blob: afbf0ff9b6644e9679ad900163c2e4fab8d6c4c0 [file] [log] [blame]
Damien George04b91472014-05-03 23:27:38 +01001/*
2 * This file is part of the Micro Python 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
Damien429d7192013-10-04 19:53:11 +010027#include <assert.h>
28#include <string.h>
Damien Georgeea0461d2015-02-10 11:02:28 +000029#include <stdio.h>
Damien429d7192013-10-04 19:53:11 +010030
Damien Georgeb4b10fd2015-01-01 23:30:53 +000031#include "py/mpstate.h"
Damien George51dfcb42015-01-01 20:27:54 +000032#include "py/qstr.h"
33#include "py/gc.h"
Damien429d7192013-10-04 19:53:11 +010034
Damien Georgeeb7bfcb2014-01-04 15:57:35 +000035// NOTE: we are using linear arrays to store and search for qstr's (unique strings, interned strings)
36// ultimately we will replace this with a static hash table of some kind
37// also probably need to include the length in the string data, to allow null bytes in the string
38
39#if 0 // print debugging info
Paul Sokolovsky44739e22014-02-16 18:11:42 +020040#define DEBUG_printf DEBUG_printf
Damien Georgeeb7bfcb2014-01-04 15:57:35 +000041#else // don't print debugging info
Damien George1dc76af2014-02-26 16:57:08 +000042#define DEBUG_printf(...) (void)0
Damien Georgeeb7bfcb2014-01-04 15:57:35 +000043#endif
44
Damien George55baff42014-01-21 21:40:13 +000045// A qstr is an index into the qstr pool.
Damien Georgec3bd9412015-07-20 11:03:13 +000046// The data for a qstr contains (hash, length, data):
47// - hash (configurable number of bytes)
48// - length (configurable number of bytes)
49// - data ("length" number of bytes)
50// - \0 terminated (so they can be printed using printf)
Damien George55baff42014-01-21 21:40:13 +000051
Damien Georgec3bd9412015-07-20 11:03:13 +000052#if MICROPY_QSTR_BYTES_IN_HASH == 1
53 #define Q_HASH_MASK (0xff)
54 #define Q_GET_HASH(q) ((mp_uint_t)(q)[0])
55 #define Q_SET_HASH(q, hash) do { (q)[0] = (hash); } while (0)
56#elif MICROPY_QSTR_BYTES_IN_HASH == 2
57 #define Q_HASH_MASK (0xffff)
58 #define Q_GET_HASH(q) ((mp_uint_t)(q)[0] | ((mp_uint_t)(q)[1] << 8))
59 #define Q_SET_HASH(q, hash) do { (q)[0] = (hash); (q)[1] = (hash) >> 8; } while (0)
60#else
61 #error unimplemented qstr hash decoding
62#endif
63#define Q_GET_ALLOC(q) (MICROPY_QSTR_BYTES_IN_HASH + MICROPY_QSTR_BYTES_IN_LEN + Q_GET_LENGTH(q) + 1)
64#define Q_GET_DATA(q) ((q) + MICROPY_QSTR_BYTES_IN_HASH + MICROPY_QSTR_BYTES_IN_LEN)
Damien George95836f82015-01-11 22:27:30 +000065#if MICROPY_QSTR_BYTES_IN_LEN == 1
Damien Georgec3bd9412015-07-20 11:03:13 +000066 #define Q_GET_LENGTH(q) ((q)[MICROPY_QSTR_BYTES_IN_HASH])
67 #define Q_SET_LENGTH(q, len) do { (q)[MICROPY_QSTR_BYTES_IN_HASH] = (len); } while (0)
Damien George95836f82015-01-11 22:27:30 +000068#elif MICROPY_QSTR_BYTES_IN_LEN == 2
Damien Georgec3bd9412015-07-20 11:03:13 +000069 #define Q_GET_LENGTH(q) ((q)[MICROPY_QSTR_BYTES_IN_HASH] | ((q)[MICROPY_QSTR_BYTES_IN_HASH + 1] << 8))
70 #define Q_SET_LENGTH(q, len) do { (q)[MICROPY_QSTR_BYTES_IN_HASH] = (len); (q)[MICROPY_QSTR_BYTES_IN_HASH + 1] = (len) >> 8; } while (0)
Damien George95836f82015-01-11 22:27:30 +000071#else
72 #error unimplemented qstr length decoding
73#endif
Damien George55baff42014-01-21 21:40:13 +000074
