1/*
2 * This file is part of the MicroPython project, http://micropython.org/
3 *
4 * The MIT License (MIT)
5 *
6 * Copyright (c) 2013, 2014 Damien P. George
7 *
8 * Permission is hereby granted, free of charge, to any person obtaining a copy
9 * of this software and associated documentation files (the "Software"), to deal
10 * in the Software without restriction, including without limitation the rights
11 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
12 * copies of the Software, and to permit persons to whom the Software is
13 * furnished to do so, subject to the following conditions:
14 *
15 * The above copyright notice and this permission notice shall be included in
16 * all copies or substantial portions of the Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
19 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
20 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
21 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
22 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
23 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24 * THE SOFTWARE.
25 */
26
27#include <stdlib.h>
28#include <assert.h>
29#include <string.h>
30
31#include "py/parsenum.h"
32#include "py/smallint.h"
33#include "py/objint.h"
34#include "py/objstr.h"
35#include "py/runtime.h"
36#include "py/binary.h"
37
38#if MICROPY_PY_BUILTINS_FLOAT
39#include <math.h>
40#endif
41
42// This dispatcher function is expected to be independent of the implementation of long int
43STATIC mp_obj_t mp_obj_int_make_new(const mp_obj_type_t *type_in, size_t n_args, size_t n_kw, const mp_obj_t *args) {
44 (void)type_in;
45 mp_arg_check_num(n_args, n_kw, 0, 2, false);
46
47 switch (n_args) {
48 case 0:
49 return MP_OBJ_NEW_SMALL_INT(0);
50
51 case 1:
52 if (mp_obj_is_int(args[0])) {
53 // already an int (small or long), just return it
54 return args[0];
55 } else if (mp_obj_is_str_or_bytes(args[0])) {
56 // a string, parse it
57 size_t l;
58 const char *s = mp_obj_str_get_data(args[0], &l);
59 return mp_parse_num_integer(s, l, 0, NULL);
60 #if MICROPY_PY_BUILTINS_FLOAT
61 } else if (mp_obj_is_float(args[0])) {
62 return mp_obj_new_int_from_float(mp_obj_float_get(args[0]));
63 #endif
64 } else {
65 return mp_unary_op(MP_UNARY_OP_INT, args[0]);
66 }
67
68 case 2:
69 default: {
70 // should be a string, parse it
71 size_t l;
72 const char *s = mp_obj_str_get_data(args[0], &l);
73 return mp_parse_num_integer(s, l, mp_obj_get_int(args[1]), NULL);
74 }
75 }
76}
77
78#if MICROPY_PY_BUILTINS_FLOAT
79
80typedef enum {
81 MP_FP_CLASS_FIT_SMALLINT,
82 MP_FP_CLASS_FIT_LONGINT,
83 MP_FP_CLASS_OVERFLOW
84} mp_fp_as_int_class_t;
85
86STATIC mp_fp_as_int_class_t mp_classify_fp_as_int(mp_float_t val) {
87 union {
88 mp_float_t f;
89 #if MICROPY_FLOAT_IMPL == MICROPY_FLOAT_IMPL_FLOAT
90 uint32_t i;
91 #elif MICROPY_FLOAT_IMPL == MICROPY_FLOAT_IMPL_DOUBLE
92 uint32_t i[2];
93 #endif
94 } u = {val};
95
96 uint32_t e;
97 #if MICROPY_FLOAT_IMPL == MICROPY_FLOAT_IMPL_FLOAT
