len) {
found=false;
break;
};
if (src[read_pos]!=cmp[j]) {
found=false;
break;
}
}
if (found)
break;
}
if (found)
return i;
}
return -1;
}
int String::findn(String p_str,int p_from) const {
if (p_from<0)
return -1;
int src_len=p_str.length();
if(src_len==0 || length()==0)
return -1; //wont find anything!
const CharType *srcd = c_str();
for (int i=p_from;i<=(length()-src_len);i++) {
bool found=true;
for (int j=0;j<src_len;j++) {
int read_pos=i+j;
if (read_pos>=length()) {
ERR_PRINT("read_pos>=length()");
return -1;
};
CharType src=_find_lower(srcd[read_pos]);
CharType dst=_find_lower(p_str[j]);
if (src!=dst) {
found=false;
break;
}
}
if (found)
return i;
}
return -1;
}
int String::rfind(String p_str,int p_from) const {
//stabilish a limit
int limit = length()-p_str.length();
if (limit<0)
return -1;
//stabilish a starting point
if (p_from<0)
p_from=limit;
else if (p_from>limit)
p_from=limit;
int src_len=p_str.length();
int len=length();
if(src_len==0 || len==0)
return -1; //wont find anything!
const CharType *src = c_str();
for (int i=p_from;i>=0;i--) {
bool found=true;
for (int j=0;j<src_len;j++) {
int read_pos=i+j;
if (read_pos>=len) {
ERR_PRINT("read_pos>=len");
return -1;
};
if (src[read_pos]!=p_str[j]) {
found=false;
break;
}
}
if (found)
return i;
}
return -1;
}
int String::rfindn(String p_str,int p_from) const {
//stabilish a limit
int limit = length()-p_str.length();
if (limit<0)
return -1;
//stabilish a starting point
if (p_from<0)
p_from=limit;
else if (p_from>limit)
p_from=limit;
int src_len=p_str.length();
int len=length();
if(src_len==0 || len==0)
return -1; //wont find anything!
const CharType *src = c_str();
for (int i=p_from;i>=0;i--) {
bool found=true;
for (int j=0;j<src_len;j++) {
int read_pos=i+j;
if (read_pos>=len) {
ERR_PRINT("read_pos>=len");
return -1;
};
CharType srcc=_find_lower(src[read_pos]);
CharType dstc=_find_lower(p_str[j]);
if (srcc!=dstc) {
found=false;
break;
}
}
if (found)
return i;
}
return -1;
}
bool String::ends_with(const String& p_string) const {
int pos = find_last(p_string);
if (pos==-1)
return false;
return pos+p_string.length()==length();
}
bool String::begins_with(const String& p_string) const {
if (p_string.length() > length())
return false;
int l=p_string.length();
if (l==0)
return true;
const CharType *src=&p_string[0];
const CharType *str=&operator[](0);
int i = 0;
for (;i<l;i++) {
if (src[i]!=str[i])
return false;
}
// only if i == l the p_string matches the beginning
return i == l;
}
bool String::begins_with(const char* p_string) const {
int l=length();
if (l==0||!p_string)
return false;
const CharType *str=&operator[](0);
int i=0;
while (*p_string && i<l) {
if (*p_string != str[i])
return false;
i++;
p_string++;
}
return *p_string == 0;
}
bool String::is_subsequence_of(const String& p_string) const {
return _base_is_subsequence_of(p_string, false);
}
bool String::is_subsequence_ofi(const String& p_string) const {
return _base_is_subsequence_of(p_string, true);
}
bool String::_base_is_subsequence_of(const String& p_string, bool case_insensitive) const {
int len=length();
if (len == 0) {
// Technically an empty string is subsequence of any string
return true;
}
if (len > p_string.length()) {
return false;
}
const CharType *src = &operator[](0);
