commit 9c18854a0f3609515d8fb0a8d01c0021b666a948 Author: Willem Jan Palenstijn Date: Mon May 11 01:06:08 2009 +0200 Add Common::String::printf to format a string diff --git a/common/module.mk b/common/module.mk index bdf9590..fb145cb 100644 --- a/common/module.mk +++ b/common/module.mk @@ -11,6 +11,7 @@ MODULE_OBJS := \ memorypool.o \ md5.o \ mutex.o \ + printf.o \ str.o \ stream.o \ util.o \ diff --git a/common/printf.cpp b/common/printf.cpp new file mode 100644 index 0000000..f788b4d --- /dev/null +++ b/common/printf.cpp @@ -0,0 +1,599 @@ +/* + * The code in this file is a slightly modified version of + * snprintf.c from the snprintf package from Mark Martinec, available at + * http://www.ijs.si/software/snprintf/ + * + * Quoting that webpage: + * == + * This program is free software; it is dual licensed, + * the terms of the "Frontier Artistic License" or the + * "GNU General Public License" can be chosen at your discretion. + * The chosen license then applies solely and in its entirety + * == + * + * The modifications to the source consist of only using the + * snprintf function, without the various related functions, and + * replacing the output manipulation functions fast_memcpy and fast_memset + * with String-specific functions. + */ + + + + + + +/* + * snprintf.c - a portable implementation of snprintf + * + * AUTHOR + * Mark Martinec , April 1999. + * + * Copyright 1999, Mark Martinec. All rights reserved. + * + * TERMS AND CONDITIONS + * This program is free software; you can redistribute it and/or modify + * it under the terms of the "Frontier Artistic License" which comes + * with this Kit. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty + * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. + * See the Frontier Artistic License for more details. + * + * You should have received a copy of the Frontier Artistic License + * with this Kit in the file named LICENSE.txt . + * If not, I'll be glad to provide one. + * + * FEATURES + * - careful adherence to specs regarding flags, field width and precision; + * - good performance for large string handling (large format, large + * argument or large paddings). Performance is similar to system's sprintf + * and in several cases significantly better (make sure you compile with + * optimizations turned on, tell the compiler the code is strict ANSI + * if necessary to give it more freedom for optimizations); + * - return value semantics per ISO/IEC 9899:1999 ("ISO C99"); + * - written in standard ISO/ANSI C - requires an ANSI C compiler. + * + * SUPPORTED CONVERSION SPECIFIERS AND DATA TYPES + * + * This snprintf only supports the following conversion specifiers: + * s, c, d, u, o, x, X, p (and synonyms: i, D, U, O - see below) + * with flags: '-', '+', ' ', '0' and '#'. + * An asterisk is supported for field width as well as precision. + * + * Length modifiers 'h' (short int), 'l' (long int), + * and 'll' (long long int) are supported. + * NOTE: + * If macro SNPRINTF_LONGLONG_SUPPORT is not defined (default) the + * length modifier 'll' is recognized but treated the same as 'l', + * which may cause argument value truncation! Defining + * SNPRINTF_LONGLONG_SUPPORT requires that your system's sprintf also + * handles length modifier 'll'. long long int is a language extension + * which may not be portable. + * + * Conversion of numeric data (conversion specifiers d, u, o, x, X, p) + * with length modifiers (none or h, l, ll) is left to the system routine + * sprintf, but all handling of flags, field width and precision as well as + * c and s conversions is done very carefully by this portable routine. + * If a string precision (truncation) is specified (e.g. %.8s) it is + * guaranteed the string beyond the specified precision will not be referenced. + * + * Length modifiers h, l and