diff options
author | Damien George <damien.p.george@gmail.com> | 2014-04-01 10:48:58 +0100 |
---|---|---|
committer | Damien George <damien.p.george@gmail.com> | 2014-04-01 10:48:58 +0100 |
commit | 46330bd9b52bdd8865be62c1f6c534ef7709b842 (patch) | |
tree | dc531b159f6e5d3c3940e5fad752df9692e4ec60 /py | |
parent | e44d26ae0c1b5d248fa4db112cdeabe404944f3c (diff) | |
parent | baf6f14deb567ab626c1b05213af346108f41700 (diff) | |
download | micropython-46330bd9b52bdd8865be62c1f6c534ef7709b842.tar.gz micropython-46330bd9b52bdd8865be62c1f6c534ef7709b842.zip |
Merge pull request #407 from dhylands/str-format
Enhance str.format support
Diffstat (limited to 'py')
-rw-r--r-- | py/formatfloat.c | 1 | ||||
-rw-r--r-- | py/obj.c | 4 | ||||
-rw-r--r-- | py/objstr.c | 384 | ||||
-rw-r--r-- | py/pfenv.c | 208 | ||||
-rw-r--r-- | py/pfenv.h | 23 | ||||
-rw-r--r-- | py/py.mk | 1 |
6 files changed, 610 insertions, 11 deletions
diff --git a/py/formatfloat.c b/py/formatfloat.c index 83aee1ef74..b1c62fa6d5 100644 --- a/py/formatfloat.c +++ b/py/formatfloat.c @@ -14,6 +14,7 @@ ***********************************************************************/ #include <stdlib.h> +#include <stdint.h> #include "mpconfig.h" @@ -198,6 +198,10 @@ machine_int_t mp_obj_get_int(mp_obj_t arg) { return MP_OBJ_SMALL_INT_VALUE(arg); } else if (MP_OBJ_IS_TYPE(arg, &mp_type_int)) { return mp_obj_int_get_checked(arg); +#if MICROPY_ENABLE_FLOAT + } else if (MP_OBJ_IS_TYPE(arg, &mp_type_float)) { + return mp_obj_float_get(arg); +#endif } else { nlr_jump(mp_obj_new_exception_msg_varg(&mp_type_TypeError, "can't convert %s to int", mp_obj_get_type_str(arg))); } diff --git a/py/objstr.c b/py/objstr.c index 936542b0e9..8389bb0bdf 100644 --- a/py/objstr.c +++ b/py/objstr.c @@ -9,6 +9,7 @@ #include "obj.h" #include "runtime0.h" #include "runtime.h" +#include "pfenv.h" typedef struct _mp_obj_str_t { mp_obj_base_t base; @@ -492,28 +493,389 @@ STATIC mp_obj_t str_strip(uint n_args, const mp_obj_t *args) { return mp_obj_new_str(orig_str + first_good_char_pos, stripped_len, false); } +// Takes an int arg, but only parses unsigned numbers, and only changes +// *num if at least one digit was parsed. +static int str_to_int(const char *str, int *num) { + const char *s = str; + if (unichar_isdigit(*s)) { + *num = 0; + do { + *num = *num * 10 + (*s - '0'); + s++; + } + while (unichar_isdigit(*s)); + } + return s - str; +} + +static bool isalignment(char ch) { + return ch && strchr("<>=^", ch) != NULL; +} + +static bool istype(char ch) { + return ch && strchr("bcdeEfFgGnosxX%", ch) != NULL; +} + +static bool arg_looks_integer(mp_obj_t arg) { + return MP_OBJ_IS_TYPE(arg, &mp_type_bool) || MP_OBJ_IS_INT(arg); +} + +static bool arg_looks_numeric(mp_obj_t arg) { + return arg_looks_integer(arg) +#if MICROPY_ENABLE_FLOAT + || MP_OBJ_IS_TYPE(arg, &mp_type_float) +#endif + ; +} + mp_obj_t str_format(uint n_args, const mp_obj_t *args) { assert(MP_OBJ_IS_STR(args[0])); GET_STR_DATA_LEN(args[0], str, len); - int arg_i = 1; + int arg_i = 0; vstr_t *vstr = vstr_new(); + pfenv_t pfenv_vstr; + pfenv_vstr.data = vstr; + pfenv_vstr.print_strn = pfenv_vstr_add_strn; + for (const byte *top = str + len; str < top; str++) { - if (*str == '{') { + if (*str == '}') { str++; - if (str < top && *str == '{') { - vstr_add_char(vstr, '{'); + if (str < top && *str == '}') { + vstr_add_char(vstr, '}'); + continue; + } + nlr_jump(mp_obj_new_exception_msg(&mp_type_ValueError, "Single '}' encountered in format string")); + } + if (*str != '{') { + vstr_add_char(vstr, *str); + continue; + } + + str++; + if (str < top && *str == '{') { + vstr_add_char(vstr, '{'); + continue; + } + + // replacement_field ::= "{" [field_name] ["!" conversion] [":" format_spec] "}" + + vstr_t *field_name = NULL; + char conversion = '\0'; + vstr_t *format_spec = NULL; + + if (str < top && *str != '}' && *str != '!' && *str != ':') { + field_name = vstr_new(); + while (str < top && *str != '}' && *str != '!' && *str != ':') { + vstr_add_char(field_name, *str++); + } + vstr_add_char(field_name, '\0'); + } + + // conversion ::= "r" | "s" + + if (str < top && *str == '!') { + str++; + if (str < top && (*str == 'r' || *str == 's')) { + conversion = *str++; } else { - while (str < top && *str != '}') str++; - if (arg_i >= n_args) { - nlr_jump(mp_obj_new_exception_msg(&mp_type_IndexError, "tuple index out of range")); + nlr_jump(mp_obj_new_exception_msg(&mp_type_ValueError, "end of format while looking for conversion specifier")); + } + } + + if (str < top && *str == ':') { + str++; + // {:} is the same as {}, which is the same as {!s} + // This makes a difference when passing in a True or False + // '{}'.format(True) returns 'True' + // '{:d}'.format(True) returns '1' + // So we treat {:} as {} and this later gets treated to be {!s} + if (*str != '}') { + format_spec = vstr_new(); + while (str < top && *str != '}') { + vstr_add_char(format_spec, *str++); } - // TODO: may be PRINT_REPR depending on formatting code - mp_obj_print_helper((void (*)(void*, const char*, ...))vstr_printf, vstr, args[arg_i], PRINT_STR); - arg_i++; + vstr_add_char(format_spec, '\0'); } + } + if (str >= top) { + nlr_jump(mp_obj_new_exception_msg(&mp_type_ValueError, "unmatched '{' in format")); + } + if (*str != '}') { + nlr_jump(mp_obj_new_exception_msg(&mp_type_ValueError, "expected ':' after format specifier")); + } + + mp_obj_t arg = mp_const_none; + + if (field_name) { + if (arg_i > 0) { + nlr_jump(mp_obj_new_exception_msg(&mp_type_ValueError, "cannot switch from automatic field numbering to manual field specification")); + } + int index; + if (str_to_int(vstr_str(field_name), &index) != vstr_len(field_name) - 1) { + nlr_jump(mp_obj_new_exception_msg(&mp_type_KeyError, "attributes not supported yet")); + } + if (index >= n_args - 1) { + nlr_jump(mp_obj_new_exception_msg(&mp_type_IndexError, "tuple index out of range")); + } + arg = args[index + 1]; + arg_i = -1; + vstr_free(field_name); + field_name = NULL; } else { - vstr_add_char(vstr, *str); + if (arg_i < 0) { + nlr_jump(mp_obj_new_exception_msg(&mp_type_ValueError, "cannot switch from manual field specification to automatic field numbering")); + } + if (arg_i >= n_args - 1) { + nlr_jump(mp_obj_new_exception_msg(&mp_type_IndexError, "tuple