Damien George6e628c42014-03-25 15:27:15 +000075// this must match the equivalent function in makeqstrdata.py
Damien George39dc1452014-10-03 19:52:22 +010076mp_uint_t qstr_compute_hash(const byte *data, mp_uint_t len) {
Damien George6e628c42014-03-25 15:27:15 +000077 // djb2 algorithm; see http://www.cse.yorku.ca/~oz/hash.html
Damien George40f3c022014-07-03 13:25:24 +010078 mp_uint_t hash = 5381;
Damien George55baff42014-01-21 21:40:13 +000079 for (const byte *top = data + len; data < top; data++) {
Damien George6e628c42014-03-25 15:27:15 +000080 hash = ((hash << 5) + hash) ^ (*data); // hash * 33 ^ data
Damien George55baff42014-01-21 21:40:13 +000081 }
Damien Georgec3bd9412015-07-20 11:03:13 +000082 hash &= Q_HASH_MASK;
Paul Sokolovsky59e269c2014-04-14 01:43:01 +030083 // Make sure that valid hash is never zero, zero means "hash not computed"
84 if (hash == 0) {
85 hash++;
86 }
87 return hash;
Damien George55baff42014-01-21 21:40:13 +000088}
89
Damien George73c98d82014-06-11 19:18:03 +010090STATIC const qstr_pool_t const_pool = {
Damien Georgeeb7bfcb2014-01-04 15:57:35 +000091 NULL, // no previous pool
92 0, // no previous pool
93 10, // set so that the first dynamically allocated pool is twice this size; must be <= the len (just below)
94 MP_QSTR_number_of, // corresponds to number of strings in array just below
95 {
Damien George6942f802015-01-11 17:52:45 +000096#define QDEF(id, str) str,
Damien Georged553be52014-04-17 18:03:27 +010097#include "genhdr/qstrdefs.generated.h"
Damien George6942f802015-01-11 17:52:45 +000098#undef QDEF
Damien Georgeeb7bfcb2014-01-04 15:57:35 +000099 },
100};
101
Damien8b3a7c22013-10-23 20:20:17 +0100102void qstr_init(void) {
Damien Georgeb4b10fd2015-01-01 23:30:53 +0000103 MP_STATE_VM(last_pool) = (qstr_pool_t*)&const_pool; // we won't modify the const_pool since it has no allocated room left
Damien Georgeade9a052015-06-13 21:53:22 +0100104 MP_STATE_VM(qstr_last_chunk) = NULL;
Damien429d7192013-10-04 19:53:11 +0100105}
106
Paul Sokolovsky520e2f52014-02-12 18:31:30 +0200107STATIC const byte *find_qstr(qstr q) {
Damien George55baff42014-01-21 21:40:13 +0000108 // search pool for this qstr
Damien Georgeb4b10fd2015-01-01 23:30:53 +0000109 for (qstr_pool_t *pool = MP_STATE_VM(last_pool); pool != NULL; pool = pool->prev) {
Damien George55baff42014-01-21 21:40:13 +0000110 if (q >= pool->total_prev_len) {
111 return pool->qstrs[q - pool->total_prev_len];
112 }
113 }
114
115 // not found
116 return 0;
117}
118
Paul Sokolovsky520e2f52014-02-12 18:31:30 +0200119STATIC qstr qstr_add(const byte *q_ptr) {
Damien George55baff42014-01-21 21:40:13 +0000120 DEBUG_printf("QSTR: add hash=%d len=%d data=%.*s\n", Q_GET_HASH(q_ptr), Q_GET_LENGTH(q_ptr), Q_GET_LENGTH(q_ptr), Q_GET_DATA(q_ptr));
Damien Georgeeb7bfcb2014-01-04 15:57:35 +0000121
122 // make sure we have room in the pool for a new qstr
Damien Georgeb4b10fd2015-01-01 23:30:53 +0000123 if (MP_STATE_VM(last_pool)->len >= MP_STATE_VM(last_pool)->alloc) {
124 qstr_pool_t *pool = m_new_obj_var(qstr_pool_t, const char*, MP_STATE_VM(last_pool)->alloc * 2);
125 pool->prev = MP_STATE_VM(last_pool);
126 pool->total_prev_len = MP_STATE_VM(last_pool)->total_prev_len + MP_STATE_VM(last_pool)->len;
127 pool->alloc = MP_STATE_VM(last_pool)->alloc * 2;
Damien Georgeeb7bfcb2014-01-04 15:57:35 +0000128 pool->len = 0;
