source: s10k/CommonLibs/zlib-1.2.8/contrib/infback9/inftree9.c @ 1096

Last change on this file since 1096 was 1096, checked in by s10k, 19 months ago

Added zlib, quazip, basicxmlsyntaxhighlighter, conditionalsemaphore and linenumberdisplay libraries. zlib and quazip are pre-compiled, but you can compile them yourself, just delete the dll files (or equivalent binary files to your OS)

File size: 13.1 KB
Line 
1/* inftree9.c -- generate Huffman trees for efficient decoding
2 * Copyright (C) 1995-2013 Mark Adler
3 * For conditions of distribution and use, see copyright notice in zlib.h
4 */
5
6#include "zutil.h"
7#include "inftree9.h"
8
9#define MAXBITS 15
10
11const char inflate9_copyright[] =
12   " inflate9 1.2.8 Copyright 1995-2013 Mark Adler ";
13/*
14  If you use the zlib library in a product, an acknowledgment is welcome
15  in the documentation of your product. If for some reason you cannot
16  include such an acknowledgment, I would appreciate that you keep this
17  copyright string in the executable of your product.
18 */
19
20/*
21   Build a set of tables to decode the provided canonical Huffman code.
22   The code lengths are lens[0..codes-1].  The result starts at *table,
23   whose indices are 0..2^bits-1.  work is a writable array of at least
24   lens shorts, which is used as a work area.  type is the type of code
25   to be generated, CODES, LENS, or DISTS.  On return, zero is success,
26   -1 is an invalid code, and +1 means that ENOUGH isn't enough.  table
27   on return points to the next available entry's address.  bits is the
28   requested root table index bits, and on return it is the actual root
29   table index bits.  It will differ if the request is greater than the
30   longest code or if it is less than the shortest code.
31 */
32int inflate_table9(type, lens, codes, table, bits, work)
33codetype type;
34unsigned short FAR *lens;
35unsigned codes;
36code FAR * FAR *table;
37unsigned FAR *bits;
38unsigned short FAR *work;
39{
40    unsigned len;               /* a code's length in bits */
41    unsigned sym;               /* index of code symbols */
42    unsigned min, max;          /* minimum and maximum code lengths */
43    unsigned root;              /* number of index bits for root table */
44    unsigned curr;              /* number of index bits for current table */
45    unsigned drop;              /* code bits to drop for sub-table */
46    int left;                   /* number of prefix codes available */
47    unsigned used;              /* code entries in table used */
48    unsigned huff;              /* Huffman code */
49    unsigned incr;              /* for incrementing code, index */
50    unsigned fill;              /* index for replicating entries */
51    unsigned low;               /* low bits for current root entry */
52    unsigned mask;              /* mask for low root bits */
53    code this;                  /* table entry for duplication */
54    code FAR *next;             /* next available space in table */
55    const unsigned short FAR *base;     /* base value table to use */
56    const unsigned short FAR *extra;    /* extra bits table to use */
57    int end;                    /* use base and extra for symbol > end */
58    unsigned short count[MAXBITS+1];    /* number of codes of each length */
59    unsigned short offs[MAXBITS+1];     /* offsets in table for each length */
60    static const unsigned short lbase[31] = { /* Length codes 257..285 base */
61        3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17,
62        19, 23, 27, 31, 35, 43, 51, 59, 67, 83, 99, 115,
63        131, 163, 195, 227, 3, 0, 0};
64    static const unsigned short lext[31] = { /* Length codes 257..285 extra */
65        128, 128, 128, 128, 128, 128, 128, 128, 129, 129, 129, 129,
66        130, 130, 130, 130, 131, 131, 131, 131, 132, 132, 132, 132,
67        133, 133, 133, 133, 144, 72, 78};
68    static const unsigned short dbase[32] = { /* Distance codes 0..31 base */
69        1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49,
70        65, 97, 129, 193, 257, 385, 513, 769, 1025, 1537, 2049, 3073,
71        4097, 6145, 8193, 12289, 16385, 24577, 32769, 49153};
72    static const unsigned short dext[32] = { /* Distance codes 0..31 extra */
73        128, 128, 128, 128, 129, 129, 130, 130, 131, 131, 132, 132,
74        133, 133, 134, 134, 135, 135, 136, 136, 137, 137, 138, 138,
75        139, 139, 140, 140, 141, 141, 142, 142};
76
77    /*
78       Process a set of code lengths to create a canonical Huffman code.  The
79       code lengths are lens[0..codes-1].  Each length corresponds to the
80       symbols 0..codes-1.  The Huffman code is generated by first sorting the
81       symbols by length from short to long, and retaining the symbol order
82       for codes with equal lengths.  Then the code starts with all zero bits
83       for the first code of the shortest length, and the codes are integer
84       increments for the same length, and zeros are appended as the length
85       increases.  For the deflate format, these bits are stored backwards
86       from their more natural integer increment ordering, and so when the
87       decoding tables are built in the large loop below, the integer codes
88       are incremented backwards.
