static char sccsid
[] = "@(#)compress.c 5.1 (Berkeley) %G%";
* Set USERMEM to the maximum amount of physical user memory available
* in bytes. USERMEM is used to determine the maximum BITS that can be used
* for compression. If USERMEM is big enough, use fast compression algorithm.
* SACREDMEM is the amount of physical memory saved for others; compress
# define BITS 12 /* max bits/code for 16-bit machine */
# define NO_UCHAR /* also if "unsigned char" functions as signed char */
# define USERMEM 750000 /* default user memory */
* Define FBITS for machines with several MB of physical memory, to use
* table lookup for (b <= FBITS). If FBITS is made too large, performance
* will decrease due to increased swapping/paging. Since the program minus
* the fast lookup table is about a half Meg, we can allocate the rest of
* available physical memory to the fast lookup table.
* If FBITS is set to 12, a 2 MB array is allocated, but only 1 MB is
* addressed for parity-free input (i.e. text).
* FBITS=10 yields 1/2 meg lookup table + 4K code memory
* FBITS=11 yields 1 meg lookup table + 8K code memory
* FBITS=12 yields 2 meg lookup table + 16K code memory
* FBITS=13 yields 4 meg lookup table + 32K code memory
# if USERMEM >= (2621440+SACREDMEM)
# if USERMEM >= (4718592+SACREDMEM)
# else 2.5M <= USERMEM < 4.5M
# if USERMEM >= (1572864+SACREDMEM)
# if USERMEM >= (1048576+SACREDMEM)
# if USERMEM >= (631808+SACREDMEM)
# if USERMEM >= (329728+SACREDMEM)
# if USERMEM >= (178176+SACREDMEM)
# if USERMEM >= (99328+SACREDMEM)
#ifdef PBITS /* Preferred BITS for this memory size */
# define HSIZE 69001 /* 95% occupancy */
# define HSIZE 35023 /* 94% occupancy */
# define HSIZE 18013 /* 91% occupancy */
# define HSIZE 9001 /* 91% occupancy */
# define HSIZE 5003 /* 80% occupancy */
# define HSIZE 2591 /* 79% occupancy */
# define HSIZE 1291 /* 79% occupancy */
# define HSIZE 691 /* 74% occupancy */
/* BITS < 9 will cause an error */
* a code_int must be able to hold 2**BITS values of type int, and also -1
typedef long int code_int
;
typedef unsigned long int count_int
;
typedef unsigned short int count_short
;
typedef long int count_int
;
typedef unsigned char char_type
;
char_type magic_header
[] = { "\037\235" }; /* 1F 9D */
/* Defines for third byte of header */
/* Masks 0x40 and 0x20 are free. I think 0x20 should mean that there is
a fourth header byte (for expansion).
#define INIT_BITS 9 /* initial number of bits/code */
* compress.c - File compression ala IEEE Computer June 1984.
* Authors: Spencer W. Thomas (decvax!harpo!utah-cs!utah-gr!thomas)
* Jim McKie (decvax!mcvax!jim)
* Steve Davies (decvax!vax135!petsd!peora!srd)
* Ken Turkowski (decvax!decwrl!turtlevax!ken)
* James A. Woods (decvax!ihnp4!ames!jaw)
* Joe Orost (decvax!vax135!petsd!joe)
* $Header: compress.c,v 3.0 84/11/27 11:50:00 joe Exp $
* Revision 3.0 84/11/27 11:50:00 petsd!joe
* Set HSIZE depending on BITS. Set BITS depending on USERMEM. Unrolled
* loops in clear routines. Added "-C" flag for 2.0 compatability. Used
* unsigned compares on Perkin-Elmer. Fixed foreground check.
* Revision 2.7 84/11/16 19:35:39 ames!jaw
* Cache common hash codes based on input statistics; this improves
* performance for low-density raster images. Pass on #ifdef bundle
* Revision 2.6 84/11/05 19:18:21 ames!jaw
* Vary size of hash tables to reduce time for small files.
