leavelet/fmt
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Implement formatting of integers.

Browse Source
This commit is contained in:
Victor Zverovich
2012-12-10 11:08:16 -08:00
parent c5dce1caf9
commit b98b2e9842
3 changed files with 310 additions and 185 deletions
Download Patch File Download Diff File Expand all files Collapse all files
format.cc
+198 -127
View File
@@ -5,6 +5,8 @@
#include "format.h"
#include <stdint.h>
#include <cassert>
#include <climits>
#include <cstring>
@@ -14,16 +16,16 @@ using std::size_t;
namespace {
// Flags.
enum { PLUS_FLAG = 1, ZERO_FLAG = 2, HEX_PREFIX_FLAG = 4 };
// Throws Exception(message) if format contains '}', otherwise throws
// FormatError reporting unmatched '{'. The idea is that unmatched '{'
// should override other errors.
template <typename Exception>
void Throw(const char *s, const char *message) {
void ReportError(const char *s, const std::string &message) {
while (*s && *s != '}')
++s;
if (!*s)
throw fmt::FormatError("unmatched '{' in format");
throw Exception(message);
throw fmt::FormatError(*s ? message : std::string("unmatched '{' in format"));
}
// Parses an unsigned integer advancing s to the end of the parsed input.
@@ -34,29 +36,168 @@ unsigned ParseUInt(const char *&s) {
do {
unsigned new_value = value * 10 + (*s++ - '0');
if (new_value < value) // Check if value wrapped around.
Throw<fmt::FormatError>(s, "number is too big in format");
ReportError(s, "number is too big in format");
value = new_value;
} while ('0' <= *s && *s <= '9');
return value;
}
// Maps an integer type T to its unsigned counterpart.
template <typename T>
struct GetUnsigned;
template <>
struct GetUnsigned<int> {
typedef unsigned Type;
};
template <>
struct GetUnsigned<unsigned> {
typedef unsigned Type;
};
template <>
struct GetUnsigned<long> {
typedef unsigned long Type;
};
template <>
struct GetUnsigned<unsigned long> {
typedef unsigned long Type;
};
template <typename T>
struct IsLongDouble { enum {VALUE = 0}; };
template <>
struct IsLongDouble<long double> { enum {VALUE = 1}; };
}
template <typename T>
void fmt::Formatter::FormatBuiltinArg(
const char *format, const T &arg, int width, int precision) {
void fmt::Formatter::FormatInt(T value, unsigned flags, int width, char type) {
int size = 0;
char sign = 0;
typedef typename GetUnsigned<T>::Type UnsignedType;
UnsignedType abs_value = value;
if (value < 0) {
sign = '-';
++size;
abs_value = -value;
} else if ((flags & PLUS_FLAG) != 0) {
sign = '+';
++size;
}
char fill = (flags & ZERO_FLAG) != 0 ? '0' : ' ';
size_t start = buffer_.size();
char *p = 0;
switch (type) {
case 0: case 'd': {
UnsignedType n = abs_value;
do {
++size;
} while ((n /= 10) != 0);
width = std::max(width, size);
buffer_.resize(buffer_.size() + width, fill);
p = &buffer_.back();
n = abs_value;
do {
*p-- = '0' + (n % 10);
} while ((n /= 10) != 0);
break;
}
case 'x': case 'X': {
UnsignedType n = abs_value;
bool print_prefix = (flags & HEX_PREFIX_FLAG) != 0;
if (print_prefix) size += 2;
do {
++size;
} while ((n >>= 4) != 0);
width = std::max(width, size);
buffer_.resize(buffer_.size() + width, fill);
p = &buffer_.back();
n = abs_value;
const char *digits = type == 'x' ? "0123456789abcdef" : "0123456789ABCDEF";
do {
*p-- = digits[n & 0xf];
} while ((n >>= 4) != 0);
if (print_prefix) {
*p-- = type;
*p-- = '0';
}
break;
}
case 'o': {
UnsignedType n = abs_value;
do {
++size;
} while ((n >>= 3) != 0);
width = std::max(width, size);
buffer_.resize(buffer_.size() + width, fill);
p = &buffer_.back();
n = abs_value;
do {
*p-- = '0' + (n & 7);
} while ((n >>= 3) != 0);
break;
}
default:
throw FormatError(
str(fmt::Format("unknown format code '{0}' for integer") << type));
}
if (sign) {
if ((flags & ZERO_FLAG) != 0)
buffer_[start] = sign;
else
*p = sign;
}
}
template <typename T>
void fmt::Formatter::FormatDouble(
T value, unsigned flags, int width, int precision, char type) {
// Check type.
switch (type) {
case 'e': case 'E': case 'f': case 'F': case 'g': case 'G':
break;
default:
// TODO: error
break;
}
// Build format string.