98 e = u.i;
99 #elif MICROPY_FLOAT_IMPL == MICROPY_FLOAT_IMPL_DOUBLE
100 e = u.i[MP_ENDIANNESS_LITTLE];
101 #endif
102#define MP_FLOAT_SIGN_SHIFT_I32 ((MP_FLOAT_FRAC_BITS + MP_FLOAT_EXP_BITS) % 32)
103#define MP_FLOAT_EXP_SHIFT_I32 (MP_FLOAT_FRAC_BITS % 32)
104
105 if (e & (1U << MP_FLOAT_SIGN_SHIFT_I32)) {
106 #if MICROPY_FLOAT_IMPL == MICROPY_FLOAT_IMPL_DOUBLE
107 e |= u.i[MP_ENDIANNESS_BIG] != 0;
108 #endif
109 if ((e & ~(1 << MP_FLOAT_SIGN_SHIFT_I32)) == 0) {
110 // handle case of -0 (when sign is set but rest of bits are zero)
111 e = 0;
112 } else {
113 e += ((1 << MP_FLOAT_EXP_BITS) - 1) << MP_FLOAT_EXP_SHIFT_I32;
114 }
115 } else {
116 e &= ~((1 << MP_FLOAT_EXP_SHIFT_I32) - 1);
117 }
118 // 8 * sizeof(uintptr_t) counts the number of bits for a small int
119 // TODO provide a way to configure this properly
120 if (e <= ((8 * sizeof(uintptr_t) + MP_FLOAT_EXP_BIAS - 3) << MP_FLOAT_EXP_SHIFT_I32)) {
121 return MP_FP_CLASS_FIT_SMALLINT;
122 }
123 #if MICROPY_LONGINT_IMPL == MICROPY_LONGINT_IMPL_LONGLONG
124 if (e <= (((sizeof(long long) * MP_BITS_PER_BYTE) + MP_FLOAT_EXP_BIAS - 2) << MP_FLOAT_EXP_SHIFT_I32)) {
125 return MP_FP_CLASS_FIT_LONGINT;
126 }
127 #endif
128 #if MICROPY_LONGINT_IMPL == MICROPY_LONGINT_IMPL_MPZ
129 return MP_FP_CLASS_FIT_LONGINT;
130 #else
131 return MP_FP_CLASS_OVERFLOW;
132 #endif
133}
134#undef MP_FLOAT_SIGN_SHIFT_I32
135#undef MP_FLOAT_EXP_SHIFT_I32
136
137mp_obj_t mp_obj_new_int_from_float(mp_float_t val) {
138 mp_float_union_t u = {val};
139 // IEEE-754: if biased exponent is all 1 bits...
140 if (u.p.exp == ((1 << MP_FLOAT_EXP_BITS) - 1)) {
141 // ...then number is Inf (positive or negative) if fraction is 0, else NaN.
142 if (u.p.frc == 0) {
143 mp_raise_msg(&mp_type_OverflowError, MP_ERROR_TEXT("can't convert inf to int"));
144 } else {
145 mp_raise_ValueError(MP_ERROR_TEXT("can't convert NaN to int"));
146 }
147 } else {
148 mp_fp_as_int_class_t icl = mp_classify_fp_as_int(val);
149 if (icl == MP_FP_CLASS_FIT_SMALLINT) {
150 return MP_OBJ_NEW_SMALL_INT((mp_int_t)val);
151 #if MICROPY_LONGINT_IMPL == MICROPY_LONGINT_IMPL_MPZ
152 } else {
153 mp_obj_int_t *o = mp_obj_int_new_mpz();
154 mpz_set_from_float(&o->mpz, val);
155 return MP_OBJ_FROM_PTR(o);
156 }
157 #else
158 #if MICROPY_LONGINT_IMPL == MICROPY_LONGINT_IMPL_LONGLONG
159 } else if (icl == MP_FP_CLASS_FIT_LONGINT) {
160 return mp_obj_new_int_from_ll((long long)val);
161 #endif
162 } else {
163 mp_raise_ValueError(MP_ERROR_TEXT("float too big"));
164 }
165 #endif
166 }
167}
168
169#endif
170
171#if MICROPY_LONGINT_IMPL == MICROPY_LONGINT_IMPL_LONGLONG
172typedef mp_longint_impl_t fmt_int_t;
173typedef unsigned long long fmt_uint_t;
174#else
175typedef mp_int_t fmt_int_t;
176typedef mp_uint_t fmt_uint_t;
177#endif
178
179void mp_obj_int_print(const mp_print_t *print, mp_obj_t self_in, mp_print_kind_t kind) {
180 (void)kind;
181 // The size of this buffer is rather arbitrary. If it's not large
182 // enough, a dynamic one will be allocated.