* Tune PDP-11 hash function.
* Revision 2.5 84/10/30 20:15:14 ames!jaw
* Junk chaining; replace with the simpler (and, on the VAX, faster)
* double hashing, discussed within. Make block compression standard.
* Revision 2.4 84/10/16 11:11:11 ames!jaw
* Introduce adaptive reset for block compression, to boost the rate
* another several percent. (See mailing list notes.)
* Revision 2.3 84/09/22 22:00:00 petsd!joe
* Implemented "-B" block compress. Implemented REVERSE sorting of tab_next.
* Bug fix for last bits. Changed fwrite to putchar loop everywhere.
* Revision 2.2 84/09/18 14:12:21 ames!jaw
* Fold in news changes, small machine typedef from thomas,
* #ifdef interdata from joe.
* Revision 2.1 84/09/10 12:34:56 ames!jaw
* Configured fast table lookup for 32-bit machines.
* This cuts user time in half for b <= FBITS, and is useful for news batching
* from VAX to PDP sites. Also sped up decompress() [fwrite->putc] and
* added signal catcher [plus beef in writeerr()] to delete effluvia.
* Revision 2.0 84/08/28 22:00:00 petsd!joe
* Add check for foreground before prompting user. Insert maxbits into
* compressed file. Force file being uncompressed to end with ".Z".
* Added "-c" flag and "zcat". Prepared for release.
* Revision 1.10 84/08/24 18:28:00 turtlevax!ken
* Will only compress regular files (no directories), added a magic number
* header (plus an undocumented -n flag to handle old files without headers),
* added -f flag to force overwriting of possibly existing destination file,
* otherwise the user is prompted for a response. Will tack on a .Z to a
* filename if it doesn't have one when decompressing. Will only replace
* file if it was compressed.
* Revision 1.9 84/08/16 17:28:00 turtlevax!ken
* Removed scanargs(), getopt(), added .Z extension and unlimited number of
* filenames to compress. Flags may be clustered (-Ddvb12) or separated
* (-D -d -v -b 12), or combination thereof. Modes and other status is
* copied with copystat(). -O bug for 4.2 seems to have disappeared with
* Revision 1.8 84/08/09 23:15:00 joe
* Made it compatible with vax version, installed jim's fixes/enhancements
* Revision 1.6 84/08/01 22:08:00 joe
* Sped up algorithm significantly by sorting the compress chain.
* Revision 1.5 84/07/13 13:11:00 srd
* Added C version of vax asm routines. Changed structure to arrays to
* save much memory. Do unsigned compares where possible (faster on
* Revision 1.4 84/07/05 03:11:11 thomas
* Clean up the code a little and lint it. (Lint complains about all
* the regs used in the asm, but I'm not going to "fix" this.)
* Revision 1.3 84/07/05 02:06:54 thomas
* Revision 1.2 84/07/05 00:27:27 thomas
* Add variable bit length output.
static char rcs_ident
[] = "$Header: compress.c,v 3.0 84/11/27 11:50:00 joe Exp$";
#define ARGVAL() (*++(*argv) || (--argc && *++argv))
int n_bits
; /* number of bits/code */
int maxbits
= BITS
; /* user settable max # bits/code */
code_int maxcode
; /* maximum code, given n_bits */
code_int maxmaxcode
= 1 << BITS
; /* should NEVER generate this code */
#ifdef COMPATIBLE /* But wrong! */
# define MAXCODE(n_bits) (1 << (n_bits) - 1)
# define MAXCODE(n_bits) ((1 << (n_bits)) - 1)
* One code could conceivably represent (1<<BITS) characters, but
* to get a code of length N requires an input string of at least
* N*(N-1)/2 characters. With 5000 chars in the stack, an input
* file would have to contain a 25Mb string of a single character.