ll are ignored for c and s conversions (data + * types wint_t and wchar_t are not supported). + * + * The following common synonyms for conversion characters are supported: + * - i is a synonym for d + * - D is a synonym for ld, explicit length modifiers are ignored + * - U is a synonym for lu, explicit length modifiers are ignored + * - O is a synonym for lo, explicit length modifiers are ignored + * The D, O and U conversion characters are nonstandard, they are supported + * for backward compatibility only, and should not be used for new code. + * + * The following is specifically NOT supported: + * - flag ' (thousands' grouping character) is recognized but ignored + * - numeric conversion specifiers: f, e, E, g, G and synonym F, + * as well as the new a and A conversion specifiers + * - length modifier 'L' (long double) and 'q' (quad - use 'll' instead) + * - wide character/string conversions: lc, ls, and nonstandard + * synonyms C and S + * - writeback of converted string length: conversion character n + * - the n$ specification for direct reference to n-th argument + * - locales + * + * It is permitted for str_m to be zero, and it is permitted to specify NULL + * pointer for resulting string argument if str_m is zero (as per ISO C99). + * + * [...] + * + * AVAILABILITY + * http://www.ijs.si/software/snprintf/ + * + */ + +#include "common/str.h" + +namespace Common { + + +#ifdef isdigit +#undef isdigit +#endif +#define isdigit(c) ((c) >= '0' && (c) <= '9') + +#define SNPRINTF_LONGLONG_SUPPORT + +void String::appendChars(char c, uint32 n) +{ + if (n == 0) + return; + + ensureCapacity(_size + n, true); + for (uint i = 0; i < n; ++i) + _str[_size++] = c; + _str[_size] = 0; +} + +void String::appendSubstring(const char *s, uint32 n) +{ + if (n == 0) + return; + + ensureCapacity(_size + n, true); + memcpy(_str + _size, s, n); + _size += n; + _str[_size] = 0; +} + +String String::printf(const char *fmt, /*args*/ ...) { + va_list ap; + + size_t str_l = 0; + const char *p = fmt; + String output; + +/* In contrast with POSIX, the ISO C99 now says + * that str can be NULL and str_m can be 0. + * This is more useful than the old: if (str_m < 1) return -1; */ + + va_start(ap, fmt); + + if (!p) p = ""; + while (*p) { + if (*p != '%') { + /* if (str_l < str_m) str[str_l++] = *p++; -- this would be sufficient */ + /* but the following code achieves better performance for cases + * where format string is long and contains few conversions */ + const char *q = strchr(p+1,'%'); + size_t n = !q ? strlen(p) : (q-p); + output.appendSubstring(p, n); + p += n; str_l += n; + } else { + const char *starting_p; + size_t min_field_width = 0, precision = 0; + int zero_padding = 0, precision_specified = 0, justify_left = 0; + int alternate_form = 0, force_sign = 0; + int space_for_positive = 1; /* If both the ' ' and '+' flags appear, + the ' ' flag should be ignored. */ + char length_modifier = '\0'; /* allowed values: \0, h, l, L */ + char tmp[32];/* temporary buffer for simple numeric->string conversion */ + + const char *str_arg; /* string address in case of string argument */ + size_t str_arg_l; /* natural field width of arg without padding + and sign */ + unsigned char uchar_arg; + /* unsigned char argument value - only defined for c conversion. + N.B. standard explicitly states the char argument for + the c conversion is unsigned */ + + size_t number_of_zeros_to_pad = 0; + /* number of zeros to be inserted for numeric conversions + as required by the precision or minimal field width */ + + size_t zero_padding_insertion_ind = 0; + /* index into tmp where zero padding is to be inserted */ + + char fmt_spec = '\0'; + /* current conversion specifier character */ + + str_arg = NULL; + starting_p = p; p++; /* skip '%' */ + /* parse flags */ + while (*p == '0' || *p == '-' || *p == '+' || + *p == ' ' || *p == '#' || *p == '\'') { + switch (*p) { + case '0': zero_padding = 1; break; + case '-': justify_left = 1; break; + case '+': force_sign = 1; space_for_positive = 0; break; + case ' ': force_sign = 1; + /* If both the ' ' and '+' flags appear, the ' ' flag should be ignored */ +#ifdef PERL_COMPATIBLE + /* ... but in Perl the last of ' ' and '+' applies */ + space_for_positive = 1; +#endif + break; + case '#': alternate_form = 1; break; + case '\'': break; + } + p++; + } + /* If the '0' and '-' flags both appear, the '0' flag should be ignored. */ + + /* parse field width */ + if (*p == '*') { + int j; + p++; j = va_arg(ap, int); + if (j >= 0) min_field_width = j; + else { min_field_width = -j; justify_left = 1; } + } else if (isdigit((int)(*p))) { + /* size_t could be wider than unsigned int; + make sure we treat argument like common implementations do */ + unsigned int uj = *p++ - '0'; + while (isdigit((int)(*p))) uj = 10*uj + (unsigned int)(*p++ - '0'); + min_field_width = uj; + } + /* parse precision */ + if (*p == '.') { + p++; precision_specified = 1; + if (*p == '*') { + int j = va_arg(ap, int); + p++; + if (j >= 0) precision = j; + else { + precision_specified = 0; precision = 0; + /* NOTE: + * Solaris 2.6 man page claims that in this case the precision + * should be set to 0. Digital Unix 4.0, HPUX 10 and BSD man page + * claim that this case should be treated as unspecified precision, + * which is what we do here. + */ + } + } else if (isdigit((int)(*p))) { + /* size_t could be wider than unsigned int; + make sure we treat argument like common implementations do */ + unsigned int uj = *p++ - '0'; + while (isdigit((int)(*p))) uj = 10*uj + (unsigned int)(*p++ - '0'); + precision = uj; + } + } + /* parse 'h', 'l' and 'll' length modifiers */ + if (*p == 'h' || *p == 'l') { + length_modifier = *p; p++; + if (length_modifier == 'l' && *p == 'l') { /* double l = long long */ +#ifdef SNPRINTF_LONGLONG_SUPPORT + length_modifier = '2'; /* double l encoded as '2' */ +#else + length_modifier = 'l'; /* treat it as a single 'l' */ +#endif + p++; + } + } + fmt_spec = *p; + /* common synonyms: */ + switch (fmt_spec) { + case 'i': fmt_spec = 'd'; break; + case 'D': fmt_spec = 'd'; length_modifier = 'l'; break; + case 'U': fmt_spec = 'u'; length_modifier = 'l'; break; + case 'O': fmt_spec = 'o'; length_modifier = 'l'; break; + default: break; + } + /* get parameter value, do initial processing */ + switch (fmt_spec) { + case '%': /* % behaves similar to 's' regarding flags and field widths */ + case 'c': /* c behaves similar to 's' regarding flags and field widths */ + case 's': + length_modifier = '\0'; /* wint_t and wchar_t not supported */ + /* the result of zero padding flag with non-numeric conversion specifier*/ + /* is undefined. Solaris and HPUX 10 does zero padding in this case, */ + /* Digital Unix and Linux does not. */ +#if !defined(SOLARIS_COMPATIBLE) && !defined(HPUX_COMPATIBLE) + zero_padding = 0; /* turn zero padding off for string conversions */ +#endif + str_arg_l = 1; + switch (fmt_spec) { + case '%': + str_arg = p; break; + case 'c': { + int j = va_arg(ap, int); + uchar_arg = (unsigned char) j; /* standard demands unsigned char */ + str_arg = (const char *) &uchar_arg; + break; + } + case 's': + str_arg = va_arg(ap, const char *); + if (!str_arg) str_arg_l = 0; + /* make sure not to address string beyond the specified precision !!! */ + else if (!precision_specified) str_arg_l = strlen(str_arg); + /* truncate string if necessary as requested by precision */ + else if (precision == 0) str_arg_l = 0; + else { + /* memchr on HP does not like n > 2^31 !!! */ + const char *q = (const char*)memchr(str_arg, '\0', + precision <= 0x7fffffff ? precision : 0x7fffffff); + str_arg_l = !q ? precision : (q-str_arg); + } + break; + default: break; + } + break; + case 'd': case 'u': case 'o': case 'x': case 'X': case 'p': { + /* NOTE: the u, o, x, X and p conversion specifiers imply + the value is unsigned; d implies a signed value */ + + int arg_sign = 0; + /* 0 if numeric argument is zero (or if pointer is NULL for 'p'), + +1 if greater than zero (or nonzero for unsigned arguments), + -1 if negative (unsigned argument is never negative) */ + + int int_arg = 0; unsigned int uint_arg = 0; + /* only defined for length modifier h, or for no length modifiers */ + + long int long_arg = 0; unsigned long int ulong_arg = 0; + /* only defined for length modifier l */ + + void *ptr_arg = NULL; + /* pointer argument value -only defined for p conversion */ + +#ifdef SNPRINTF_LONGLONG_SUPPORT + long long int long_long_arg = 0; + unsigned long long int ulong_long_arg = 0; + /* only defined for length modifier ll */ +#endif + if (fmt_spec == 'p') { + /* HPUX 10: An l, h, ll or L before any other conversion character + * (other than d, i, u, o, x, or X) is ignored. + * Digital Unix: + * not specified, but seems to behave as HPUX does. + * Solaris: If an h, l, or L appears before any other conversion + * specifier (other than d, i, u, o, x, or X), the behavior + * is undefined. (Actually %hp converts only 16-bits of address + * and %llp treats address as 64-bit data which is incompatible + * with (void *) argument on a 32-bit system). + */ +#ifdef SOLARIS_COMPATIBLE +# ifdef SOLARIS_BUG_COMPATIBLE + /* keep length modifiers even if it represents 'll' */ +# else + if (length_modifier == '2') length_modifier = '\0'; +# endif +#else + length_modifier = '\0'; +#endif + ptr_arg = va_arg(ap, void *); + if (ptr_arg != NULL) arg_sign = 1; + } else if (fmt_spec == 'd') { /* signed */ + switch (length_modifier) { + case '\0': + case 'h': + /* It is non-portable to specify a second argument of char or short + * to va_arg, because arguments seen by the called function + * are not char or short. C converts char and short arguments + * to int before passing them to a function. + */ + int_arg = va_arg(ap, int); + if (int_arg > 0) arg_sign = 1; + else if (int_arg < 0) arg_sign = -1; + break; + case 'l': + long_arg = va_arg(ap, long int); + if (long_arg > 0) arg_sign = 1; + else if (long_arg < 0) arg_sign = -1; + break; +#ifdef SNPRINTF_LONGLONG_SUPPORT + case '2': + long_long_arg = va_arg(ap, long long int); + if (long_long_arg > 0) arg_sign = 1; + else if (long_long_arg < 0) arg_sign = -1; + break; +#endif + } + } else { /* unsigned */ + switch (length_modifier) { + case '\0': + case 'h': + uint_arg = va_arg(ap, unsigned int); + if (uint_arg) arg_sign = 1; + break; + case 'l': + ulong_arg = va_arg(ap, unsigned long int); + if (ulong_arg) arg_sign = 1; + break; +#ifdef SNPRINTF_LONGLONG_SUPPORT + case '2': + ulong_long_arg = va_arg(ap, unsigned long long int); + if (ulong_long_arg) arg_sign = 1; + break; +#endif + } + } + str_arg = tmp; str_arg_l = 0; + /* NOTE: + * For d, i, u, o, x, and X conversions, if precision is specified, + * the '0' flag should be ignored. This is so with Solaris 2.6, + * Digital UNIX 4.0, HPUX 10, Linux, FreeBSD, NetBSD; but not with Perl. + */ +#ifndef PERL_COMPATIBLE + if (precision_specified) zero_padding = 0; +#endif + if (fmt_spec == 'd') { + if (force_sign && arg_sign >= 0) + tmp[str_arg_l++] = space_for_positive ? ' ' : '+'; + /* leave negative numbers for sprintf to handle, + to avoid handling tricky cases like (short int)(-32768) */ +#ifdef LINUX_COMPATIBLE + } else if (fmt_spec == 'p' && force_sign && arg_sign > 0) { + tmp[str_arg_l++] = space_for_positive ? ' ' : '+'; +#endif + } else if (alternate_form) { + if (arg_sign != 0 && (fmt_spec == 'x' || fmt_spec == 'X') ) + { tmp[str_arg_l++] = '0'; tmp[str_arg_l++] = fmt_spec; } + /* alternate