enum { MAX_FORMAT_SIZE = 9}; // longest format: %+0*.*Lg
char format[MAX_FORMAT_SIZE];
char *format_ptr = format;
*format_ptr++ = '%';
if ((flags & PLUS_FLAG) != 0)
*format_ptr++ = '+';
if ((flags & ZERO_FLAG) != 0)
*format_ptr++ = '0';
if (width > 0)
*format_ptr++ = '*';
if (precision >= 0) {
*format_ptr++ = '.';
*format_ptr++ = '*';
}
if (IsLongDouble<T>::VALUE)
*format_ptr++ = 'L';
*format_ptr++ = type ? type : 'g';
*format_ptr = '\0';
// Format using snprintf.
size_t offset = buffer_.size();
buffer_.resize(buffer_.capacity());
for (;;) {
size_t size = buffer_.size() - offset;
int n = 0;
if (width < 0) {
if (width <= 0) {
n = precision < 0 ?
snprintf(&buffer_[offset], size, format, arg) :
snprintf(&buffer_[offset], size, format, precision, arg);
snprintf(&buffer_[offset], size, format, value) :
snprintf(&buffer_[offset], size, format, precision, value);
} else {
n = precision < 0 ?
snprintf(&buffer_[offset], size, format, width, arg) :
snprintf(&buffer_[offset], size, format, width, precision, arg);
snprintf(&buffer_[offset], size, format, width, value) :
snprintf(&buffer_[offset], size, format, width, precision, value);
}
if (n >= 0 && offset + n < buffer_.size()) {
buffer_.resize(offset + n);
@@ -77,92 +218,64 @@ void fmt::Formatter::Format() {
// Parse argument index.
if (*s < '0' || *s > '9')
Throw<fmt::FormatError>(s, "missing argument index in format string");
ReportError(s, "missing argument index in format string");
unsigned arg_index = ParseUInt(s);
if (arg_index >= args_.size())
Throw<std::out_of_range>(s, "argument index is out of range in format");
ReportError(s, "argument index is out of range in format");
Arg &arg = args_[arg_index];
enum { MAX_FORMAT_SIZE = 10}; // longest format: %+0-*.*ld
char arg_format[MAX_FORMAT_SIZE];
char *arg_format_ptr = arg_format;
*arg_format_ptr++ = '%';
int width = -1;
unsigned flags = 0;
int width = 0;
int precision = -1;
char type = 0;
bool is_floating_point = false;
if (*s == ':') {
++s;
if (*s == '+') {
if (arg.type > LAST_NUMERIC_TYPE) {
Throw<FormatError>(s,
"format specifier '+' requires numeric argument");
}
++s;
if (arg.type > LAST_NUMERIC_TYPE)
ReportError(s, "format specifier '+' requires numeric argument");
if (arg.type == UINT || arg.type == ULONG) {
Throw<FormatError>(s,
ReportError(s,
"format specifier '+' requires signed argument");
}
*arg_format_ptr++ = *s++;
flags |= PLUS_FLAG;
}
if (*s == '0') {
if (arg.type > LAST_NUMERIC_TYPE) {
Throw<FormatError>(s,
"format specifier '0' requires numeric argument");
}
*arg_format_ptr++ = *s++;
++s;
if (arg.type > LAST_NUMERIC_TYPE)
ReportError(s, "format specifier '0' requires numeric argument");
flags |= ZERO_FLAG;
}
// Parse width.
if ('0' <= *s && *s <= '9') {
if (arg.type > LAST_NUMERIC_TYPE && arg.type != POINTER)
*arg_format_ptr++ = '-';
*arg_format_ptr++ = '*';
unsigned value = ParseUInt(s);
if (value > INT_MAX)
Throw<FormatError>(s, "number is too big in format");
ReportError(s, "number is too big in format");
width = value;
}
// Parse precision.
if (*s == '.') {
*arg_format_ptr++ = '.';
*arg_format_ptr++ = '*';
++s;
precision = 0;
if ('0' <= *s && *s <= '9') {
unsigned value = ParseUInt(s);
if (value > INT_MAX)
Throw<FormatError>(s, "number is too big in format");
ReportError(s, "number is too big in format");
precision = value;
} else {
Throw<FormatError>(s, "missing precision in format");
ReportError(s, "missing precision in format");
}
if (arg.type > LAST_NUMERIC_TYPE ||
(*s == '}' && arg.type != DOUBLE && arg.type != LONG_DOUBLE)) {
Throw<FormatError>(s,
"precision specifier requires floating-point type");
if (arg.type != DOUBLE && arg.type != LONG_DOUBLE) {
ReportError(s,
"precision specifier requires floating-point argument");
}
}
// Parse type.