# define MAXSTACK 5000 /* size of output stack */
# define MAXSTACK 8000 /* size of output stack */
unsigned short codetab
[HSIZE
];
code_int hsize
= HSIZE
; /* for dynamic table sizing */
#define tab_prefix codetab /* prefix code for this entry */
char_type tab_suffix
[1<<BITS
]; /* last char in this entry */
short ftable
[(1 << FBITS
) * 256];
count_int fcodemem
[1 << FBITS
];
code_int free_ent
= 0; /* first unused entry */
fprintf(stderr
,"Usage: compress [-dDvqfFc] [-b maxbits] [file ...]\n");
fprintf(stderr
,"Usage: compress [-dfFqc] [-b maxbits] [file ...]\n");
int nomagic
= 0; /* Use a 2 byte magic number header, unless old file */
int zcat_flg
= 0; /* Write output on stdout, suppress messages */
int quiet
= 0; /* don't tell me about compression */
* block compression parameters -- after all codes are used up,
* and compression rate changes, start over.
int block_compress
= BLOCK_MASK
;
double ratio
= 0.0; /* compression ratio for last block */
#define CHECK_GAP 10000 /* ratio check interval */
count_int checkpoint
= CHECK_GAP
;
* the next two codes should not be changed lightly, as they must not
* lie within the contiguous general code space.
#define FIRST 257 /* first free entry */
#define CLEAR 256 /* table clear output code */
/*****************************************************************
* Algorithm from "A Technique for High Performance Data Compression",
* Terry A. Welch, IEEE Computer Vol 17, No 6 (June 1984), pp 8-19.
* Usage: compress [-dfFqc] [-b bits] [file ...]
* -d: If given, decompression is done instead.
* -c: Write output on stdout, don't remove original.
* -b: Parameter limits the max number of bits/code.
* -f: Forces output file to be generated, even if one already
* exists; if -f is not used, the user will be prompted if
* the stdin is a tty, otherwise, the output file will not
* -F: Forces output file to be generated, even if no space is
* -q: No output, unless error
* file ...: Files to be compressed. If none specified, stdin
* file.Z: Compressed form of file with same mode, owner, and utimes
* or stdout (if stdin used as input)
* When filenames are given, replaces with the compressed version
* (.Z suffix) only if the file decreased in size.
* Modified Lempel-Ziv method (LZW). Basically finds common
* substrings and replaces them with a variable size code. This is
* deterministic, and can be done on the fly. Thus, the decompression
* procedure needs no input table, but tracks the way the table was
register int argc
; char **argv
;
int overwrite
= 0; /* Do not overwrite unless given -f flag */
char **filelist
, **fileptr
;
if ( (bgnd_flag
= signal ( SIGINT
, SIG_IGN
)) != SIG_IGN
)
signal ( SIGINT
, onintr
);
nomagic
= 1; /* Original didn't have a magic number */
filelist
= fileptr
= (char **)(malloc(argc
* sizeof(*argv
)));
if((cp
= rindex(argv
[0], '/')) != 0) {
if(strcmp(cp
, "uncompress") == 0) {
} else if(strcmp(cp
, "zcat") == 0) {
/* 4.2BSD dependent - take it out if not */
* All flags are optional.
* -f => force overwrite of output file
* -n => no header: useful to uncompress old files
* -b maxbits => maxbits. If -b is specified, then maxbits MUST be
* -c => cat all output to stdout
* -C => generate output compatable with compress 2.0.