Damien Georgeb4b10fd2015-01-01 23:30:53 +0000129 MP_STATE_VM(last_pool) = pool;
130 DEBUG_printf("QSTR: allocate new pool of size %d\n", MP_STATE_VM(last_pool)->alloc);
Damien429d7192013-10-04 19:53:11 +0100131 }
Damien Georgeeb7bfcb2014-01-04 15:57:35 +0000132
133 // add the new qstr
Damien Georgeb4b10fd2015-01-01 23:30:53 +0000134 MP_STATE_VM(last_pool)->qstrs[MP_STATE_VM(last_pool)->len++] = q_ptr;
Damien Georgeeb7bfcb2014-01-04 15:57:35 +0000135
136 // return id for the newly-added qstr
Damien Georgeb4b10fd2015-01-01 23:30:53 +0000137 return MP_STATE_VM(last_pool)->total_prev_len + MP_STATE_VM(last_pool)->len - 1;
Damien429d7192013-10-04 19:53:11 +0100138}
139
Damien George39dc1452014-10-03 19:52:22 +0100140qstr qstr_find_strn(const char *str, mp_uint_t str_len) {
Damien George55baff42014-01-21 21:40:13 +0000141 // work out hash of str
Damien George40f3c022014-07-03 13:25:24 +0100142 mp_uint_t str_hash = qstr_compute_hash((const byte*)str, str_len);
Damien George55baff42014-01-21 21:40:13 +0000143
144 // search pools for the data
Damien Georgeb4b10fd2015-01-01 23:30:53 +0000145 for (qstr_pool_t *pool = MP_STATE_VM(last_pool); pool != NULL; pool = pool->prev) {
Damien George55baff42014-01-21 21:40:13 +0000146 for (const byte **q = pool->qstrs, **q_top = pool->qstrs + pool->len; q < q_top; q++) {
Damien George1e708fe2014-01-23 18:27:51 +0000147 if (Q_GET_HASH(*q) == str_hash && Q_GET_LENGTH(*q) == str_len && memcmp(Q_GET_DATA(*q), str, str_len) == 0) {
Damien George55baff42014-01-21 21:40:13 +0000148 return pool->total_prev_len + (q - pool->qstrs);
Damien Georgeeb7bfcb2014-01-04 15:57:35 +0000149 }
Damien429d7192013-10-04 19:53:11 +0100150 }
151 }
Damien George55baff42014-01-21 21:40:13 +0000152
153 // not found; return null qstr
154 return 0;
Damien429d7192013-10-04 19:53:11 +0100155}
156
Damien George55baff42014-01-21 21:40:13 +0000157qstr qstr_from_str(const char *str) {
158 return qstr_from_strn(str, strlen(str));
Damien429d7192013-10-04 19:53:11 +0100159}
160
Damien George39dc1452014-10-03 19:52:22 +0100161qstr qstr_from_strn(const char *str, mp_uint_t len) {
Damien George4c81ba82015-01-13 16:21:23 +0000162 assert(len < (1 << (8 * MICROPY_QSTR_BYTES_IN_LEN)));
Damien George2617eeb2014-05-25 22:27:57 +0100163 qstr q = qstr_find_strn(str, len);
Damien George55baff42014-01-21 21:40:13 +0000164 if (q == 0) {
Damien Georgeade9a052015-06-13 21:53:22 +0100165 // qstr does not exist in interned pool so need to add it
166
167 // compute number of bytes needed to intern this string
Damien Georgec3bd9412015-07-20 11:03:13 +0000168 mp_uint_t n_bytes = MICROPY_QSTR_BYTES_IN_HASH + MICROPY_QSTR_BYTES_IN_LEN + len + 1;
Damien Georgeade9a052015-06-13 21:53:22 +0100169
170 if (MP_STATE_VM(qstr_last_chunk) != NULL && MP_STATE_VM(qstr_last_used) + n_bytes > MP_STATE_VM(qstr_last_alloc)) {
171 // not enough room at end of previously interned string so try to grow
172 byte *new_p = m_renew_maybe(byte, MP_STATE_VM(qstr_last_chunk), MP_STATE_VM(qstr_last_alloc), MP_STATE_VM(qstr_last_alloc) + n_bytes, false);
173 if (new_p == NULL) {
174 // could not grow existing memory; shrink it to fit previous
175 (void)m_renew(byte, MP_STATE_VM(qstr_last_chunk), MP_STATE_VM(qstr_last_alloc), MP_STATE_VM(qstr_last_used));
