]> git.tdb.fi Git - libs/core.git/blobdiff - source/lexicalcast.cpp
Rewrite lexical_cast to use internal conversion routines
[libs/core.git] / source / lexicalcast.cpp
diff --git a/source/lexicalcast.cpp b/source/lexicalcast.cpp
new file mode 100644 (file)
index 0000000..687a37b
--- /dev/null
@@ -0,0 +1,619 @@
+/* $Id$
+
+This file is part of libmspstrings
+Copyright © 2006-2008 Mikko Rasa
+Distributed under the LGPL
+*/
+
+#include <cmath>
+#include <limits>
+#include "lexicalcast.h"
+
+using namespace std;
+
+namespace {
+
+using namespace Msp;
+
+template<typename T>
+struct IsSigned
+{ enum { result=!(static_cast<T>(-1)>0) }; };
+
+template<typename T, bool f=(sizeof(T)>sizeof(unsigned long))>
+struct Temporary
+{ typedef unsigned long Type; };
+
+template<typename T>
+struct Temporary<T, true>
+{
+#ifdef WIN32
+       typedef __int64 Type;
+#else
+       typedef unsigned long long Type;
+#endif
+};
+
+/* Helper to avoid warnings about an unsigned type never being < 0 */
+template<typename T, bool f=IsSigned<T>::result>
+struct IsNegative
+{ static bool eval(T v) { return v<0; } };
+
+template<typename T>
+struct IsNegative<T, false>
+{ static bool eval(T) { return false; } };
+
+/* Helper to avoid errors about ambiguous function calls since there are no
+overloads of abs for unsigned types */
+template<typename T, bool f=IsSigned<T>::result>
+struct Absolute
+{ static T eval(T v) { return v<0 ? -v : v; } };
+
+template<typename T>
+struct Absolute<T, false>
+{ static T eval(T v) { return v; } };
+
+
+/*** Integer conversions ***/
+
+const char udigits[]="0123456789ABCDEF";
+const char ldigits[]="0123456789abcdef";
+
+template<typename T>
+char *int_to_str(T v, const Fmt &f, char *end)
+{
+       if(f.get_type()==Fmt::CHAR)
+       {
+               *--end=v;
+               return end;
+       }
+
+       char *ptr=end;
+
+       // Find out the base to use
+       unsigned base=f.get_base();
+       if(!base)
+               base=10;
+
+       // Format the number, starting from the least significant digit
+       const char *digits=(f.get_uppercase() ? udigits : ldigits);
+       if(v)
+       {
+               typename Temporary<T>::Type w=Absolute<T>::eval(v);
+               while(w)
+               {
+                       *--ptr=digits[w%base];
+                       w/=base;
+               }
+       }
+       else
+               *--ptr=digits[0];
+
+       char sign=(IsNegative<T>::eval(v) ? '-' : f.get_showpos() ? '+' : 0);
+       if(f.get_fill()=='0')
+       {
+               /* Zero-fill, taking base/sign size into account.  The expression is a
+               bit ugly, but saves having to write code for creating the prefix both
+               ways. */
+               unsigned pfxsize=((f.get_showbase() && base!=10) ? base==8 ? 1 : 2 : 0) + (sign!=0);
+               for(unsigned i=(end-ptr)+pfxsize; i<f.get_width(); ++i)
+                       *--ptr='0';
+       }
+
+       if(f.get_showbase() && v!=0)
+       {
+               // Add base indicator
+               if(base==2)
+                       *--ptr=(f.get_uppercase() ? 'B' : 'b');
+               else if(base==16)
+                       *--ptr=(f.get_uppercase() ? 'X' : 'x');
+               if(base!=10)
+                       *--ptr='0';
+       }
+
+       if(sign)
+               *--ptr=sign;
+
+       return ptr;
+}
+
+template<typename T>
+string int_to_str(T v, const Fmt &f)
+{
+       unsigned size=max(f.get_width(), max(f.get_precision(), sizeof(T)*8+3));
+       char *buf=new char[size];
+       string result(int_to_str(v, f, buf+size), buf+size);
+       delete[] buf;
+       return result;
+}
+
+template<typename T>
+T str_to_int(const std::string &s, const Fmt &f)
+{
+       if(s.empty())
+               throw LexicalError("Empty input in integer conversion");
+
+       std::string::const_iterator i=s.begin();
+
+       // See if the input starts with a sign
+       bool neg=false;
+       if(*i=='-')
+       {
+               if(!IsSigned<T>::result)
+                       throw LexicalError("Negative sign in unsigned integer conversion");
+               neg=true;
+               ++i;
+       }
+       else if(*i=='+')
+               ++i;
+
+       // Must have some digits to convert
+       if(i==s.end())
+               throw LexicalError("Missing digits in integer conversion");
+
+       T base=f.get_base();
+       if(!base && i!=s.end())
+       {
+               // Automatic base detection requested, figure it out
+               if(*i=='0' && ++i!=s.end())
+               {
+                       if(*i=='x' || *i=='X')
+                       {
+                               base=16;
+                               ++i;
+                       }
+                       else if(*i=='b' || *i=='B')
+                       {
+                               base=2;
+                               ++i;
+                       }
+                       else
+                               base=8;
+               }
+               else
+                       base=10;
+       }
+
+       // Parse the digits
+       T result=0;
+       for(; i!=s.end(); ++i)
+       {
+               T digit=base;
+               if(*i>='0' && *i<='9')
+                       digit=*i-'0';
+               else if(*i>='A' && *i<='F')