* if a string is left, must be an input filename.
for (argc
--, argv
++; argc
> 0; argc
--, argv
++) {
if (**argv
== '-') { /* A flag argument */
while (*++(*argv
)) { /* Process all flags in this arg */
fprintf(stderr
, "Missing maxbits\n");
fprintf(stderr
, "Unknown flag: '%c'; ", **argv
);
else { /* Input file name */
*fileptr
++ = *argv
; /* Build input file list */
if(maxbits
< INIT_BITS
) maxbits
= INIT_BITS
;
if (maxbits
> BITS
) maxbits
= BITS
;
maxmaxcode
= 1 << maxbits
;
for (fileptr
= filelist
; *fileptr
; fileptr
++) {
if (do_decomp
!= 0) { /* DECOMPRESSION */
/* Check for .Z suffix */
if (strcmp(*fileptr
+ strlen(*fileptr
) - 2, ".Z") != 0) {
strcpy(tempname
, *fileptr
);
if ((freopen(*fileptr
, "r", stdin
)) == NULL
) {
perror(*fileptr
); continue;
/* Check the magic number */
if ((getchar() != (magic_header
[0] & 0xFF))
|| (getchar() != (magic_header
[1] & 0xFF))) {
fprintf(stderr
, "%s: not in compressed format\n",
maxbits
= getchar(); /* set -b from file */
block_compress
= maxbits
& BLOCK_MASK
;
maxmaxcode
= 1 << maxbits
;
"%s: compressed with %d bits, can only handle %d bits\n",
*fileptr
, maxbits
, BITS
);
/* Generate output filename */
strcpy(ofname
, *fileptr
);
ofname
[strlen(*fileptr
) - 2] = '\0'; /* Strip off .Z */
} else { /* COMPRESSION */
if (strcmp(*fileptr
+ strlen(*fileptr
) - 2, ".Z") == 0) {
fprintf(stderr
, "%s: already has .Z suffix -- no change\n",
if ((freopen(*fileptr
, "r", stdin
)) == NULL
) {
perror(*fileptr
); continue;
stat ( *fileptr
, &statbuf
);
fsize
= (long) statbuf
.st_size
;
* tune hash table size for small files -- ad hoc
else if ( fsize
< (1 << 13) )
else if ( fsize
< (1 << 14) )
else if ( fsize
< (1 << 15) )
else if ( fsize
< 47000 )
/* Generate output filename */
strcpy(ofname
, *fileptr
);
#ifndef BSD4_2 /* Short filenames */
if ((cp
=rindex(ofname
,'/')) != NULL
) cp
++;
fprintf(stderr
,"%s: filename too long to tack on .Z\n",cp
);
#endif BSD4_2 /* Long filenames allowed */
/* Check for overwrite of existing file */
if (overwrite
== 0 && zcat_flg
== 0) {
if (stat(ofname
, &statbuf
) == 0) {
fprintf(stderr
, "%s already exists;", ofname
);
fprintf(stderr
, " do you wish to overwrite (y or n)? ",
while (response
[1] != '\n') {
if (read(2, response
+1, 1) < 0) { /* Ack! */
if (response
[0] != 'y') {
fprintf(stderr
, "\tnot overwritten\n");
if(zcat_flg
== 0) { /* Open output file */
if (freopen(ofname
, "w", stdout
) == NULL
) {
fprintf(stderr
, "%s: ", *fileptr
);
/* Actually do the compression/decompression */
if (do_decomp
== 0) compress();
else if (debug
== 0) decompress();
copystat(*fileptr
, ofname
); /* Copy stats */
if(exit_stat
|| (!quiet
))
} else { /* Standard input */
/* Check the magic number */
if ((getchar()!=(magic_header
[0] & 0xFF))
|| (getchar()!=(magic_header
[1] & 0xFF))) {
fprintf(stderr
, "stdin: not in compressed format\n");
maxbits
= getchar(); /* set -b from file */
block_compress
= maxbits
& BLOCK_MASK
;
maxmaxcode
= 1 << maxbits
;
fsize
= 100000; /* assume stdin large for USERMEM */
"stdin: compressed with %d bits, can only handle %d bits\n",
if (debug
== 0) decompress();
long int in_count
= 1; /* length of input */
long int bytes_out
; /* length of compressed output */
long int out_count
= 0; /* # of codes output (for debugging) */
#define HOG_CHECK ((count_int) 2000) /* Number of chars to read b4 check */