index out of range")); + } + arg = args[arg_i + 1]; + arg_i++; + } + if (!format_spec && !conversion) { + conversion = 's'; + } + if (conversion) { + mp_print_kind_t print_kind; + if (conversion == 's') { + print_kind = PRINT_STR; + } else if (conversion == 'r') { + print_kind = PRINT_REPR; + } else { + nlr_jump(mp_obj_new_exception_msg_varg(&mp_type_ValueError, "Unknown conversion specifier %c", conversion)); + } + vstr_t *arg_vstr = vstr_new(); + mp_obj_print_helper((void (*)(void*, const char*, ...))vstr_printf, arg_vstr, arg, print_kind); + arg = mp_obj_new_str((const byte *)vstr_str(arg_vstr), vstr_len(arg_vstr), false); + vstr_free(arg_vstr); + } + + char sign = '\0'; + char fill = '\0'; + char align = '\0'; + int width = -1; + int precision = -1; + char type = '\0'; + int flags = 0; + + if (format_spec) { + // The format specifier (from http://docs.python.org/2/library/string.html#formatspec) + // + // [[fill]align][sign][#][0][width][,][.precision][type] + // fill ::= <any character> + // align ::= "<" | ">" | "=" | "^" + // sign ::= "+" | "-" | " " + // width ::= integer + // precision ::= integer + // type ::= "b" | "c" | "d" | "e" | "E" | "f" | "F" | "g" | "G" | "n" | "o" | "s" | "x" | "X" | "%" + + const char *s = vstr_str(format_spec); + if (isalignment(*s)) { + align = *s++; + } else if (*s && isalignment(s[1])) { + fill = *s++; + align = *s++; + } + if (*s == '+' || *s == '-' || *s == ' ') { + if (*s == '+') { + flags |= PF_FLAG_SHOW_SIGN; + } else if (*s == ' ') { + flags |= PF_FLAG_SPACE_SIGN; + } + sign = *s++; + } + if (*s == '#') { + flags |= PF_FLAG_SHOW_PREFIX; + s++; + } + if (*s == '0') { + if (!align) { + align = '='; + } + if (!fill) { + fill = '0'; + } + } + s += str_to_int(s, &width); + if (*s == ',') { + flags |= PF_FLAG_SHOW_COMMA; + s++; + } + if (*s == '.') { + s++; + s += str_to_int(s, &precision); + } + if (istype(*s)) { + type = *s++; + } + if (*s) { + nlr_jump(mp_obj_new_exception_msg(&mp_type_KeyError, "Invalid conversion specification")); + } + vstr_free(format_spec); + format_spec = NULL; + } + if (!align) { + if (arg_looks_numeric(arg)) { + align = '>'; + } else { + align = '<'; + } + } + if (!fill) { + fill = ' '; + } + + if (sign) { + if (type == 's') { + nlr_jump(mp_obj_new_exception_msg(&mp_type_ValueError, "Sign not allowed in string format specifier")); + } + if (type == 'c') { + nlr_jump(mp_obj_new_exception_msg(&mp_type_ValueError, "Sign not allowed with integer format specifier 'c'")); + } + } else { + sign = '-'; + } + + switch (align) { + case '<': flags |= PF_FLAG_LEFT_ADJUST; break; + case '=': flags |= PF_FLAG_PAD_AFTER_SIGN; break; + case '^': flags |= PF_FLAG_CENTER_ADJUST; break; + } + + if (arg_looks_integer(arg)) { + switch (type) { + case 'b': + pfenv_print_int(&pfenv_vstr, mp_obj_get_int(arg), 1, 2, 'a', flags, fill, width); + continue; + + case 'c': + { + char ch = mp_obj_get_int(arg); + pfenv_print_strn(&pfenv_vstr, &ch, 1, flags, fill, width); + continue; + } + + case '\0': // No explicit format type implies 'd' + case 'n': // I don't think we support locales in uPy so use 'd' + case 