176 MP_STATE_VM(qstr_last_chunk) = NULL;
177 } else {
178 // could grow existing memory
179 MP_STATE_VM(qstr_last_alloc) += n_bytes;
180 }
181 }
182
183 if (MP_STATE_VM(qstr_last_chunk) == NULL) {
184 // no existing memory for the interned string so allocate a new chunk
185 mp_uint_t al = n_bytes;
186 if (al < MICROPY_ALLOC_QSTR_CHUNK_INIT) {
187 al = MICROPY_ALLOC_QSTR_CHUNK_INIT;
188 }
189 MP_STATE_VM(qstr_last_chunk) = m_new_maybe(byte, al);
190 if (MP_STATE_VM(qstr_last_chunk) == NULL) {
191 // failed to allocate a large chunk so try with exact size
192 MP_STATE_VM(qstr_last_chunk) = m_new(byte, n_bytes);
193 al = n_bytes;
194 }
195 MP_STATE_VM(qstr_last_alloc) = al;
196 MP_STATE_VM(qstr_last_used) = 0;
197 }
198
199 // allocate memory from the chunk for this new interned string's data
200 byte *q_ptr = MP_STATE_VM(qstr_last_chunk) + MP_STATE_VM(qstr_last_used);
201 MP_STATE_VM(qstr_last_used) += n_bytes;
202
203 // store the interned strings' data
Damien George40f3c022014-07-03 13:25:24 +0100204 mp_uint_t hash = qstr_compute_hash((const byte*)str, len);
Damien Georgec3bd9412015-07-20 11:03:13 +0000205 Q_SET_HASH(q_ptr, hash);
Damien George95836f82015-01-11 22:27:30 +0000206 Q_SET_LENGTH(q_ptr, len);
Damien Georgec3bd9412015-07-20 11:03:13 +0000207 memcpy(q_ptr + MICROPY_QSTR_BYTES_IN_HASH + MICROPY_QSTR_BYTES_IN_LEN, str, len);
208 q_ptr[MICROPY_QSTR_BYTES_IN_HASH + MICROPY_QSTR_BYTES_IN_LEN + len] = '\0';
Damien George55baff42014-01-21 21:40:13 +0000209 q = qstr_add(q_ptr);
Damien429d7192013-10-04 19:53:11 +0100210 }
Damien George55baff42014-01-21 21:40:13 +0000211 return q;
Damien429d7192013-10-04 19:53:11 +0100212}
213
Damien George39dc1452014-10-03 19:52:22 +0100214byte *qstr_build_start(mp_uint_t len, byte **q_ptr) {
Damien George95836f82015-01-11 22:27:30 +0000215 assert(len < (1 << (8 * MICROPY_QSTR_BYTES_IN_LEN)));
Damien Georgec3bd9412015-07-20 11:03:13 +0000216 *q_ptr = m_new(byte, MICROPY_QSTR_BYTES_IN_HASH + MICROPY_QSTR_BYTES_IN_LEN + len + 1);
Damien George95836f82015-01-11 22:27:30 +0000217 Q_SET_LENGTH(*q_ptr, len);
Damien George55baff42014-01-21 21:40:13 +0000218 return Q_GET_DATA(*q_ptr);
219}
220
221qstr qstr_build_end(byte *q_ptr) {
Damien George2617eeb2014-05-25 22:27:57 +0100222 qstr q = qstr_find_strn((const char*)Q_GET_DATA(q_ptr), Q_GET_LENGTH(q_ptr));
Damien George55baff42014-01-21 21:40:13 +0000223 if (q == 0) {
Damien George40f3c022014-07-03 13:25:24 +0100224 mp_uint_t len = Q_GET_LENGTH(q_ptr);
225 mp_uint_t hash = qstr_compute_hash(Q_GET_DATA(q_ptr), len);
Damien Georgec3bd9412015-07-20 11:03:13 +0000226 Q_SET_HASH(q_ptr, hash);
227 q_ptr[MICROPY_QSTR_BYTES_IN_HASH + MICROPY_QSTR_BYTES_IN_LEN + len] = '\0';
Damien George55baff42014-01-21 21:40:13 +0000228 q = qstr_add(q_ptr);
229 } else {
230 m_del(byte, q_ptr, Q_GET_ALLOC(q_ptr));
231 }
232 return q;
233}
234
Damien George40f3c022014-07-03 13:25:24 +0100235mp_uint_t qstr_hash(qstr q) {
Damien George55baff42014-01-21 21:40:13 +0000236 return Q_GET_HASH(find_qstr(q));
237}
238
Damien George39dc1452014-10-03 19:52:22 +0100239mp_uint_t qstr_len(qstr q) {
Damien George55baff42014-01-21 21:40:13 +0000240 const byte *qd = find_qstr(q);
241 return Q_GET_LENGTH(qd);
242}
243
244// XXX to remove!
