1 ///////////////////////////////////////////////////////////////////////////////
4 /// \brief LZ out window
6 // Authors: Igor Pavlov
9 // This file has been put into the public domain.
10 // You can do whatever you want with this file.
12 ///////////////////////////////////////////////////////////////////////////////
14 #ifndef LZMA_LZ_DECODER_H
15 #define LZMA_LZ_DECODER_H
21 /// Pointer to the dictionary buffer. It can be an allocated buffer
22 /// internal to liblzma, or it can a be a buffer given by the
23 /// application when in single-call mode (not implemented yet).
26 /// Write position in dictionary. The next byte will be written to
30 /// Indicates how full the dictionary is. This is used by
31 /// dict_is_distance_valid() to detect corrupt files that would
32 /// read beyond the beginning of the dictionary.
38 /// Size of the dictionary
41 /// True when dictionary should be reset before decoding more data.
49 const uint8_t *preset_dict;
50 size_t preset_dict_size;
55 /// Data specific to the LZ-based decoder
58 /// Function to decode from in[] to *dict
59 lzma_ret (*code)(void *coder,
60 lzma_dict *restrict dict, const uint8_t *restrict in,
61 size_t *restrict in_pos, size_t in_size);
63 void (*reset)(void *coder, const void *options);
65 /// Set the uncompressed size
66 void (*set_uncompressed)(void *coder, lzma_vli uncompressed_size);
68 /// Free allocated resources
69 void (*end)(void *coder, const lzma_allocator *allocator);
74 #define LZMA_LZ_DECODER_INIT \
79 .set_uncompressed = NULL, \
84 extern lzma_ret lzma_lz_decoder_init(lzma_next_coder *next,
85 const lzma_allocator *allocator,
86 const lzma_filter_info *filters,
87 lzma_ret (*lz_init)(lzma_lz_decoder *lz,
88 const lzma_allocator *allocator, const void *options,
89 lzma_lz_options *lz_options));
91 extern uint64_t lzma_lz_decoder_memusage(size_t dictionary_size);
93 extern void lzma_lz_decoder_uncompressed(
94 void *coder, lzma_vli uncompressed_size);
97 //////////////////////
98 // Inline functions //
99 //////////////////////
101 /// Get a byte from the history buffer.
102 static inline uint8_t
103 dict_get(const lzma_dict *const dict, const uint32_t distance)
105 return dict->buf[dict->pos - distance - 1
106 + (distance < dict->pos ? 0 : dict->size)];
110 /// Test if dictionary is empty.
112 dict_is_empty(const lzma_dict *const dict)
114 return dict->full == 0;
118 /// Validate the match distance
120 dict_is_distance_valid(const lzma_dict *const dict, const size_t distance)
122 return dict->full > distance;
126 /// Repeat *len bytes at distance.
128 dict_repeat(lzma_dict *dict, uint32_t distance, uint32_t *len)
130 // Don't write past the end of the dictionary.
131 const size_t dict_avail = dict->limit - dict->pos;
132 uint32_t left = my_min(dict_avail, *len);
135 // Repeat a block of data from the history. Because memcpy() is faster
136 // than copying byte by byte in a loop, the copying process gets split
138 if (distance < left) {
139 // Source and target areas overlap, thus we can't use
140 // memcpy() nor even memmove() safely.
142 dict->buf[dict->pos] = dict_get(dict, distance);
144 } while (--left > 0);
146 } else if (distance < dict->pos) {
147 // The easiest and fastest case
148 memcpy(dict->buf + dict->pos,
149 dict->buf + dict->pos - distance - 1,
154 // The bigger the dictionary, the more rare this
155 // case occurs. We need to "wrap" the dict, thus
156 // we might need two memcpy() to copy all the data.
157 assert(dict->full == dict->size);
158 const uint32_t copy_pos
159 = dict->pos - distance - 1 + dict->size;
160 uint32_t copy_size = dict->size - copy_pos;
162 if (copy_size < left) {
163 memmove(dict->buf + dict->pos, dict->buf + copy_pos,
165 dict->pos += copy_size;
166 copy_size = left - copy_size;
167 memcpy(dict->buf + dict->pos, dict->buf, copy_size);
168 dict->pos += copy_size;
170 memmove(dict->buf + dict->pos, dict->buf + copy_pos,
176 // Update how full the dictionary is.
177 if (dict->full < dict->pos)
178 dict->full = dict->pos;
180 return unlikely(*len != 0);
184 /// Puts one byte into the dictionary. Returns true if the dictionary was
185 /// already full and the byte couldn't be added.
187 dict_put(lzma_dict *dict, uint8_t byte)
189 if (unlikely(dict->pos == dict->limit))
192 dict->buf[dict->pos++] = byte;
194 if (dict->pos > dict->full)
195 dict->full = dict->pos;
201 /// Copies arbitrary amount of data into the dictionary.
203 dict_write(lzma_dict *restrict dict, const uint8_t *restrict in,
204 size_t *restrict in_pos, size_t in_size,
205 size_t *restrict left)
207 // NOTE: If we are being given more data than the size of the
208 // dictionary, it could be possible to optimize the LZ decoder
209 // so that not everything needs to go through the dictionary.
210 // This shouldn't be very common thing in practice though, and
211 // the slowdown of one extra memcpy() isn't bad compared to how
212 // much time it would have taken if the data were compressed.
214 if (in_size - *in_pos > *left)
215 in_size = *in_pos + *left;
217 *left -= lzma_bufcpy(in, in_pos, in_size,
218 dict->buf, &dict->pos, dict->limit);
220 if (dict->pos > dict->full)
221 dict->full = dict->pos;
228 dict_reset(lzma_dict *dict)
230 dict->need_reset = true;