#define MAX_CACHE ((count_int) 1<<BITS) /* Next line is this constant too */
unsigned short hashcache
[1<<BITS
]; /* common hash short circuit cache */
count_int cfreq
[256]; /* character counts */
char chog
; /* most common character from input */
# define CHOG ' ' /* Assume space is most frequent */
int chog
; /* char arith slow on VAX */
# define CHOG (int) ' ' /* Assume space is most frequent */
int cstat_flg
= 0; /* on after determining char hog */
* compress stdin to stdout
* Algorithm: on large machines, for maxbits <= FBITS, use fast direct table
* lookup on the prefix code / next character combination. For smaller code
* size, use open addressing modular division double hashing (no chaining), ala
* Knuth vol. 3, sec. 6.4 Algorithm D, along with G. Knott's relatively-prime
* secondary probe. Do block compression with an adaptive reset, whereby the
* code table is cleared when the compression ratio decreases, but after the
* table fills. The variable-length output codes are re-sized at this point,
* and a special CLEAR code is generated for the decompressor. For the
* megamemory version, the sparse array is cleared indirectly through a
* "shadow" output code history. Late additions: for the hashing code,
* construct the table according to file size for noticeable speed improvement
* on small files. Also detect and cache codes associated with the most
* common character to bypass hash calculation on these codes (a characteristic
* of highly-compressable raster images). Please direct questions about this
* implementation to ames!jaw.
register code_int hsize_reg
;
putchar(magic_header
[0]); putchar(magic_header
[1]);
putchar((char)(maxbits
| block_compress
));
maxcode
= MAXCODE(n_bits
= INIT_BITS
);
free_ent
= ((block_compress
) ? FIRST
: 256 );
if ( maxbits
<= FBITS
&& (fsize
>= 30000) ) { /* use hashing on small files */
while ( (c
= getchar()) != (unsigned) EOF
) {
fcode
= (long) (((long) c
<< maxbits
) + ent
);
if ( ftable
[fcode
] != 0 ) /* test for code in "string" table */
output ( (code_int
) ent
);
if ( free_ent
>= maxmaxcode
) {
if ( (count_int
)in_count
< checkpoint
|| (!block_compress
) )
} else { /* put code in table */
ftable
[fcode
] = (short) free_ent
++;
fcodemem
[i
++] = fcode
; /* memorize for block compression */
chog
= CHOG
; /* assumed character for the hog */
cl_hash(hsize_reg
); /* clear hash tables */
while ( (c
= getchar()) != (unsigned) EOF
) {
cfreq
[c
]++; /* gather frequencies at start of input */
if ( (count_int
)in_count
> HOG_CHECK
) {
chog
= hogtally(); /* compute char hog */
if(chog
!= CHOG
) /* fixup for wrong assumption */
creset( (count_int
) free_ent
);
if ( (i
= hashcache
[ent
]) ) { /* cache -> code */
fcode
= (long) (((long) c
<< maxbits
) + ent
);
i
= (((c
+ 12347) * ent
) & 077777) % HSIZE
; /* avoid 'lrem' call */
i
= fcode
% hsize_reg
; /* division hashing */
if ( htab
[i
] == fcode
) {
} else if ( (long)htab
[i
] < 0 ) /* empty slot */
disp
= hsize_reg
- i
; /* secondary hash (G. Knott) */
if ( htab
[i
] == fcode
) {
if ( (long)htab
[i
] > 0 )
output ( (code_int
) ent
);
if ( (unsigned) free_ent
< (unsigned) maxmaxcode
) {
if ( free_ent
< maxmaxcode
) {
if ( c
== chog
) /* code -> cache */
hashcache
[ent
] = free_ent
;
codetab
[i
] = free_ent
++;
else if ( (count_int
)in_count
>= checkpoint
&& block_compress
)
* Put out the final code.