+                       digit=*i-'A'+10;
+               else if(*i>='a' && *i<='f')
+                       digit=*i-'a'+10;
+               if(digit>=base)
+                       throw LexicalError("Invalid digit in integer conversion");
+               T next=result*base+digit;
+               if(next/base!=result)
+                       throw LexicalError("Overflow in integer conversion");
+               result=next;
+       }
+
+       if(neg)
+               result=-result;
+
+       return result;
+}
+
+
+/*** Boolean conversions ***/
+
+string bool_to_str(bool b, const Fmt &f)
+{
+       if(f.get_type()==Fmt::STR)
+               return b ? "true" : "false";
+       else
+               return b ? "1" : "0";
+}
+
+bool str_to_bool(const string &s)
+{
+       if(s=="1" || s=="true" || s=="yes" || s=="on")
+               return true;
+       else if(s=="0" || s=="false" || s=="no" || s=="off")
+               return true;
+       throw LexicalError("Invalid input in boolean conversion");
+}
+
+
+/*** Floating-point conversions ***/
+
+template<typename T>
+string flt_to_str(T v, const Fmt &f)
+{
+       if(f.get_type()==Fmt::CHAR)
+               throw LexicalError("Character format in floating-point conversion");
+
+       Fmt::FloatMode mode=f.get_floatmode();
+       long double w=abs(v);
+       char sign=(v<0 ? '-' : f.get_showpos() ? '+' : 0);
+
+       // Handle infinity and not-a-number as special cases
+       if(!(w+w>w) && w!=0)
+       {
+               string result;
+               if(sign)
+                       result+=sign;
+               if(!(w>=0))
+                       result+=(f.get_uppercase() ? "NAN" : "nan");
+               else
+                       result+=(f.get_uppercase() ? "INF" : "inf");
+               if(result.size()<f.get_width())
+                       result=string(f.get_width()-result.size(), ' ')+result;
+               return result;
+       }
+
+       /* Find out the base-10 exponent.  Building up the multiplier / divisor
+       first helps with accuracy in some cases. */
+       int exp=0;
+       if(w>=10)
+       {
+               long double div=1;
+               while(div*10<w)
+               {
+                       ++exp;
+                       div*=10;
+               }
+               w/=div;
+       }
+       else if(mode!=Fmt::FIXED && w<1 && w!=0)
+       {
+               long double mul=1;
+               while(w*mul<1)
+               {
+                       --exp;
+                       mul*=10;
+               }
+               w*=mul;
+       }
+
+       // Decide how to format the number
+       unsigned digits;
+       unsigned point=1;
+       bool showexp=false;
+       if(mode==Fmt::FIXED)
+       {
+               point=exp+1;
+               digits=point+f.get_precision();
+       }
+       else if(mode==Fmt::SCI)
+       {
+               digits=f.get_precision()+1;
+               showexp=true;
+       }
+       else
+       {
+               digits=max(f.get_precision(), 1U);
+               if(exp<-4 || exp>=static_cast<int>(digits))
+               {
+                       point=1;
+                       showexp=true;
+               }
+               else
+               {
+                       point=max(exp, 0)+1;
+                       if(exp<0)
+                               digits+=-exp;
+               }
+       }
+
+       // Apply rounding
+       w+=5.0l/pow(10.0l, digits);
+       if(w>10)
+       {
+               // Rounding bumped us to the next exponent, deal with it
+               w/=10;
+               if(mode==Fmt::AUTOFLT && exp+1==static_cast<int>(digits))
+               {
+                       point=1;
+                       showexp=true;
+               }
+               if(!showexp)
+               {
+                       ++digits;
+                       ++point;
+               }
+               else
+                       ++exp;
+       }
+
+       // Create a buffer and start from the end
+       unsigned size=max(f.get_width(), digits+8);
+       char *buf=new char[size];
+       char *end=buf+size;
+       char *ptr=end;
+
+       // Format exponent
+       if(showexp)
+       {
+               ptr=int_to_str(exp, Fmt().showpos().fill('0').width(3), ptr);
+               *--ptr=(f.get_uppercase() ? 'E' : 'e');
+       }
+
+       // Format mantissa left-to-right
+       char *eptr=ptr;
+       ptr-=digits+(point<digits || f.get_showpoint());
+       char *mptr=ptr;
+       for(unsigned i=0; i<digits; ++i)
+       {
+               if(i==point)
+                       *mptr++='.';
+               if(showexp || static_cast<int>(i)>=-exp)
+               {
+                       int digit=static_cast<int>(w);
+                       *mptr++='0'+digit;
+                       w=(w-digit)*10;
+               }
+               else
+                       *mptr++='0';
+       }
+
+       if(f.get_showpoint())
+       {
+               // Radix point requested but not displayed yet, add it
+               if(digits<=point)
+                       *mptr++='.';
+       }
+       else if(mode==Fmt::AUTOFLT && digits>point)
+       {
+               // Remove trailing zeroes from fraction and a lone radix point
+               while(mptr[-1]=='0')
+                       --mptr;
+               if(mptr[-1]=='.')