const CharType *tgt = &p_string[0];
for (;*src && *tgt;tgt++) {
bool match = false;
if (case_insensitive) {
CharType srcc = _find_lower(*src);
CharType tgtc = _find_lower(*tgt);
match = srcc == tgtc;
} else {
match = *src == *tgt;
}
if (match) {
src++;
if(!*src) {
return true;
}
}
}
return false;
}
Vector<String> String::bigrams() const {
int n_pairs = length() - 1;
Vector<String> b;
if(n_pairs <= 0) {
return b;
}
b.resize(n_pairs);
for(int i = 0; i < n_pairs; i++) {
b[i] = substr(i,2);
}
return b;
}
// Similarity according to Sorensen-Dice coefficient
float String::similarity(const String& p_string) const {
if(operator==(p_string)) {
// Equal strings are totally similar
return 1.0f;
}
if (length() < 2 || p_string.length() < 2) {
// No way to calculate similarity without a single bigram
return 0.0f;
}
Vector<String> src_bigrams = bigrams();
Vector<String> tgt_bigrams = p_string.bigrams();
int src_size = src_bigrams.size();
int tgt_size = tgt_bigrams.size();
float sum = src_size + tgt_size;
float inter = 0;
for (int i = 0; i < src_size; i++) {
for (int j = 0; j < tgt_size; j++) {
if (src_bigrams[i] == tgt_bigrams[j]) {
inter++;
break;
}
}
}
return (2.0f * inter)/sum;
}
static bool _wildcard_match(const CharType* p_pattern, const CharType* p_string,bool p_case_sensitive) {
switch (*p_pattern) {
case '\0':
return !*p_string;
case '*' :
return _wildcard_match(p_pattern+1, p_string,p_case_sensitive) || (*p_string && _wildcard_match(p_pattern, p_string+1,p_case_sensitive));
case '?' :
return *p_string && (*p_string != '.') && _wildcard_match(p_pattern+1, p_string+1,p_case_sensitive);
default :
return (p_case_sensitive?(*p_string==*p_pattern):(_find_upper(*p_string)==_find_upper(*p_pattern))) && _wildcard_match(p_pattern+1, p_string+1,p_case_sensitive);
}
}
bool String::match(const String& p_wildcard) const {
if (!p_wildcard.length() || !length())
return false;
return _wildcard_match(p_wildcard.c_str(),c_str(),true);
}
bool String::matchn(const String& p_wildcard) const {
if (!p_wildcard.length() || !length())
return false;
return _wildcard_match(p_wildcard.c_str(),c_str(),false);
}
String String::replace(String p_key,String p_with) const {
String new_string;
int search_from=0;
int result=0;
while( (result=find(p_key,search_from))>=0 ) {
new_string+=substr(search_from,result-search_from);
new_string+=p_with;
search_from=result+p_key.length();
}
new_string+=substr(search_from,length()-search_from);
return new_string;
}
String String::replace_first(String p_key,String p_with) const {
String new_string;
int search_from=0;
int result=0;
while( (result=find(p_key,search_from))>=0 ) {
new_string+=substr(search_from,result-search_from);
new_string+=p_with;
search_from=result+p_key.length();
break;
}
new_string+=substr(search_from,length()-search_from);
return new_string;
}
String String::replacen(String p_key,String p_with) const {
String new_string;
int search_from=0;
int result=0;
while( (result=findn(p_key,search_from))>=0 ) {
new_string+=substr(search_from,result-search_from);
new_string+=p_with;
search_from=result+p_key.length();
}
new_string+=substr(search_from,length()-search_from);
return new_string;
}
String String::left(int p_pos) const {
if (p_pos<=0)
return "";
if (p_pos>=length())
return *this;