89
90       This routine assumes, but does not check, that all of the entries in
91       lens[] are in the range 0..MAXBITS.  The caller must assure this.
92       1..MAXBITS is interpreted as that code length.  zero means that that
93       symbol does not occur in this code.
94
95       The codes are sorted by computing a count of codes for each length,
96       creating from that a table of starting indices for each length in the
97       sorted table, and then entering the symbols in order in the sorted
98       table.  The sorted table is work[], with that space being provided by
99       the caller.
100
101       The length counts are used for other purposes as well, i.e. finding
102       the minimum and maximum length codes, determining if there are any
103       codes at all, checking for a valid set of lengths, and looking ahead
104       at length counts to determine sub-table sizes when building the
105       decoding tables.
106     */
107
108    /* accumulate lengths for codes (assumes lens[] all in 0..MAXBITS) */
109    for (len = 0; len <= MAXBITS; len++)
110        count[len] = 0;
111    for (sym = 0; sym < codes; sym++)
112        count[lens[sym]]++;
113
114    /* bound code lengths, force root to be within code lengths */
115    root = *bits;
116    for (max = MAXBITS; max >= 1; max--)
117        if (count[max] != 0) break;
118    if (root > max) root = max;
119    if (max == 0) return -1;            /* no codes! */
120    for (min = 1; min <= MAXBITS; min++)
121        if (count[min] != 0) break;
122    if (root < min) root = min;
123
124    /* check for an over-subscribed or incomplete set of lengths */
125    left = 1;
126    for (len = 1; len <= MAXBITS; len++) {
127        left <<= 1;
128        left -= count[len];
129        if (left < 0) return -1;        /* over-subscribed */
130    }
131    if (left > 0 && (type == CODES || max != 1))
132        return -1;                      /* incomplete set */
133
134    /* generate offsets into symbol table for each length for sorting */
135    offs[1] = 0;
136    for (len = 1; len < MAXBITS; len++)
137        offs[len + 1] = offs[len] + count[len];
138
139    /* sort symbols by length, by symbol order within each length */
140    for (sym = 0; sym < codes; sym++)
141        if (lens[sym] != 0) work[offs[lens[sym]]++] = (unsigned short)sym;
142
143    /*
144       Create and fill in decoding tables.  In this loop, the table being
145       filled is at next and has curr index bits.  The code being used is huff
146       with length len.  That code is converted to an index by dropping drop
147       bits off of the bottom.  For codes where len is less than drop + curr,
148       those top drop + curr - len bits are incremented through all values to
149       fill the table with replicated entries.
150
151       root is the number of index bits for the root table.  When len exceeds
152       root, sub-tables are created pointed to by the root entry with an index
153       of the low root bits of huff.  This is saved in low to check for when a
154       new sub-table should be started.  drop is zero when the root table is
155       being filled, and drop is root when sub-tables are being filled.
156
157       When a new sub-table is needed, it is necessary to look ahead in the
158       code lengths to determine what size sub-table is needed.  The length
159       counts are used for this, and so count[] is decremented as codes are
160       entered in the tables.
161
162       used keeps track of how many table entries have been allocated from the
163       provided *table space.  It is checked for LENS and DIST tables against
164       the constants ENOUGH_LENS and ENOUGH_DISTS to guard against changes in
165       the initial root table size constants.  See the comments in inftree9.h
166       for more information.
167
168       sym increments through all symbols, and the loop terminates when
169       all codes of length max, i.e. all codes, have been processed.  This
170       routine permits incomplete codes, so another loop after this one fills
171       in the rest of the decoding tables with invalid code markers.