if (*s != '}' && *s) {
if (*s != '}' && *s)
type = *s++;
if (arg.type <= LAST_NUMERIC_TYPE) {
switch (type) {
case 'd': case 'o': case 'x': case 'X':
// TODO: check that argument is integer
break;
case 'e': case 'E': case 'f': case 'F': case 'g': case 'G':
is_floating_point = true;
break;
default:
// TODO: error
break;
}
}
// TODO: check non-numeric types (string, character)
}
}
if (*s++ != '}')
@@ -172,87 +285,45 @@ void fmt::Formatter::Format() {
// Format argument.
switch (arg.type) {
case INT:
*arg_format_ptr++ = type ? type : 'd';
*arg_format_ptr = '\0';
if (is_floating_point)
FormatBuiltinArg<double>(arg_format, arg.int_value, width, precision);
else
FormatBuiltinArg(arg_format, arg.int_value, width, precision);
FormatInt(arg.int_value, flags, width, type);
break;
case UINT:
*arg_format_ptr++ = type ? type : 'u';
*arg_format_ptr = '\0';
if (is_floating_point)
FormatBuiltinArg<double>(arg_format, arg.uint_value, width, precision);
else
FormatBuiltinArg(arg_format, arg.uint_value, width, precision);
FormatInt(arg.uint_value, flags, width, type);
break;
case LONG:
*arg_format_ptr++ = 'l';
*arg_format_ptr++ = type ? type : 'd';
*arg_format_ptr = '\0';
if (is_floating_point)
FormatBuiltinArg<double>(arg_format, arg.long_value, width, precision);
else
FormatBuiltinArg(arg_format, arg.long_value, width, precision);
FormatInt(arg.long_value, flags, width, type);
break;
case ULONG:
*arg_format_ptr++ = 'l';
*arg_format_ptr++ = type ? type : 'u';
*arg_format_ptr = '\0';
if (is_floating_point)
FormatBuiltinArg<double>(arg_format, arg.ulong_value, width, precision);
else
FormatBuiltinArg(arg_format, arg.ulong_value, width, precision);
FormatInt(arg.ulong_value, flags, width, type);
break;
case DOUBLE:
*arg_format_ptr++ = type ? type : 'g';
*arg_format_ptr = '\0';
FormatBuiltinArg(arg_format, arg.double_value, width, precision);
FormatDouble(arg.double_value, flags, width, precision, type);
break;
case LONG_DOUBLE:
*arg_format_ptr++ = 'L';
*arg_format_ptr++ = type ? type : 'g';
*arg_format_ptr = '\0';
FormatBuiltinArg(arg_format, arg.long_double_value, width, precision);
break;
case POINTER:
// TODO: don't allow any type specifiers other than 'p'
*arg_format_ptr++ = 'p';
*arg_format_ptr = '\0';
FormatBuiltinArg(arg_format, arg.pointer_value, width, precision);
FormatDouble(arg.long_double_value, flags, width, precision, type);
break;
case CHAR:
// TODO: check if type is 'c' or none
if (width <= 1) {
buffer_.push_back(arg.int_value);
break;
}
*arg_format_ptr++ = 'c';
*arg_format_ptr = '\0';
FormatBuiltinArg(arg_format, arg.int_value, width, precision);
buffer_.reserve(std::max(width, 1));
buffer_.push_back(arg.int_value);
if (width > 1)
buffer_.resize(buffer_.size() + width - 1, ' ');
break;
case STRING:
// TODO: check if type is 's' or none
if (width == -1 || width <= arg.size) {
const char *str = arg.string_value;
size_t size = arg.size;
if (size == 0 && *str)
size = std::strlen(str);
buffer_.reserve(buffer_.size() + size + 1);
buffer_.insert(buffer_.end(), str, str + size);
break;
}
*arg_format_ptr++ = 's';
*arg_format_ptr = '\0';
FormatBuiltinArg(arg_format, arg.string_value, width, precision);
case STRING: {
const char *str = arg.string_value;
size_t size = arg.size;
if (size == 0 && *str)
size = std::strlen(str);
buffer_.reserve(buffer_.size() + std::max<size_t>(width, size));
buffer_.insert(buffer_.end(), str, str + size);
if (width > size)
buffer_.resize(buffer_.size() + width - size, ' ');
break;
case WSTRING:
// TODO: check if type is 's' or none
*arg_format_ptr++ = 'l';
*arg_format_ptr++ = 's';
*arg_format_ptr = '\0';
FormatBuiltinArg(arg_format, arg.wstring_value, width, precision);
}
case POINTER:
// TODO: don't allow type specifiers other than 'p'
FormatInt(reinterpret_cast<uintptr_t>(
arg.pointer_value), HEX_PREFIX_FLAG, width, 'x');
break;
case CUSTOM:
// TODO: check if type is 's' or none