return substr(0,p_pos);
}
String String::right(int p_pos) const {
if (p_pos>=size())
return *this;
if (p_pos<0)
return "";
return substr(p_pos,(length()-p_pos));
}
CharType String::ord_at(int p_idx) const {
ERR_FAIL_INDEX_V(p_idx, length(), 0);
return operator[](p_idx);
}
String String::strip_edges(bool left, bool right) const {
int len=length();
int beg=0,end=len;
if(left) {
for (int i=0;i<len;i++) {
if (operator[](i)<=32)
beg++;
else
break;
}
}
if(right) {
for (int i=(int)(len-1);i>=0;i--) {
if (operator[](i)<=32)
end--;
else
break;
}
}
if (beg==0 && end==len)
return *this;
return substr(beg,end-beg);
}
String String::strip_escapes() const {
int len=length();
int beg=0,end=len;
for (int i=0;i<length();i++) {
if (operator[](i)<=31)
beg++;
else
break;
}
for (int i=(int)(length()-1);i>=0;i--) {
if (operator[](i)<=31)
end--;
else
break;
}
if (beg==0 && end==len)
return *this;
return substr(beg,end-beg);
}
String String::simplify_path() const {
String s = *this;
String drive;
if (s.begins_with("local://")) {
drive="local://";
s=s.substr(8,s.length());
} else if (s.begins_with("res://")) {
drive="res://";
s=s.substr(6,s.length());
} else if (s.begins_with("user://")) {
drive="user://";
s=s.substr(6,s.length());
} else if (s.begins_with("/") || s.begins_with("\\")) {
drive=s.substr(0,1);
s=s.substr(1,s.length()-1);
} else {
int p = s.find(":/");
if (p==-1)
p=s.find(":\\");
if (p!=-1 && p < s.find("/")) {
drive=s.substr(0,p+2);
s=s.substr(p+2,s.length());
}
}
s =s.replace("\\","/");
while(true){ // in case of using 2 or more slash
String compare = s.replace("//","/");
if (s==compare) break;
else s=compare;
}
Vector<String> dirs = s.split("/",false);
for(int i=0;i<dirs.size();i++) {
String d = dirs[i];
if (d==".") {
dirs.remove(i);
i--;
} else if (d=="..") {
if (i==0) {
dirs.remove(i);
i--;
} else {
dirs.remove(i);
dirs.remove(i-1);
i-=2;
}
}
}
s="";
for(int i=0;i<dirs.size();i++) {
if (i>0)
s+="/";
s+=dirs[i];
}
return drive+s;
}
static int _humanize_digits(int p_num) {
if (p_num<10)
return 2;
else if (p_num<100)
return 2;
else if (p_num<1024)
return 1;
else
return 0;
}
String String::humanize_size(size_t p_size) {
uint64_t _div=1;
static const char* prefix[]={" Bytes"," KB"," MB"," GB","TB"," PB","HB",""};
int prefix_idx=0;
while(p_size>(_div*1024) && prefix[prefix_idx][0]) {
_div*=1024;
prefix_idx++;
}
int digits=prefix_idx>0?_humanize_digits(p_size/_div):0;
double divisor = prefix_idx>0?_div:1;
return String::num(p_size/divisor,digits)+prefix[prefix_idx];
}
bool String::is_abs_path() const {
if (length()>1)
return (operator[](0)=='/' || operator[](0)=='\\' || find(":/")!=-1 || find(":\\")!=-1);
else if ((length())==1)
return (operator[](0)=='/' || operator[](0)=='\\');
else
return false;
}
bool String::is_valid_identifier() const {
int len = length();
if (len==0)
return false;
const wchar_t * str = &operator[](0);
for(int i=0;i<len;i++) {
if (i==0) {
if (str[0]>='0' && str[0]<='9')
return false; // no start with number plz
}
bool valid_char = (str[i]>='0' && str[i]<='9') || (str[i]>='a' && str[i]<='z') || (str[i]>='A' && str[i]<='Z') || str[i]=='_';
if (!valid_char)
return false;
}
return true;
}
//kind of poor should be rewritten properly
String String::word_wrap(int p_chars_per_line) const {
int from=0;