245const char *qstr_str(qstr q) {
246 const byte *qd = find_qstr(q);
247 return (const char*)Q_GET_DATA(qd);
248}
249
Damien George39dc1452014-10-03 19:52:22 +0100250const byte *qstr_data(qstr q, mp_uint_t *len) {
Damien George55baff42014-01-21 21:40:13 +0000251 const byte *qd = find_qstr(q);
252 *len = Q_GET_LENGTH(qd);
253 return Q_GET_DATA(qd);
Damien429d7192013-10-04 19:53:11 +0100254}
Damien George4d5b28c2014-01-29 18:56:46 +0000255
Damien George39dc1452014-10-03 19:52:22 +0100256void qstr_pool_info(mp_uint_t *n_pool, mp_uint_t *n_qstr, mp_uint_t *n_str_data_bytes, mp_uint_t *n_total_bytes) {
Damien George4d5b28c2014-01-29 18:56:46 +0000257 *n_pool = 0;
258 *n_qstr = 0;
259 *n_str_data_bytes = 0;
260 *n_total_bytes = 0;
Damien Georgeb4b10fd2015-01-01 23:30:53 +0000261 for (qstr_pool_t *pool = MP_STATE_VM(last_pool); pool != NULL && pool != &const_pool; pool = pool->prev) {
Damien George4d5b28c2014-01-29 18:56:46 +0000262 *n_pool += 1;
263 *n_qstr += pool->len;
264 for (const byte **q = pool->qstrs, **q_top = pool->qstrs + pool->len; q < q_top; q++) {
Damien George0b13f3e2014-10-24 23:12:25 +0100265 #if MICROPY_ENABLE_GC
266 *n_str_data_bytes += gc_nbytes(*q); // this counts actual bytes used in heap
267 #else
Damien George4d5b28c2014-01-29 18:56:46 +0000268 *n_str_data_bytes += Q_GET_ALLOC(*q);
Damien George0b13f3e2014-10-24 23:12:25 +0100269 #endif
Damien George4d5b28c2014-01-29 18:56:46 +0000270 }
Damien George0b13f3e2014-10-24 23:12:25 +0100271 #if MICROPY_ENABLE_GC
272 *n_total_bytes += gc_nbytes(pool); // this counts actual bytes used in heap
273 #else
Damien George4d5b28c2014-01-29 18:56:46 +0000274 *n_total_bytes += sizeof(qstr_pool_t) + sizeof(qstr) * pool->alloc;
Damien George0b13f3e2014-10-24 23:12:25 +0100275 #endif
Damien George4d5b28c2014-01-29 18:56:46 +0000276 }
277 *n_total_bytes += *n_str_data_bytes;
278}
Damien Georgeea0461d2015-02-10 11:02:28 +0000279
280#if MICROPY_PY_MICROPYTHON_MEM_INFO
281void qstr_dump_data(void) {
282 for (qstr_pool_t *pool = MP_STATE_VM(last_pool); pool != NULL && pool != &const_pool; pool = pool->prev) {
283 for (const byte **q = pool->qstrs, **q_top = pool->qstrs + pool->len; q < q_top; q++) {
Damien Georgee72cda92015-04-11 12:15:47 +0100284 mp_printf(&mp_plat_print, "Q(%s)\n", Q_GET_DATA(*q));
Damien Georgeea0461d2015-02-10 11:02:28 +0000285 }
286 }
287}
288#endif