183 char stack_buf[sizeof(fmt_int_t) * 4];
184 char *buf = stack_buf;
185 size_t buf_size = sizeof(stack_buf);
186 size_t fmt_size;
187
188 char *str = mp_obj_int_formatted(&buf, &buf_size, &fmt_size, self_in, 10, NULL, '\0', '\0');
189 mp_print_str(print, str);
190
191 if (buf != stack_buf) {
192 m_del(char, buf, buf_size);
193 }
194}
195
196STATIC const uint8_t log_base2_floor[] = {
197 0, 1, 1, 2,
198 2, 2, 2, 3,
199 3, 3, 3, 3,
200 3, 3, 3, 4,
201 /* if needed, these are the values for higher bases
202 4, 4, 4, 4,
203 4, 4, 4, 4,
204 4, 4, 4, 4,
205 4, 4, 4, 5
206 */
207};
208
209size_t mp_int_format_size(size_t num_bits, int base, const char *prefix, char comma) {
210 assert(2 <= base && base <= 16);
211 size_t num_digits = num_bits / log_base2_floor[base - 1] + 1;
212 size_t num_commas = comma ? num_digits / 3 : 0;
213 size_t prefix_len = prefix ? strlen(prefix) : 0;
214 return num_digits + num_commas + prefix_len + 2; // +1 for sign, +1 for null byte
215}
216
217// This routine expects you to pass in a buffer and size (in *buf and *buf_size).
218// If, for some reason, this buffer is too small, then it will allocate a
219// buffer and return the allocated buffer and size in *buf and *buf_size. It
220// is the callers responsibility to free this allocated buffer.
221//
222// The resulting formatted string will be returned from this function and the
223// formatted size will be in *fmt_size.
224char *mp_obj_int_formatted(char **buf, size_t *buf_size, size_t *fmt_size, mp_const_obj_t self_in,
225 int base, const char *prefix, char base_char, char comma) {
226 fmt_int_t num;
227 #if MICROPY_LONGINT_IMPL == MICROPY_LONGINT_IMPL_NONE
228 // Only have small ints; get the integer value to format.
229 num = MP_OBJ_SMALL_INT_VALUE(self_in);
230 #else
231 if (mp_obj_is_small_int(self_in)) {
232 // A small int; get the integer value to format.
233 num = MP_OBJ_SMALL_INT_VALUE(self_in);
234 } else {
235 assert(mp_obj_is_type(self_in, &mp_type_int));
236 // Not a small int.
237 #if MICROPY_LONGINT_IMPL == MICROPY_LONGINT_IMPL_LONGLONG
238 const mp_obj_int_t *self = self_in;
239 // Get the value to format; mp_obj_get_int truncates to mp_int_t.
240 num = self->val;
241 #else
242 // Delegate to the implementation for the long int.
243 return mp_obj_int_formatted_impl(buf, buf_size, fmt_size, self_in, base, prefix, base_char, comma);
244 #endif
245 }
246 #endif
247
248 char sign = '\0';
249 if (num < 0) {
250 num = -num;
251 sign = '-';
252 }
253
254 size_t needed_size = mp_int_format_size(sizeof(fmt_int_t) * 8, base, prefix, comma);
255 if (needed_size > *buf_size) {
256 *buf = m_new(char, needed_size);
257 *buf_size = needed_size;
258 }
259 char *str = *buf;
260
261 char *b = str + needed_size;
262 *(--b) = '\0';
263 char *last_comma = b;
264
265 if (num == 0) {
266 *(--b) = '0';
267 } else {
268 do {
269 // The cast to fmt_uint_t is because num is positive and we want unsigned arithmetic
270 int c = (fmt_uint_t)num % base;
271 num = (fmt_uint_t)num / base;
272 if (c >= 10) {
273 c += base_char - 10;
274 } else {
275 c += '0';
276 }
277 *(--b) = c;
278 if (comma && num != 0 && b > str && (last_comma - b) == 3) {
279 *(--b) = comma;
280 last_comma = b;
281 }
282 }