+                       --mptr;
+               if(mptr!=eptr)
+               {
+                       while(mptr!=ptr)
+                               *--eptr=*--mptr;
+                       ptr=eptr;
+               }
+       }
+
+       // Add filling and sign
+       if(f.get_fill()=='0')
+       {
+               unsigned pfxlen=(sign!=0);
+               while(end-ptr+pfxlen<f.get_width())
+                       *--ptr='0';
+       }
+       if(sign)
+               *--ptr=sign;
+
+       string result(ptr, end);
+       delete[] buf;
+       return result;
+}
+
+template<typename T>
+T str_to_flt(const string &s, const Fmt &)
+{
+       if(s.empty())
+               throw LexicalError("Empty input in floating-point conversion");
+
+       std::string::const_iterator i=s.begin();
+
+       // See if the input starts with a sign
+       bool neg=false;
+       if(*i=='-')
+       {
+               neg=true;
+               ++i;
+       }
+       else if(*i=='+')
+               ++i;
+
+       // Must have some digits to convert
+       if(i==s.end())
+               throw LexicalError("Missing digits in floating-point conversion");
+
+       long double v=0;
+       int exp=0;
+
+       // Parse mantissa
+       bool point_seen=false;
+       for(; i!=s.end(); ++i)
+       {
+               if(*i=='.')
+               {
+                       if(point_seen)
+                               throw LexicalError("Extra point in floating-point conversion");
+                       point_seen=true;
+               }
+               else if(*i>='0' && *i<='9')
+               {
+                       v=v*10+(*i-'0');
+                       if(point_seen)
+                               --exp;
+               }
+               else if(*i=='e' || *i=='E')
+               {
+                       // We have an exponent
+                       ++i;
+
+                       exp+=str_to_int<int>(string(i, s.end()), Fmt());
+                       // str_to_int has eaten the rest of the input or thrown
+                       break;
+               }
+               else
+                       throw LexicalError("Invalid digit in floating-point conversion");
+       }
+
+       // Scale and negate the result as needed
+       while(exp>0)
+       {
+               v*=10;
+               --exp;
+       }
+       while(exp<0)
+       {
+               v/=10;
+               ++exp;
+       }
+
+       if(neg)
+               v=-v;
+
+       return v;
+}
+
+
+/*** String conversions ***/
+
+string str_to_str(const string &s, const Fmt &f)
+{
+       if(f.get_type()==Fmt::NUM)
+               throw LexicalError("Numeric format in string conversion");
+       return s;
+}
+
+}
+
+namespace Msp {
+
+void LexicalConverter::result(const string &s)
+{
+       if(s.size()<fmt.get_width())
+       {
+               if(fmt.get_align()==Fmt::RIGHT)
+                       buf=string(fmt.get_width()-s.size(), fmt.get_fill())+s;
+               else
+                       buf=s+string(fmt.get_width()-s.size(), fmt.get_fill());
+       }
+       else
+               buf=s;
+}
+
+
+/*** operator<< ***/
+
+void operator<<(LexicalConverter &c, char v)
+{
+       Fmt::Type type=c.get_fmt().get_type();
+       if(type==Fmt::NUM)
+               c.result(int_to_str(v, c.get_fmt()));
+       else
+               c.result(string(1, v));
+}
+
+void operator<<(LexicalConverter &c, signed char v)
+{ c.result(int_to_str(v, c.get_fmt())); }
+
+void operator<<(LexicalConverter &c, short v)
+{ c.result(int_to_str(v, c.get_fmt())); }
+
+void operator<<(LexicalConverter &c, int v)
+{ c.result(int_to_str(v, c.get_fmt())); }