int last_space=0;
String ret;
for(int i=0;i<length();i++) {
if (i-from>=p_chars_per_line) {
if (last_space==-1) {
ret+=substr(from,i-from+1)+"\n";
} else {
ret+=substr(from,last_space-from)+"\n";
i=last_space; //rewind
}
from=i+1;
last_space=-1;
} else if (operator[](i)==' ' || operator[](i)=='\t') {
last_space=i;
} else if (operator[](i)=='\n') {
ret+=substr(from,i-from)+"\n";
from=i+1;
last_space=-1;
}
}
if (from<length()) {
ret+=substr(from,length());
}
return ret;
}
String String::http_escape() const {
const CharString temp = utf8();
String res;
for (int i = 0; i < length(); ++i) {
CharType ord = temp[i];
if (ord == '.' || ord == '-' || ord == '_' || ord == '~' ||
(ord >= 'a' && ord <= 'z') ||
(ord >= 'A' && ord <= 'Z') ||
(ord >= '0' && ord <= '9')) {
res += ord;
} else {
char h_Val[3];
#if defined(__GNUC__) || defined(_MSC_VER)
snprintf(h_Val, 3, "%.2X", ord);
#else
sprintf(h_Val, "%.2X", ord);
#endif
res += "%";
res += h_Val;
}
}
return res;
}
String String::http_unescape() const {
String res;
for (int i = 0; i < length(); ++i) {
if (ord_at(i) == '%' && i+2 < length()) {
CharType ord1 = ord_at(i+1);
if ((ord1 >= '0' && ord1 <= '9') || (ord1 >= 'A' && ord1 <= 'Z')) {
CharType ord2 = ord_at(i+2);
if ((ord2 >= '0' && ord2 <= '9') || (ord2 >= 'A' && ord2 <= 'Z')) {
char bytes[2] = {ord1, ord2};
res += (char)strtol(bytes, NULL, 16);
i+=2;
}
} else {
res += ord_at(i);
}
} else {
res += ord_at(i);
}
}
return String::utf8(res.ascii());
}
String String::c_unescape() const {
String escaped=*this;
escaped=escaped.replace("\\a","\a");
escaped=escaped.replace("\\b","\b");
escaped=escaped.replace("\\f","\f");
escaped=escaped.replace("\\n","\n");
escaped=escaped.replace("\\r","\r");
escaped=escaped.replace("\\t","\t");
escaped=escaped.replace("\\v","\v");
escaped=escaped.replace("\\'","\'");
escaped=escaped.replace("\\\"","\"");
escaped=escaped.replace("\\?","\?");
escaped=escaped.replace("\\\\","\\");
return escaped;
}
String String::c_escape() const {
String escaped=*this;
escaped=escaped.replace("\\","\\\\");
escaped=escaped.replace("\a","\\a");
escaped=escaped.replace("\b","\\b");
escaped=escaped.replace("\f","\\f");
escaped=escaped.replace("\n","\\n");
escaped=escaped.replace("\r","\\r");
escaped=escaped.replace("\t","\\t");
escaped=escaped.replace("\v","\\v");
escaped=escaped.replace("\'","\\'");
escaped=escaped.replace("\?","\\?");
escaped=escaped.replace("\"","\\\"");
return escaped;
}
String String::c_escape_multiline() const {
String escaped=*this;
escaped=escaped.replace("\\","\\\\");
escaped=escaped.replace("\"","\\\"");
return escaped;
}
String String::json_escape() const {
String escaped=*this;
escaped=escaped.replace("\\","\\\\");
escaped=escaped.replace("\b","\\b");
escaped=escaped.replace("\f","\\f");
escaped=escaped.replace("\n","\\n");
escaped=escaped.replace("\r","\\r");
escaped=escaped.replace("\t","\\t");
escaped=escaped.replace("\v","\\v");
escaped=escaped.replace("\"","\\\"");
return escaped;
}
String String::xml_escape(bool p_escape_quotes) const {
String str=*this;
str=str.replace("&","&amp;");
str=str.replace("<","&lt;");
str=str.replace(">","&gt;");
if (p_escape_quotes) {
str=str.replace("'","&apos;");
str=str.replace("\"","&quot;");
}
/*
for (int i=1;i<32;i++) {
