1 ///////////////////////////////////////////////////////////////////////////////
3 /// \file simple_coder.c
4 /// \brief Wrapper for simple filters
6 /// Simple filters don't change the size of the data i.e. number of bytes
7 /// in equals the number of bytes out.
9 // Copyright (C) 2007 Lasse Collin
11 // This library is free software; you can redistribute it and/or
12 // modify it under the terms of the GNU Lesser General Public
13 // License as published by the Free Software Foundation; either
14 // version 2.1 of the License, or (at your option) any later version.
16 // This library is distributed in the hope that it will be useful,
17 // but WITHOUT ANY WARRANTY; without even the implied warranty of
18 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 // Lesser General Public License for more details.
21 ///////////////////////////////////////////////////////////////////////////////
23 #include "simple_private.h"
26 /// Copied or encodes/decodes more data to out[].
28 copy_or_code(lzma_coder *coder, lzma_allocator *allocator,
29 const uint8_t *restrict in, size_t *restrict in_pos,
30 size_t in_size, uint8_t *restrict out,
31 size_t *restrict out_pos, size_t out_size, lzma_action action)
33 assert(!coder->end_was_reached);
35 if (coder->next.code == NULL) {
36 lzma_bufcpy(in, in_pos, in_size, out, out_pos, out_size);
38 // Check if end of stream was reached.
39 if (coder->is_encoder && action == LZMA_FINISH
40 && *in_pos == in_size)
41 coder->end_was_reached = true;
44 // Call the next coder in the chain to provide us some data.
45 // We don't care about uncompressed_size here, because
46 // the next filter in the chain will do it for us (since
47 // we don't change the size of the data).
48 const lzma_ret ret = coder->next.code(
49 coder->next.coder, allocator,
51 out, out_pos, out_size, action);
53 if (ret == LZMA_STREAM_END) {
54 assert(!coder->is_encoder
55 || action == LZMA_FINISH);
56 coder->end_was_reached = true;
58 } else if (ret != LZMA_OK) {
68 call_filter(lzma_coder *coder, uint8_t *buffer, size_t size)
70 const size_t filtered = coder->filter(coder->simple,
71 coder->now_pos, coder->is_encoder,
73 coder->now_pos += filtered;
79 simple_code(lzma_coder *coder, lzma_allocator *allocator,
80 const uint8_t *restrict in, size_t *restrict in_pos,
81 size_t in_size, uint8_t *restrict out,
82 size_t *restrict out_pos, size_t out_size, lzma_action action)
84 // TODO: Add partial support for LZMA_SYNC_FLUSH. We can support it
85 // in cases when the filter is able to filter everything. With most
86 // simple filters it can be done at offset that is a multiple of 2,
87 // 4, or 16. With x86 filter, it needs good luck, and thus cannot
88 // be made to work predictably.
89 if (action == LZMA_SYNC_FLUSH)
90 return LZMA_OPTIONS_ERROR;
92 // Flush already filtered data from coder->buffer[] to out[].
93 if (coder->pos < coder->filtered) {
94 lzma_bufcpy(coder->buffer, &coder->pos, coder->filtered,
95 out, out_pos, out_size);
97 // If we couldn't flush all the filtered data, return to
98 // application immediatelly.
99 if (coder->pos < coder->filtered)
102 if (coder->end_was_reached) {
103 assert(coder->filtered == coder->size);
104 return LZMA_STREAM_END;
108 // If we get here, there is no filtered data left in the buffer.
111 assert(!coder->end_was_reached);
113 // If there is more output space left than there is unfiltered data
114 // in coder->buffer[], flush coder->buffer[] to out[], and copy/code
115 // more data to out[] hopefully filling it completely. Then filter
116 // the data in out[]. This step is where most of the data gets
117 // filtered if the buffer sizes used by the application are reasonable.
118 const size_t out_avail = out_size - *out_pos;
119 const size_t buf_avail = coder->size - coder->pos;
120 if (out_avail > buf_avail) {
121 // Store the old position so that we know from which byte
122 // to start filtering.
123 const size_t out_start = *out_pos;
125 // Flush data from coder->buffer[] to out[], but don't reset
126 // coder->pos and coder->size yet. This way the coder can be
127 // restarted if the next filter in the chain returns e.g.
129 memcpy(out + *out_pos, coder->buffer + coder->pos, buf_avail);
130 *out_pos += buf_avail;
132 // Copy/Encode/Decode more data to out[].