+
+void operator<<(LexicalConverter &c, long v)
+{ c.result(int_to_str(v, c.get_fmt())); }
+
+void operator<<(LexicalConverter &c, unsigned char v)
+{ c.result(int_to_str(v, c.get_fmt())); }
+
+void operator<<(LexicalConverter &c, unsigned short v)
+{ c.result(int_to_str(v, c.get_fmt())); }
+
+void operator<<(LexicalConverter &c, unsigned v)
+{ c.result(int_to_str(v, c.get_fmt())); }
+
+void operator<<(LexicalConverter &c, unsigned long v)
+{ c.result(int_to_str(v, c.get_fmt())); }
+
+#ifdef __GNUC__
+void operator<<(LexicalConverter &c, long long v)
+{ c.result(int_to_str(v, c.get_fmt())); }
+
+void operator<<(LexicalConverter &c, unsigned long long v)
+{ c.result(int_to_str(v, c.get_fmt())); }
+#endif
+
+void operator<<(LexicalConverter &c, bool v)
+{ c.result(bool_to_str(v, c.get_fmt())); }
+
+void operator<<(LexicalConverter &c, float v)
+{ c.result(flt_to_str(v, c.get_fmt())); }
+
+void operator<<(LexicalConverter &c, double v)
+{ c.result(flt_to_str(v, c.get_fmt())); }
+
+void operator<<(LexicalConverter &c, long double v)
+{ c.result(flt_to_str(v, c.get_fmt())); }
+
+void operator<<(LexicalConverter &c, const string &s)
+{ c.result(str_to_str(s, c.get_fmt())); }
+
+void operator<<(LexicalConverter &c, const char *s)
+{ c.result(str_to_str(s, c.get_fmt())); }
+
+void operator<<(LexicalConverter &c, const void *p)
+{ c.result(int_to_str(reinterpret_cast<unsigned long>(p), c.get_fmt())); }
+
+
+/*** operator>> ***/
+
+void operator>>(const LexicalConverter &c, char &v)
+{
+       if(c.get_fmt().get_type()==Fmt::NUM)
+               v=str_to_int<char>(c.get(), c.get_fmt());
+       else
+       {
+               const std::string &s=c.get();
+               if(s.empty())
+                       throw LexicalError("Empty input in character conversion");
+               if(s.size()>1)
+                       throw LexicalError("Extra input in character conversion");
+               v=s[0];
+       }
+}
+
+void operator>>(const LexicalConverter &c, signed char &v)
+{ v=str_to_int<signed char>(c.get(), c.get_fmt()); }
+
+void operator>>(const LexicalConverter &c, short &v)
+{ v=str_to_int<short>(c.get(), c.get_fmt()); }
+
+void operator>>(const LexicalConverter &c, int &v)
+{ v=str_to_int<int>(c.get(), c.get_fmt()); }
+
+void operator>>(const LexicalConverter &c, long &v)
+{ v=str_to_int<long>(c.get(), c.get_fmt()); }
+
+void operator>>(const LexicalConverter &c, unsigned char &v)
+{ v=str_to_int<unsigned char>(c.get(), c.get_fmt()); }
+
+void operator>>(const LexicalConverter &c, unsigned short &v)
+{ v=str_to_int<unsigned short>(c.get(), c.get_fmt()); }
+
+void operator>>(const LexicalConverter &c, unsigned int &v)
+{ v=str_to_int<unsigned int>(c.get(), c.get_fmt()); }
+
+void operator>>(const LexicalConverter &c, unsigned long &v)
+{ v=str_to_int<unsigned long>(c.get(), c.get_fmt()); }
+
+#ifdef __GNUC__
+void operator>>(const LexicalConverter &c, long long &v)
+{ v=str_to_int<long long>(c.get(), c.get_fmt()); }
+
+void operator>>(const LexicalConverter &c, unsigned long long &v)
+{ v=str_to_int<unsigned long long>(c.get(), c.get_fmt()); }
+#endif
+
+void operator>>(const LexicalConverter &c, bool &v)
+{ v=str_to_bool(c.get()); }
+
+void operator>>(const LexicalConverter &c, float &v)
+{ v=str_to_flt<float>(c.get(), c.get_fmt()); }
+
+void operator>>(const LexicalConverter &c, double &v)
+{ v=str_to_flt<double>(c.get(), c.get_fmt()); }
+
+void operator>>(const LexicalConverter &c, long double &v)
+{ v=str_to_flt<long double>(c.get(), c.get_fmt()); }
+
+void operator>>(const LexicalConverter &c, string &s)
+{ s=str_to_str(c.get(), c.get_fmt()); }
+
+} // namespace Msp