char chr[2]={i,0};
str=str.replace(chr,"&#"+String::num(i)+";");
}*/
return str;
}
static _FORCE_INLINE_ int _xml_unescape(const CharType *p_src,int p_src_len,CharType *p_dst) {
int len=0;
while(p_src_len) {
if (*p_src=='&') {
int eat=0;
if (p_src_len>=4 && p_src[1]=='#') {
CharType c=0;
for(int i=2;i<p_src_len;i++) {
eat=i+1;
CharType ct=p_src[i];
if (ct==';') {
break;
} else if (ct>='0' && ct<='9') {
ct=ct-'0';
} else if (ct>='a' && ct<='f') {
ct=(ct-'a')+10;
} else if (ct>='A' && ct<='F') {
ct=(ct-'A')+10;
} else {
continue;
}
c<<=4;
c|=ct;
}
if (p_dst)
*p_dst=c;
} else if (p_src_len>=4 && p_src[1]=='g' && p_src[2]=='t' && p_src[3]==';') {
if (p_dst)
*p_dst='>';
eat=4;
} else if (p_src_len>=4 && p_src[1]=='l' && p_src[2]=='t' && p_src[3]==';') {
if (p_dst)
*p_dst='<';
eat=4;
} else if (p_src_len>=5 && p_src[1]=='a' && p_src[2]=='m' && p_src[3]=='p' && p_src[4]==';') {
if (p_dst)
*p_dst='&';
eat=5;
} else if (p_src_len>=6 && p_src[1]=='q' && p_src[2]=='u' && p_src[3]=='o' && p_src[4]=='t' && p_src[5]==';') {
if (p_dst)
*p_dst='"';
eat=6;
} else if (p_src_len>=6 && p_src[1]=='a' && p_src[2]=='p' && p_src[3]=='o' && p_src[4]=='s' && p_src[5]==';') {
if (p_dst)
*p_dst='\'';
eat=6;
} else {
if (p_dst)
*p_dst=*p_src;
eat=1;
}
if (p_dst)
p_dst++;
len++;
p_src+=eat;
p_src_len-=eat;
} else {
if (p_dst) {
*p_dst=*p_src;
p_dst++;
}
len++;
p_src++;
p_src_len--;
}
}
return len;
}
String String::xml_unescape() const {
String str;
int l = length();
int len = _xml_unescape(c_str(),l,NULL);
if (len==0)
return String();
str.resize(len+1);
_xml_unescape(c_str(),l,&str[0]);
str[len]=0;
return str;
}
String String::pad_decimals(int p_digits) const {
String s=*this;
int c = s.find(".");
if (c==-1) {
if (p_digits<=0) {
return s;
}
s+=".";
c=s.length()-1;
} else {
if (p_digits<=0) {
return s.substr(0,c);
}
}
if (s.length()-(c+1) > p_digits) {
s=s.substr(0,c+p_digits+1);
} else {
while(s.length()-(c+1) < p_digits) {
s+="0";
}
}
return s;
}
String String::pad_zeros(int p_digits) const {
String s=*this;
int end = s.find(".");
if (end==-1) {
end=s.length();
}
if (end==0)
return s;
int begin=0;
while (begin<end && (s[begin]<'0' || s[begin]>'9')) {
begin++;
}
if (begin>=end)
return s;
while(end-begin < p_digits) {
s=s.insert(begin,"0");
end++;
}
return s;
}
bool String::is_valid_integer() const {
int len = length();
if (len==0)
return false;
int from=0;
if (len!=1 && (operator[](0)=='+' || operator[](0)=='-'))
from++;
for(int i=from;i<len;i++) {
if (operator[](i)<'0' || operator[](i)>'9')
return false; // no start with number plz
}
return true;
}
bool String::is_valid_hex_number(bool p_with_prefix) const {
int from = 0;
int len = length();
if (len!=1 && (operator[](0)=='+' || operator[](0)=='-'))
from++;
if (p_with_prefix) {
if (len < 2)
return false;
if (operator[](from) != '0' || operator[](from+1) != 'x') {
return false;
};
from += 2;
};
for (int i=from; i<len; i++) {
CharType c = operator[](i);
if ((c >= '0' && c <= '9') || (c >= 'a' && c <= 'f') || (c >= 'A' && c <= 'F'))
continue;
return false;
};
return true;
};
bool String::is_valid_float() const {
int len = length();
if (len==0)
return false;
int from=0;
if (operator[](0)=='+' || operator[](0)=='-') {
from++;
}
//this was pulled out of my ass, i wonder if it's correct...