172     */
173
174    /* set up for code type */
175    switch (type) {
176    case CODES:
177        base = extra = work;    /* dummy value--not used */
178        end = 19;
179        break;
180    case LENS:
181        base = lbase;
182        base -= 257;
183        extra = lext;
184        extra -= 257;
185        end = 256;
186        break;
187    default:            /* DISTS */
188        base = dbase;
189        extra = dext;
190        end = -1;
191    }
192
193    /* initialize state for loop */
194    huff = 0;                   /* starting code */
195    sym = 0;                    /* starting code symbol */
196    len = min;                  /* starting code length */
197    next = *table;              /* current table to fill in */
198    curr = root;                /* current table index bits */
199    drop = 0;                   /* current bits to drop from code for index */
200    low = (unsigned)(-1);       /* trigger new sub-table when len > root */
201    used = 1U << root;          /* use root table entries */
202    mask = used - 1;            /* mask for comparing low */
203
204    /* check available table space */
205    if ((type == LENS && used >= ENOUGH_LENS) ||
206        (type == DISTS && used >= ENOUGH_DISTS))
207        return 1;
208
209    /* process all codes and make table entries */
210    for (;;) {
211        /* create table entry */
212        this.bits = (unsigned char)(len - drop);
213        if ((int)(work[sym]) < end) {
214            this.op = (unsigned char)0;
215            this.val = work[sym];
216        }
217        else if ((int)(work[sym]) > end) {
218            this.op = (unsigned char)(extra[work[sym]]);
219            this.val = base[work[sym]];
220        }
221        else {
222            this.op = (unsigned char)(32 + 64);         /* end of block */
223            this.val = 0;
224        }
225
226        /* replicate for those indices with low len bits equal to huff */
227        incr = 1U << (len - drop);
228        fill = 1U << curr;
229        do {
230            fill -= incr;
231            next[(huff >> drop) + fill] = this;
232        } while (fill != 0);
233
234        /* backwards increment the len-bit code huff */
235        incr = 1U << (len - 1);
236        while (huff & incr)
237            incr >>= 1;
238        if (incr != 0) {
239            huff &= incr - 1;
240            huff += incr;
241        }
242        else
243            huff = 0;
244
245        /* go to next symbol, update count, len */
246        sym++;
247        if (--(count[len]) == 0) {
248            if (len == max) break;
249            len = lens[work[sym]];
250        }
251
252        /* create new sub-table if needed */
253        if (len > root && (huff & mask) != low) {
254            /* if first time, transition to sub-tables */
255            if (drop == 0)
256                drop = root;
257
258            /* increment past last table */
259            next += 1U << curr;
260
261            /* determine length of next table */
262            curr = len - drop;
263            left = (int)(1 << curr);
264            while (curr + drop < max) {
265                left -= count[curr + drop];
266                if (left <= 0) break;
267                curr++;
268                left <<= 1;
269            }
270
271            /* check for enough space */
272            used += 1U << curr;
273            if ((type == LENS && used >= ENOUGH_LENS) ||
274                (type == DISTS && used >= ENOUGH_DISTS))
275                return 1;
276
277            /* point entry in root table to sub-table */
278            low = huff & mask;
279            (*table)[low].op = (unsigned char)curr;
280            (*table)[low].bits = (unsigned char)root;
281            (*table)[low].val = (unsigned short)(next - *table);
282        }
283    }
284
285    /*
286       Fill in rest of table for incomplete codes.  This loop is similar to the
287       loop above in incrementing huff for table indices.  It is assumed that
288       len is equal to curr + drop, so there is no loop needed to increment
289       through high index bits.  When the current sub-table is filled, the loop
290       drops back to the root table to fill in any remaining entries there.
291     */
292    this.op = (unsigned char)64;                /* invalid code marker */
293    this.bits = (unsigned char)(len - drop);
294    this.val = (unsigned short)0;
295    while (huff != 0) {
296        /* when done with sub-table, drop back to root table */
297        if (drop != 0 && (huff & mask) != low) {
298            drop = 0;
299            len = root;
300            next = *table;
301            curr = root;
302            this.bits = (unsigned char)len;
303        }
304
305        /* put invalid code marker in table */
306        next[huff >> drop] = this;
307
308        /* backwards increment the len-bit code huff */
309        incr = 1U << (len - 1);
310        while (huff & incr)
311            incr >>= 1;
312        if (incr != 0) {
313            huff &= incr - 1;
314            huff += incr;
315        }
316        else
317            huff = 0;
318    }
319
320    /* set return parameters */
321    *table += used;
322    *bits = root;
323    return 0;
324}
Note: See TracBrowser for help on using the repository browser.