134 const lzma_ret ret = copy_or_code(coder, allocator,
136 out, out_pos, out_size, action);
137 assert(ret != LZMA_STREAM_END);
143 const size_t size = *out_pos - out_start;
144 const size_t filtered = call_filter(
145 coder, out + out_start, size);
147 const size_t unfiltered = size - filtered;
148 assert(unfiltered <= coder->allocated / 2);
150 // Now we can update coder->pos and coder->size, because
151 // the next coder in the chain (if any) was successful.
153 coder->size = unfiltered;
155 if (coder->end_was_reached) {
156 // The last byte has been copied to out[] already.
157 // They are left as is.
160 } else if (unfiltered > 0) {
161 // There is unfiltered data left in out[]. Copy it to
162 // coder->buffer[] and rewind *out_pos appropriately.
163 *out_pos -= unfiltered;
164 memcpy(coder->buffer, out + *out_pos, unfiltered);
166 } else if (coder->pos > 0) {
167 memmove(coder->buffer, coder->buffer + coder->pos, buf_avail);
168 coder->size -= coder->pos;
172 assert(coder->pos == 0);
174 // If coder->buffer[] isn't empty, try to fill it by copying/decoding
175 // more data. Then filter coder->buffer[] and copy the successfully
176 // filtered data to out[]. It is probable, that some filtered and
177 // unfiltered data will be left to coder->buffer[].
178 if (coder->size > 0) {
180 const lzma_ret ret = copy_or_code(coder, allocator,
182 coder->buffer, &coder->size,
183 coder->allocated, action);
184 assert(ret != LZMA_STREAM_END);
189 coder->filtered = call_filter(
190 coder, coder->buffer, coder->size);
192 // Everything is considered to be filtered if coder->buffer[]
193 // contains the last bytes of the data.
194 if (coder->end_was_reached)
195 coder->filtered = coder->size;
197 // Flush as much as possible.
198 lzma_bufcpy(coder->buffer, &coder->pos, coder->filtered,
199 out, out_pos, out_size);
202 // Check if we got everything done.
203 if (coder->end_was_reached && coder->pos == coder->size)
204 return LZMA_STREAM_END;
211 simple_coder_end(lzma_coder *coder, lzma_allocator *allocator)
213 lzma_next_end(&coder->next, allocator);
214 lzma_free(coder->simple, allocator);
215 lzma_free(coder, allocator);
221 lzma_simple_coder_init(lzma_next_coder *next, lzma_allocator *allocator,
222 const lzma_filter_info *filters,
223 size_t (*filter)(lzma_simple *simple, uint32_t now_pos,
224 bool is_encoder, uint8_t *buffer, size_t size),
225 size_t simple_size, size_t unfiltered_max, bool is_encoder)
227 // Allocate memory for the lzma_coder structure if needed.
228 if (next->coder == NULL) {
229 // Here we allocate space also for the temporary buffer. We
230 // need twice the size of unfiltered_max, because then it
231 // is always possible to filter at least unfiltered_max bytes
232 // more data in coder->buffer[] if it can be filled completely.
233 next->coder = lzma_alloc(sizeof(lzma_coder)
234 + 2 * unfiltered_max, allocator);
235 if (next->coder == NULL)
236 return LZMA_MEM_ERROR;
238 next->code = &simple_code;
239 next->end = &simple_coder_end;
241 next->coder->next = LZMA_NEXT_CODER_INIT;
242 next->coder->filter = filter;
243 next->coder->allocated = 2 * unfiltered_max;
245 // Allocate memory for filter-specific data structure.
246 if (simple_size > 0) {
247 next->coder->simple = lzma_alloc(
248 simple_size, allocator);
249 if (next->coder->simple == NULL)
250 return LZMA_MEM_ERROR;
252 next->coder->simple = NULL;
256 if (filters[0].options != NULL) {
257 const lzma_options_bcj *simple = filters[0].options;
258 next->coder->now_pos = simple->start_offset;
260 next->coder->now_pos = 0;
264 next->coder->is_encoder = is_encoder;
265 next->coder->end_was_reached = false;
266 next->coder->pos = 0;
267 next->coder->filtered = 0;
268 next->coder->size = 0;
270 return lzma_next_filter_init(
271 &next->coder->next, allocator, filters + 1);