bool exponent_found=false;
bool period_found=false;
bool sign_found=false;
bool exponent_values_found=false;
bool numbers_found=false;
for(int i=from;i<len;i++) {
if (operator[](i)>='0' && operator[](i)<='9') {
if (exponent_found)
exponent_values_found=true;
else
numbers_found=true;
} else if (numbers_found && !exponent_found && operator[](i)=='e') {
exponent_found=true;
} else if (!period_found && !exponent_found && operator[](i)=='.') {
period_found=true;
} else if ((operator[](i)=='-' || operator[](i)=='+') && exponent_found && !exponent_values_found && !sign_found) {
sign_found=true;
} else
return false; // no start with number plz
}
return numbers_found;
}
String String::path_to_file(const String& p_path) const {
String src=this->replace("\\","/").get_base_dir();
String dst=p_path.replace("\\","/").get_base_dir();
String rel = src.path_to(dst);
if (rel==dst) // failed
return p_path;
else
return rel+p_path.get_file();
}
String String::path_to(const String& p_path) const {
String src=this->replace("\\","/");
String dst=p_path.replace("\\","/");
if (!src.ends_with("/"))
src+="/";
if (!dst.ends_with("/"))
dst+="/";
String base;
if (src.begins_with("res://") && dst.begins_with("res://")) {
base="res:/";
src=src.replace("res://","/");
dst=dst.replace("res://","/");
} else if (src.begins_with("user://") && dst.begins_with("user://")) {
base="user:/";
src=src.replace("user://","/");
dst=dst.replace("user://","/");
} else if (src.begins_with("/") && dst.begins_with("/")) {
//nothing
} else {
//dos style
String src_begin=src.get_slicec('/',0);
String dst_begin=dst.get_slicec('/',0);
if (src_begin!=dst_begin)
return p_path; //impossible to do this
base=src_begin;
src=src.substr(src_begin.length(),src.length());
dst=dst.substr(dst_begin.length(),dst.length());
}
//remove leading and trailing slash and split
Vector<String> src_dirs=src.substr(1,src.length()-2).split("/");
Vector<String> dst_dirs=dst.substr(1,dst.length()-2).split("/");
//find common parent
int common_parent=0;
while(true) {
if (src_dirs.size()==common_parent)
break;
if (dst_dirs.size()==common_parent)
break;
if (src_dirs[common_parent]!=dst_dirs[common_parent])
break;
common_parent++;
}
common_parent--;
String dir;
for(int i=src_dirs.size()-1;i>common_parent;i--) {
dir+="../";
}
for(int i=common_parent+1;i<dst_dirs.size();i++) {
dir+=dst_dirs[i]+"/";
}
if (dir.length()==0)
dir="./";
return dir;
}
bool String::is_valid_html_color() const {
return Color::html_is_valid(*this);
}
bool String::is_valid_ip_address() const {
if (find(":") >= 0) {
Vector<String> ip = split(":");
for (int i=0; i<ip.size(); i++) {
String n = ip[i];
if (n.empty())
continue;
if (n.is_valid_hex_number(false)) {
int nint = n.hex_to_int(false);
if (nint < 0 || nint > 0xffff)
return false;
continue;
};
if (!n.is_valid_ip_address())
return false;
};
} else {
Vector<String> ip = split(".");
if (ip.size()!=4)
return false;
for(int i=0;i<ip.size();i++) {
String n = ip[i];
if (!n.is_valid_integer())
return false;
int val = n.to_int();
if (val<0 || val>255)
return false;
}
};
return true;
}
bool String::is_resource_file() const {
return begins_with("res://") && find("::")==-1;
}
bool String::is_rel_path() const {
return !is_abs_path();
}
String String::get_base_dir() const {
int basepos = find("://");
String rs;
String base;
if (basepos!=-1) {
int end = basepos+3;
rs = substr(end,length());
base = substr(0,end);
} else {
if (begins_with("/")) {
rs=substr(1,length());
base="/";
} else {
rs=*this;
}
}
int sep = MAX( rs.find_last("/"), rs.find_last("\\") );
if (sep==-1)