form should have no effect for p conversion, but ... */ +#ifdef HPUX_COMPATIBLE + else if (fmt_spec == 'p' + /* HPUX 10: for an alternate form of p conversion, + * a nonzero result is prefixed by 0x. */ +#ifndef HPUX_BUG_COMPATIBLE + /* Actually it uses 0x prefix even for a zero value. */ + && arg_sign != 0 +#endif + ) { tmp[str_arg_l++] = '0'; tmp[str_arg_l++] = 'x'; } +#endif + } + zero_padding_insertion_ind = str_arg_l; + if (!precision_specified) precision = 1; /* default precision is 1 */ + if (precision == 0 && arg_sign == 0 +#if defined(HPUX_BUG_COMPATIBLE) || defined(LINUX_COMPATIBLE) + && fmt_spec != 'p' + /* HPUX 10 man page claims: With conversion character p the result of + * converting a zero value with a precision of zero is a null string. + * Actually HP returns all zeroes, and Linux returns "(nil)". */ +#endif + ) { + /* converted to null string */ + /* When zero value is formatted with an explicit precision 0, + the resulting formatted string is empty (d, i, u, o, x, X, p). */ + } else { + char f[5]; int f_l = 0; + f[f_l++] = '%'; /* construct a simple format string for sprintf */ + if (!length_modifier) { } + else if (length_modifier=='2') { f[f_l++] = 'l'; f[f_l++] = 'l'; } + else f[f_l++] = length_modifier; + f[f_l++] = fmt_spec; f[f_l++] = '\0'; + if (fmt_spec == 'p') str_arg_l += sprintf(tmp+str_arg_l, f, ptr_arg); + else if (fmt_spec == 'd') { /* signed */ + switch (length_modifier) { + case '\0': + case 'h': str_arg_l+=sprintf(tmp+str_arg_l, f, int_arg); break; + case 'l': str_arg_l+=sprintf(tmp+str_arg_l, f, long_arg); break; +#ifdef SNPRINTF_LONGLONG_SUPPORT + case '2': str_arg_l+=sprintf(tmp+str_arg_l,f,long_long_arg); break; +#endif + } + } else { /* unsigned */ + switch (length_modifier) { + case '\0': + case 'h': str_arg_l+=sprintf(tmp+str_arg_l, f, uint_arg); break; + case 'l': str_arg_l+=sprintf(tmp+str_arg_l, f, ulong_arg); break; +#ifdef SNPRINTF_LONGLONG_SUPPORT + case '2': str_arg_l+=sprintf(tmp+str_arg_l,f,ulong_long_arg);break; +#endif + } + } + /* include the optional minus sign and possible "0x" + in the region before the zero padding insertion point */ + if (zero_padding_insertion_ind < str_arg_l && + tmp[zero_padding_insertion_ind] == '-') { + zero_padding_insertion_ind++; + } + if (zero_padding_insertion_ind+1 < str_arg_l && + tmp[zero_padding_insertion_ind] == '0' && + (tmp[zero_padding_insertion_ind+1] == 'x' || + tmp[zero_padding_insertion_ind+1] == 'X') ) { + zero_padding_insertion_ind += 2; + } + } + { size_t num_of_digits = str_arg_l - zero_padding_insertion_ind; + if (alternate_form && fmt_spec == 'o' +#ifdef HPUX_COMPATIBLE /* ("%#.o",0) -> "" */ + && (str_arg_l > 0) +#endif +#ifdef DIGITAL_UNIX_BUG_COMPATIBLE /* ("%#o",0) -> "00" */ +#else + /* unless zero is already the first character */ + && !(zero_padding_insertion_ind < str_arg_l + && tmp[zero_padding_insertion_ind] == '0') +#endif + ) { /* assure leading zero for alternate-form octal numbers */ + if (!precision_specified || precision < num_of_digits+1) { + /* precision is increased to force the first character to be zero, + except if a zero value is formatted with an explicit precision + of zero */ + precision = num_of_digits+1; precision_specified = 1; + } + } + /* zero padding to specified precision? */ + if (num_of_digits < precision) + number_of_zeros_to_pad = precision - num_of_digits; + } + /* zero padding to specified minimal field width? */ + if (!justify_left && zero_padding) { + int n = min_field_width - (str_arg_l+number_of_zeros_to_pad); + if (n > 0) number_of_zeros_to_pad += n; + } + break; + } + default: /* unrecognized conversion specifier, keep format string as-is*/ + zero_padding = 0; /* turn zero padding off for non-numeric convers. */ +#ifndef DIGITAL_UNIX_COMPATIBLE + justify_left = 1; min_field_width = 