283 while (b > str && num != 0);
284 }
285 if (prefix) {
286 size_t prefix_len = strlen(prefix);
287 char *p = b - prefix_len;
288 if (p > str) {
289 b = p;
290 while (*prefix) {
291 *p++ = *prefix++;
292 }
293 }
294 }
295 if (sign && b > str) {
296 *(--b) = sign;
297 }
298 *fmt_size = *buf + needed_size - b - 1;
299
300 return b;
301}
302
303#if MICROPY_LONGINT_IMPL == MICROPY_LONGINT_IMPL_NONE
304
305int mp_obj_int_sign(mp_obj_t self_in) {
306 mp_int_t val = mp_obj_get_int(self_in);
307 if (val < 0) {
308 return -1;
309 } else if (val > 0) {
310 return 1;
311 } else {
312 return 0;
313 }
314}
315
316// This is called for operations on SMALL_INT that are not handled by mp_unary_op
317mp_obj_t mp_obj_int_unary_op(mp_unary_op_t op, mp_obj_t o_in) {
318 return MP_OBJ_NULL; // op not supported
319}
320
321// This is called for operations on SMALL_INT that are not handled by mp_binary_op
322mp_obj_t mp_obj_int_binary_op(mp_binary_op_t op, mp_obj_t lhs_in, mp_obj_t rhs_in) {
323 return mp_obj_int_binary_op_extra_cases(op, lhs_in, rhs_in);
324}
325
326// This is called only with strings whose value doesn't fit in SMALL_INT
327mp_obj_t mp_obj_new_int_from_str_len(const char **str, size_t len, bool neg, unsigned int base) {
328 mp_raise_msg(&mp_type_OverflowError, MP_ERROR_TEXT("long int not supported in this build"));
329 return mp_const_none;
330}
331
332// This is called when an integer larger than a SMALL_INT is needed (although val might still fit in a SMALL_INT)
333mp_obj_t mp_obj_new_int_from_ll(long long val) {
334 mp_raise_msg(&mp_type_OverflowError, MP_ERROR_TEXT("small int overflow"));
335 return mp_const_none;
336}
337
338// This is called when an integer larger than a SMALL_INT is needed (although val might still fit in a SMALL_INT)
339mp_obj_t mp_obj_new_int_from_ull(unsigned long long val) {
340 mp_raise_msg(&mp_type_OverflowError, MP_ERROR_TEXT("small int overflow"));
341 return mp_const_none;
342}
343
344mp_obj_t mp_obj_new_int_from_uint(mp_uint_t value) {
345 // SMALL_INT accepts only signed numbers, so make sure the input
346 // value fits completely in the small-int positive range.
347 if ((value & ~MP_SMALL_INT_POSITIVE_MASK) == 0) {
348 return MP_OBJ_NEW_SMALL_INT(value);
349 }
350 mp_raise_msg(&mp_type_OverflowError, MP_ERROR_TEXT("small int overflow"));
351 return mp_const_none;
352}
353
354mp_obj_t mp_obj_new_int(mp_int_t value) {
355 if (MP_SMALL_INT_FITS(value)) {
356 return MP_OBJ_NEW_SMALL_INT(value);
357 }
358 mp_raise_msg(&mp_type_OverflowError, MP_ERROR_TEXT("small int overflow"));
359 return mp_const_none;
360}
361
362mp_int_t mp_obj_int_get_truncated(mp_const_obj_t self_in) {
363 return MP_OBJ_SMALL_INT_VALUE(self_in);
364}
365
366mp_int_t mp_obj_int_get_checked(mp_const_obj_t self_in) {
367 return MP_OBJ_SMALL_INT_VALUE(self_in);
368}
369
370#endif // MICROPY_LONGINT_IMPL == MICROPY_LONGINT_IMPL_NONE
371
372// This dispatcher function is expected to be independent of the implementation of long int
373// It handles the extra cases for integer-like arithmetic
374mp_obj_t mp_obj_int_binary_op_extra_cases(mp_binary_op_t op, mp_obj_t lhs_in, mp_obj_t rhs_in) {
375 if (rhs_in == mp_const_false) {
376 // false acts as 0