return base;
return base+rs.substr(0,sep);
}
String String::get_file() const {
int sep = MAX( find_last("/"), find_last("\\") );
if (sep==-1)
return *this;
return substr(sep+1,length());
}
String String::extension() const {
int pos = find_last(".");
if (pos<0)
return *this;
return substr( pos+1, length() );
}
String String::plus_file(const String& p_file) const {
if (length()>0 && operator [](length()-1)=='/')
return *this+p_file;
else
return *this+"/"+p_file;
}
String String::percent_encode() const {
CharString cs = utf8();
String encoded;
for(int i=0;i<cs.length();i++) {
uint8_t c = cs[i];
if ( (c>='A' && c<='Z') || (c>='a' && c<='z') || (c>='0' && c<='9') || c=='-' || c=='_' || c=='~' || c=='.') {
char p[2]={(char)c,0};
encoded+=p;
} else {
char p[4]={'%',0,0,0};
static const char hex[16]={'0','1','2','3','4','5','6','7','8','9','a','b','c','d','e','f'};
p[1]=hex[c>>4];
p[2]=hex[c&0xF];
encoded+=p;
}
}
return encoded;
}
String String::percent_decode() const {
CharString pe;
CharString cs = utf8();
for(int i=0;i<cs.length();i++) {
uint8_t c = cs[i];
if (c=='%' && i<length()-2) {
uint8_t a = LOWERCASE(cs[i+1]);
uint8_t b = LOWERCASE(cs[i+2]);
c=0;
if (a>='0' && a<='9')
c=(a-'0')<<4;
else if (a>='a' && a<='f')
c=(a-'a'+10)<<4;
else
continue;
uint8_t d=0;
if (b>='0' && b<='9')
d=(b-'0');
else if (b>='a' && b<='f')
d=(b-'a'+10);
else
continue;
c+=d;
i+=2;
}
pe.push_back(c);
}
pe.push_back(0);
return String::utf8(pe.ptr());
}
String String::basename() const {
int pos = find_last(".");
if (pos<0)
return *this;
return substr( 0, pos );
}
String itos(int64_t p_val) {
return String::num_int64(p_val);
}
String rtos(double p_val) {
return String::num(p_val);
}
String rtoss(double p_val) {
return String::num_scientific(p_val);
}
// Right-pad with a character.
String String::rpad(int min_length, const String& character) const {
String s = *this;
int padding = min_length - s.length();
if (padding > 0) {
for (int i = 0; i < padding; i++) s = s + character;
}
return s;
}
// Left-pad with a character.
String String::lpad(int min_length, const String& character) const {
String s = *this;
int padding = min_length - s.length();
if (padding > 0) {
for (int i = 0; i < padding; i++) s = character + s;
}
return s;
}
// sprintf is implemented in GDScript via:
// "fish %s pie" % "frog"
// "fish %s %d pie" % ["frog", 12]
// In case of an error, the string returned is the error description and "error" is true.
String String::sprintf(const Array& values, bool* error) const {
String formatted;
CharType* self = (CharType*)c_str();
bool in_format = false;
int value_index = 0;
int min_chars;
int min_decimals;
bool in_decimals;
bool pad_with_zeroes;
bool left_justified;
bool show_sign;
*error = true;
for (; *self; self++) {
const CharType c = *self;
if (in_format) { // We have % - lets see what else we get.
switch (c) {
case '%': { // Replace %% with %
formatted += chr(c);
in_format = false;
break;
}
case 'd': // Integer (signed)
case 'o': // Octal
case 'x': // Hexadecimal (lowercase)
case 'X': { // Hexadecimal (uppercase)
if (value_index >= values.size()) {
return "not enough arguments for format string";
}
if (!values[value_index].is_num()) {
return "a number is required";
}
int64_t value = values[value_index];
int base;
bool capitalize = false;
switch (c) {
case 'd': base = 10; break;
case 'o': base = 8; break;
case 'x': base = 16; break;
case 'X': base = 16; capitalize = true; break;
}
// Get basic number.
String str = String::num_int64(ABS(value), base, capitalize);
int number_len = str.length();
// Padding.