0; /* reset flags */ +#endif +#if defined(PERL_COMPATIBLE) || defined(LINUX_COMPATIBLE) + /* keep the entire format string unchanged */ + str_arg = starting_p; str_arg_l = p - starting_p; + /* well, not exactly so for Linux, which does something inbetween, + * and I don't feel an urge to imitate it: "%+++++hy" -> "%+y" */ +#else + /* discard the unrecognized conversion, just keep * + * the unrecognized conversion character */ + str_arg = p; str_arg_l = 0; +#endif + if (*p) str_arg_l++; /* include invalid conversion specifier unchanged + if not at end-of-string */ + break; + } + if (*p) p++; /* step over the just processed conversion specifier */ + /* insert padding to the left as requested by min_field_width; + this does not include the zero padding in case of numerical conversions*/ + if (!justify_left) { /* left padding with blank or zero */ + int n = min_field_width - (str_arg_l+number_of_zeros_to_pad); + if (n > 0) { + output.appendChars((zero_padding?'0':' '), n); + str_l += n; + } + } + /* zero padding as requested by the precision or by the minimal field width + * for numeric conversions required? */ + if (number_of_zeros_to_pad <= 0) { + /* will not copy first part of numeric right now, * + * force it to be copied later in its entirety */ + zero_padding_insertion_ind = 0; + } else { + /* insert first part of numerics (sign or '0x') before zero padding */ + int n = zero_padding_insertion_ind; + if (n > 0) { + output.appendSubstring(str_arg, n); + str_l += n; + } + /* insert zero padding as requested by the precision or min field width */ + n = number_of_zeros_to_pad; + if (n > 0) { + output.appendChars('0', n); + str_l += n; + } + } + /* insert formatted string + * (or as-is conversion specifier for unknown conversions) */ + { int n = str_arg_l - zero_padding_insertion_ind; + if (n > 0) { + output.appendSubstring(str_arg+zero_padding_insertion_ind, n); + str_l += n; + } + } + /* insert right padding */ + if (justify_left) { /* right blank padding to the field width */ + int n = min_field_width - (str_arg_l+number_of_zeros_to_pad); + if (n > 0) { + output.appendChars(' ', n); + str_l += n; + } + } + } + } + + va_end(ap); + + return output; +} + + +} // End of namespace Common + diff --git a/common/str.h b/common/str.h index b7dbd65..a1a04e0 100644 --- a/common/str.h +++ b/common/str.h @@ -218,6 +218,25 @@ public: uint hash() const; + /** + * Printf-like function. Returns a formatted String. + */ + static Common::String printf(const char *fmt, ...); + +protected: + /** + * Helper function for printf: + * Append the first n characters from s. + * s is assumed to be long enough. + */ + void appendSubstring(const char *s, uint32 n); + + /** + * Helper function for printf: + * Append n copies of c. + */ + void appendChars(char c, uint32 n); + public: typedef char * iterator; typedef const char * const_iterator; diff --git a/dists/msvc8/scummvm.vcproj b/dists/msvc8/scummvm.vcproj index 27e37ad..45ff1fd 100644 --- a/dists/msvc8/scummvm.vcproj +++ b/dists/msvc8/scummvm.vcproj @@ -70,6 +70,7 @@ + diff --git a/dists/msvc9/scummvm.vcproj b/dists/msvc9/scummvm.vcproj index edcf9c4..9040fe1 100644 --- a/dists/msvc9/scummvm.vcproj +++ b/dists/msvc9/scummvm.vcproj @@ -71,6 +71,7 @@ + diff --git a/test/common/str.h b/test/common/str.h index bf0db98..a860946 100644 --- a/test/common/str.h +++ b/test/common/str.h @@ -285,4 +285,10 @@ class StringTestSuite : public CxxTest::TestSuite TS_ASSERT(!Common::matchString("monkey.s99", "monkey.s*1")); TS_ASSERT(Common::matchString("monkey.s101", "monkey.s*1")); } + + void test_string_printf() { + TS_ASSERT( Common::String::printf("%s", "test") == "test" ); + TS_ASSERT( Common::String::printf("%s.s%.02d", "monkey", 1) == "monkey.s01" ); + TS_ASSERT( Common::String::printf("%s%X", "test", 1234) == "test4D2" ); + } };