377 return mp_binary_op(op, lhs_in, MP_OBJ_NEW_SMALL_INT(0));
378 } else if (rhs_in == mp_const_true) {
379 // true acts as 0
380 return mp_binary_op(op, lhs_in, MP_OBJ_NEW_SMALL_INT(1));
381 } else if (op == MP_BINARY_OP_MULTIPLY) {
382 if (mp_obj_is_str_or_bytes(rhs_in) || mp_obj_is_type(rhs_in, &mp_type_tuple) || mp_obj_is_type(rhs_in, &mp_type_list)) {
383 // multiply is commutative for these types, so delegate to them
384 return mp_binary_op(op, rhs_in, lhs_in);
385 }
386 }
387 return MP_OBJ_NULL; // op not supported
388}
389
390// this is a classmethod
391STATIC mp_obj_t int_from_bytes(size_t n_args, const mp_obj_t *args) {
392 // TODO: Support signed param (assumes signed=False at the moment)
393 (void)n_args;
394
395 // get the buffer info
396 mp_buffer_info_t bufinfo;
397 mp_get_buffer_raise(args[1], &bufinfo, MP_BUFFER_READ);
398
399 const byte *buf = (const byte *)bufinfo.buf;
400 int delta = 1;
401 if (args[2] == MP_OBJ_NEW_QSTR(MP_QSTR_little)) {
402 buf += bufinfo.len - 1;
403 delta = -1;
404 }
405
406 mp_uint_t value = 0;
407 size_t len = bufinfo.len;
408 for (; len--; buf += delta) {
409 #if MICROPY_LONGINT_IMPL != MICROPY_LONGINT_IMPL_NONE
410 if (value > (MP_SMALL_INT_MAX >> 8)) {
411 // Result will overflow a small-int so construct a big-int
412 return mp_obj_int_from_bytes_impl(args[2] != MP_OBJ_NEW_QSTR(MP_QSTR_little), bufinfo.len, bufinfo.buf);
413 }
414 #endif
415 value = (value << 8) | *buf;
416 }
417 return mp_obj_new_int_from_uint(value);
418}
419
420STATIC MP_DEFINE_CONST_FUN_OBJ_VAR_BETWEEN(int_from_bytes_fun_obj, 3, 4, int_from_bytes);
421STATIC MP_DEFINE_CONST_CLASSMETHOD_OBJ(int_from_bytes_obj, MP_ROM_PTR(&int_from_bytes_fun_obj));
422
423STATIC mp_obj_t int_to_bytes(size_t n_args, const mp_obj_t *args) {
424 // TODO: Support signed param (assumes signed=False)
425 (void)n_args;
426
427 mp_int_t len = mp_obj_get_int(args[1]);
428 if (len < 0) {
429 mp_raise_ValueError(NULL);
430 }
431 bool big_endian = args[2] != MP_OBJ_NEW_QSTR(MP_QSTR_little);
432
433 vstr_t vstr;
434 vstr_init_len(&vstr, len);
435 byte *data = (byte *)vstr.buf;
436 memset(data, 0, len);
437
438 #if MICROPY_LONGINT_IMPL != MICROPY_LONGINT_IMPL_NONE
439 if (!mp_obj_is_small_int(args[0])) {
440 mp_obj_int_to_bytes_impl(args[0], big_endian, len, data);
441 } else
442 #endif
443 {
444 mp_int_t val = MP_OBJ_SMALL_INT_VALUE(args[0]);
445 size_t l = MIN((size_t)len, sizeof(val));
446 mp_binary_set_int(l, big_endian, data + (big_endian ? (len - l) : 0), val);
447 }
448
449 return mp_obj_new_str_from_vstr(&mp_type_bytes, &vstr);
450}
451STATIC MP_DEFINE_CONST_FUN_OBJ_VAR_BETWEEN(int_to_bytes_obj, 3, 4, int_to_bytes);
452
453STATIC const mp_rom_map_elem_t int_locals_dict_table[] = {
454 { MP_ROM_QSTR(MP_QSTR_from_bytes), MP_ROM_PTR(&int_from_bytes_obj) },
455 { MP_ROM_QSTR(MP_QSTR_to_bytes), MP_ROM_PTR(&int_to_bytes_obj) },
456};
457
458STATIC MP_DEFINE_CONST_DICT(int_locals_dict, int_locals_dict_table);
459
460const mp_obj_type_t mp_type_int = {
461 { &mp_type_type },
462 .name = MP_QSTR_int,
463 .print = mp_obj_int_print,
464 .make_new = mp_obj_int_make_new,
465 .unary_op = mp_obj_int_unary_op,
466 .binary_op = mp_obj_int_binary_op,
467 .locals_dict = (mp_obj_dict_t *)&int_locals_dict,
468};
469