String pad_char = pad_with_zeroes ? String("0") : String(" ");
if (left_justified) {
str = str.rpad(min_chars, pad_char);
} else {
str = str.lpad(min_chars, pad_char);
}
// Sign.
if (show_sign && value >= 0) {
str = str.insert(pad_with_zeroes?0:str.length()-number_len, "+");
} else if (value < 0) {
str = str.insert(pad_with_zeroes?0:str.length()-number_len, "-");
}
formatted += str;
++value_index;
in_format = false;
break;
}
case 'f': { // Float
if (value_index >= values.size()) {
return "not enough arguments for format string";
}
if (!values[value_index].is_num()) {
return "a number is required";
}
double value = values[value_index];
String str = String::num(value, min_decimals);
// Pad decimals out.
str = str.pad_decimals(min_decimals);
// Show sign
if (show_sign && value >= 0) {
str = str.insert(0, "+");
}
// Padding
if (left_justified) {
str = str.rpad(min_chars);
} else {
str = str.lpad(min_chars);
}
formatted += str;
++value_index;
in_format = false;
break;
}
case 's': { // String
if (value_index >= values.size()) {
return "not enough arguments for format string";
}
String str = values[value_index];
// Padding.
if (left_justified) {
str = str.rpad(min_chars);
} else {
str = str.lpad(min_chars);
}
formatted += str;
++value_index;
in_format = false;
break;
}
case 'c': {
if (value_index >= values.size()) {
return "not enough arguments for format string";
}
// Convert to character.
String str;
if (values[value_index].is_num()) {
int value = values[value_index];
if (value < 0) {
return "unsigned byte integer is lower than maximum";
} else if (value > 255) {
return "unsigned byte integer is greater than maximum";
}
str = chr(values[value_index]);
} else if (values[value_index].get_type() == Variant::STRING) {
str = values[value_index];
if (str.length() != 1) {
return "%c requires number or single-character string";
}
} else {
return "%c requires number or single-character string";
}
// Padding.
if (left_justified) {
str = str.rpad(min_chars);
} else {
str = str.lpad(min_chars);
}
formatted += str;
++value_index;
in_format = false;
break;
}
case '-': { // Left justify
left_justified = true;
break;
}
case '+': { // Show + if positive.
show_sign = true;
break;
}
case '0': case '1': case '2': case '3': case '4':
case '5': case '6': case '7': case '8': case '9': {
int n = c - '0';
if (in_decimals) {
min_decimals *= 10;
min_decimals += n;
} else {
if (c == '0' && min_chars == 0) {
pad_with_zeroes = true;
} else {
min_chars *= 10;
min_chars += n;
}
}
break;
}
case '.': { // Float separtor.
if (in_decimals) {
return "too many decimal points in format";
}
in_decimals = true;
min_decimals = 0; // We want to add the value manually.
break;
}
case '*': { // Dyanmic width, based on value.
if (value_index >= values.size()) {
return "not enough arguments for format string";
}
if (!values[value_index].is_num()) {
return "* wants number";
}
int size = values[value_index];
if (in_decimals) {
min_decimals = size;
} else {
min_chars = size;
}
++value_index;
break;
}
default: {
return "unsupported format character";
}
}
} else { // Not in format string.
switch (c) {
case '%':
in_format = true;
// Back to defaults:
min_chars = 0;
min_decimals = 6;
pad_with_zeroes = false;
left_justified = false;
show_sign = false;
in_decimals = false;
break;
default:
formatted += chr(c);
}
}
}
if (in_format) {
return "incomplete format";
}
if (value_index != values.size()) {
return "not all arguments converted during string formatting";
}
*error = false;
return formatted;
}
#include "translation.h"
#ifdef TOOLS_ENABLED
String TTR(const String& p_text) {
if (TranslationServer::get_singleton()) {
return TranslationServer::get_singleton()->tool_translate(p_text);
}
return p_text;
}
#endif
String RTR(const String& p_text) {
if (TranslationServer::get_singleton()) {
String rtr = TranslationServer::get_singleton()->tool_translate(p_text);
if (rtr==String() || rtr==p_text) {
return TranslationServer::get_singleton()->translate(p_text);
} else {
return rtr;
}
}
return p_text;
}