'd': + pfenv_print_int(&pfenv_vstr, mp_obj_get_int(arg), 1, 10, 'a', flags, fill, width); + continue; + + case 'o': + pfenv_print_int(&pfenv_vstr, mp_obj_get_int(arg), 1, 8, 'a', flags, fill, width); + continue; + + case 'x': + pfenv_print_int(&pfenv_vstr, mp_obj_get_int(arg), 1, 16, 'a', flags, fill, width); + continue; + + case 'X': + pfenv_print_int(&pfenv_vstr, mp_obj_get_int(arg), 1, 16, 'A', flags, fill, width); + continue; + + case 'e': + case 'E': + case 'f': + case 'F': + case 'g': + case 'G': + case '%': + // The floating point formatters all work with anything that + // looks like an integer + break; + + default: + nlr_jump(mp_obj_new_exception_msg_varg(&mp_type_ValueError, + "Unknown format code '%c' for object of type '%s'", type, mp_obj_get_type_str(arg))); + } + } +#if MICROPY_ENABLE_FLOAT + if (arg_looks_numeric(arg)) { + if (!type) { + + // Even though the docs say that an unspecified type is the same + // as 'g', there is one subtle difference, when the exponent + // is one less than the precision. + // + // '{:10.1}'.format(0.0) ==> '0e+00' + // '{:10.1g}'.format(0.0) ==> '0' + // + // TODO: Figure out how to deal with this. + // + // A proper solution would involve adding a special flag + // or something to format_float, and create a format_double + // to deal with doubles. In order to fix this when using + // sprintf, we'd need to use the e format and tweak the + // returned result to strip trailing zeros like the g format + // does. + // + // {:10.3} and {:10.2e} with 1.23e2 both produce 1.23e+02 + // but with 1.e2 you get 1e+02 and 1.00e+02 + // + // Stripping the trailing 0's (like g) does would make the + // e format give us the right format. + // + // CPython sources say: + // Omitted type specifier. Behaves in the same way as repr(x) + // and str(x) if no precision is given, else like 'g', but with + // at least one digit after the decimal point. */ + + type = 'g'; + } + if (type == 'n') { + type = 'g'; + } + + flags |= PF_FLAG_PAD_NAN_INF; // '{:06e}'.format(float('-inf')) should give '-00inf' + switch (type) { + case 'e': + case 'E': + case 'f': + case 'F': + case 'g': + case 'G': + pfenv_print_float(&pfenv_vstr, mp_obj_get_float(arg), type, flags, fill, width, precision); + break; + + case '%': + flags |= PF_FLAG_ADD_PERCENT; + pfenv_print_float(&pfenv_vstr, mp_obj_get_float(arg) * 100.0F, 'f', flags, fill, width, precision); + break; + + default: + nlr_jump(mp_obj_new_exception_msg_varg(&mp_type_ValueError, + "Unknown format code '%c' for object of type 'float'", + type, mp_obj_get_type_str(arg))); + } +#endif + } else { + if (align == '=') { + nlr_jump(mp_obj_new_exception_msg(&mp_type_ValueError, "'=' alignment not allowed in string format specifier")); + } + switch (type) { + case '\0': + mp_obj_print_helper((void (*)(void*, const char*, ...))vstr_printf, vstr, arg, PRINT_STR); + break; + + case 's': + { + uint len; + const char *s = mp_obj_str_get_data(arg, &len); + if (precision < 0) { + precision = len; + } + if (len > precision) { + len = precision; + } + pfenv_print_strn(&pfenv_vstr, s, len, flags, fill, width); + break; + } + + default: + nlr_jump(mp_obj_new_exception_msg_varg(&mp_type_ValueError, + "Unknown format code '%c' for object of type 'str'", + type, mp_obj_get_type_str(arg))); + } } } diff --git a/py/pfenv.c b/py/pfenv.c new file mode 100644 index 0000000000..07f35c335c --- /dev/null +++ b/py/pfenv.c @@ -0,0 +1,208 @@ +#include <stdint.h> +#include <string.h> + +///#include "std.h" +#include "misc.h" +#include "mpconfig.h" +#include "qstr.h" +#include "obj.h" +#include "pfenv.h" + +#if MICROPY_FLOAT_IMPL == MICROPY_FLOAT_IMPL_DOUBLE +#include <stdio.h> +#endif + +#if MICROPY_ENABLE_FLOAT +#include "formatfloat.h" +#endif + +#define PF_PAD_SIZE 16 +static const char *pad_spaces = " "; +static const char *pad_zeroes = "0000000000000000"; + +void pfenv_vstr_add_strn(void *data, const char *str, unsigned int len){ + vstr_add_strn(data, str, len); +} + +int pfenv_print_strn(const pfenv_t *pfenv, const char *str, unsigned int len, int flags, char fill, int width) { + int left_pad = 0; + int right_pad = 0; + int pad = width - len; + char pad_fill[PF_PAD_SIZE]; + const char *pad_chars; + + if (!fill || fill == ' ' ) { + pad_chars = pad_spaces; + } else if (fill == '0') { + pad_chars = pad_zeroes; + } else { + memset(pad_fill, fill, PF_PAD_SIZE); + pad_chars = pad_fill; + } + + if (flags & PF_FLAG_CENTER_ADJUST) { + left_pad = pad / 2; + right_pad = pad - left_pad; + } else if (flags & PF_FLAG_LEFT_ADJUST) { + right_pad = pad; + } else { + left_pad = pad; + } + + if (left_pad) { + while (left_pad > 0) { + int p = left_pad; + if (p > PF_PAD_SIZE) + p = PF_PAD_SIZE; + pfenv->print_strn(pfenv->data, pad_chars, p); + left_pad -= p; + } + } + pfenv->print_strn(pfenv->data, str, len); + if (right_pad) { + while (right_pad > 0) { + int p = right_pad; + if (p > PF_PAD_SIZE) + p = PF_PAD_SIZE; + pfenv->print_strn(pfenv->data, pad_chars, p); + right_pad -= p; + } + } + return len; +} + +// enough room for 32 signed number +#define INT_BUF_SIZE (16) + +int pfenv_print_int(const pfenv_t *pfenv, unsigned int x, int sgn, int base, int base_char, int flags, char fill, int width) { + char sign = 0; + if (sgn) { + if ((int)x < 0) { + sign = '-'; + x = -x; + } else if (flags & PF_FLAG_SHOW_SIGN) { + sign = '+'; + } else if (flags & PF_FLAG_SPACE_SIGN) { + sign = ' '; + } + } + + char buf[INT_BUF_SIZE]; + char *b = buf + INT_BUF_SIZE; + + if (x == 0) { + *(--b) = '0'; + } else { + do { + int c = x % base; + x /= base; + if (c >= 10) { + c += base_char - 10; + } else { + c += '0'; + } + *(--b) = c; + } while (b > buf && x != 0); + } + + char prefix_char = '\0'; + + if (flags & PF_FLAG_SHOW_PREFIX) { + if (base == 2) { + prefix_char = base_char + 'b' - 'a'; + } else if (base == 8) { + prefix_char = base_char + 'o' - 'a'; + } else if (base == 16) { + prefix_char = base_char + 'x' - 'a'; + } + } + + int len = 0; + if (flags & PF_FLAG_PAD_AFTER_SIGN) { + if (sign) { + len += pfenv_print_strn(pfenv, &sign, 1, flags, fill, 1); + width--; + } + if (prefix_char) { + len += pfenv_print_strn(pfenv, "0", 1, flags, fill, 1); + len += pfenv_print_strn(pfenv, &prefix_char, 1, flags, fill, 1); + width -= 2; + } + } else { + if (prefix_char && b > &buf[1]) { + *(--b) = prefix_char; + *(--b) = '0'; + } + if (sign && b > buf) { + *(--b) = sign; + } + } + + len += pfenv_print_strn(pfenv, b, buf + INT_BUF_SIZE - b, flags, fill, width); + return len; +} + +#if MICROPY_ENABLE_FLOAT +int pfenv_print_float(const pfenv_t *pfenv, mp_float_t f, char fmt, int flags, char fill, int width, int prec) { + char buf[32]; + char sign = '\0'; + int chrs = 0; + + if (flags & PF_FLAG_SHOW_SIGN) { + sign = '+'; + } + else + if (flags & PF_FLAG_SPACE_SIGN) { + sign = ' '; + } + int len; +#if MICROPY_FLOAT_IMPL == MICROPY_FLOAT_IMPL_FLOAT + len = format_float(f, buf, sizeof(buf), fmt, prec, sign); +#elif MICROPY_FLOAT_IMPL == MICROPY_FLOAT_IMPL_DOUBLE + char fmt_buf[6]; + char *fmt_s = fmt_buf; + + *fmt_s++ = '%'; + if (sign) { + *fmt_s++ = sign; + } + *fmt_s++ = '.'; + *fmt_s++ = '*'; + *fmt_s++ = fmt; + *fmt_s = '\0'; + + len = snprintf(buf, sizeof(buf), fmt_buf, prec, f); +#else +#error Unknown MICROPY FLOAT IMPL +#endif + char *s = buf; + + if ((flags & PF_FLAG_ADD_PERCENT) && (len + 1) < sizeof(buf)) { + buf[len++] = '%'; + buf[len] = '\0'; + } + + // buf[0] < '0' returns true if the first character is space, + or - + if ((flags & PF_FLAG_PAD_AFTER_SIGN) && buf[0] < '0') { + // We have a sign character + s++; + if (*s <= '9' || (flags & PF_FLAG_PAD_NAN_INF)) { + // We have a number, or we have a inf/nan and PAD_NAN_INF is set + // With '{:06e}'.format(float('-inf')) you get '-00inf' + chrs += pfenv_print_strn(pfenv, &buf[0], 1, 0, 0, 1); + width--; + len--; + } + } + + if (*s > 'A' && (flags & PF_FLAG_PAD_NAN_INF) == 0) { + // We have one of the inf or nan variants, suppress zero fill. + // With printf, if you use: printf("%06e", -inf) then you get " -inf" + // so suppress the zero fill. + fill = ' '; + } + chrs += pfenv_print_strn(pfenv, s, len, flags, fill, width); + + return chrs; +} +#endif diff --git a/py/pfenv.h b/py/pfenv.h new file mode 100644 index 0000000000..edceaf3e4e --- /dev/null +++ b/py/pfenv.h @@ -0,0 +1,23 @@ +#define PF_FLAG_LEFT_ADJUST (0x001) +#define PF_FLAG_SHOW_SIGN (0x002) +#define PF_FLAG_SPACE_SIGN (0x004) +#define PF_FLAG_NO_TRAILZ (0x008) +#define PF_FLAG_SHOW_PREFIX (0x010) +#define PF_FLAG_SHOW_COMMA (0x020) +#define PF_FLAG_PAD_AFTER_SIGN (0x040) +#define PF_FLAG_CENTER_ADJUST (0x080) +#define PF_FLAG_ADD_PERCENT (0x100) +#define PF_FLAG_PAD_NAN_INF (0x200) + +typedef struct _pfenv_t { + void *data; + void (*print_strn)(void *, const char *str, unsigned int len); +} pfenv_t; + +void pfenv_vstr_add_strn(void *data, const char *str, unsigned int len); + +int pfenv_print_strn(const pfenv_t *pfenv, const char *str, unsigned int len, int flags, char fill, int width); +int pfenv_print_int(const pfenv_t *pfenv, unsigned int x, int sgn, int base, int base_char, int flags, char fill, int width); +#if MICROPY_ENABLE_FLOAT +int pfenv_print_float(const pfenv_t *pfenv, mp_float_t f, char fmt, int flags, char fill, int width, int prec); +#endif @@ -84,6 +84,7 @@ PY_O_BASENAME = \ showbc.o \ repl.o \ intdivmod.o \ + pfenv.o \ # prepend the build destination prefix to the py object files PY_O = $(addprefix $(PY_BUILD)/, $(PY_O_BASENAME)) |