* Print out stats on stderr
if(zcat_flg
== 0 && !quiet
) {
"%ld chars in, %ld codes (%ld bytes) out, compression factor %g\n",
in_count
, out_count
, bytes_out
,
(double)in_count
/ (double)bytes_out
);
fprintf( stderr
, "\tCompression as in compact: %5.2f%%\n",
100.0 * ( in_count
- bytes_out
) / (double) in_count
);
fprintf( stderr
, "\tLargest code was %d (%d bits)\n", free_ent
- 1, n_bits
);
fprintf( stderr
, "Compression: %5.2f%%",
100.0 * ( in_count
- bytes_out
) / (double) in_count
);
if(bytes_out
> in_count
) /* exit(2) if no savings */
/*****************************************************************
* code: A n_bits-bit integer. If == -1, then EOF. This assumes
* that n_bits =< (long)wordsize - 1.
* Outputs code to the file.
* Maintain a BITS character long buffer (so that 8 codes will
* fit in it exactly). Use the VAX insv instruction to insert each
* code in turn. When the buffer fills up empty it and start over.
char_type lmask
[9] = {0xff, 0xfe, 0xfc, 0xf8, 0xf0, 0xe0, 0xc0, 0x80, 0x00};
char_type rmask
[9] = {0x00, 0x01, 0x03, 0x07, 0x0f, 0x1f, 0x3f, 0x7f, 0xff};
* On the VAX, it is important to have the register declarations
* in exactly the order given, or the asm will break.
register int r_off
= offset
, bits
= n_bits
;
register char * bp
= buf
;
/* VAX DEPENDENT!! Implementation on other machines may be
* Translation: Insert BITS bits from the argument starting at
* offset bits from the beginning of buf.
0; /* C compiler bug ?? */
asm( "insv 4(ap),r11,r10,(r9)" );
/* WARNING: byte/bit numbering on the vax is simulated by the following code
* Since code is always >= 8 bits, only need to mask the first
*bp
= (*bp
& rmask
[r_off
]) | (code
<< r_off
) & lmask
[r_off
];
/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
if ( offset
== (n_bits
<< 3) ) {
fprintf( stderr
, "%5d%c", code
,
(col
+=6) >= 74 ? (col
= 0, '\n') : ' ' );
* If the next entry is going to be too big for the code size,
* then increase it, if possible.
if ( free_ent
> maxcode
|| (clear_flg
> 0))
* Write the whole buffer, because the input side won't
* discover the size increase until after it has read it.
if( fwrite( buf
, 1, n_bits
, stdout
) != n_bits
)
maxcode
= MAXCODE (n_bits
= INIT_BITS
);
maxcode
= MAXCODE(n_bits
);
fprintf( stderr
, "\nChange to %d bits\n", n_bits
);
* At EOF, write the rest of the buffer.
fwrite( buf
, 1, (offset
+ 7) / 8, stdout
);
bytes_out
+= (offset
+ 7) / 8;
register int stack_top
= MAXSTACK
;
register code_int code
, oldcode
, incode
;
* As above, initialize the first 256 entries in the table.
maxcode
= MAXCODE(n_bits
= INIT_BITS
);
for ( code
= 255; code
>= 0; code
-- ) {
tab_suffix
[code
] = (char_type
)code
;
free_ent
= ((block_compress
) ? FIRST
: 256 );
finchar
= oldcode
= getcode();
putchar( (char)finchar
); /* first code must be 8 bits = char */
while ( (code
= getcode()) != -1 ) {
if ( (code
== CLEAR
) && block_compress
) {
for ( code
= 255; code
> 0; code
-= 4 ) {
if ( (code
= getcode ()) == -1 ) /* O, untimely death! */
* Special case for KwKwK string.
if ( code
>= free_ent
) {
stack
[--stack_top
] = finchar
;
* Generate output characters in reverse order
while ( ((unsigned long)code
) >= ((unsigned long)256) ) {
stack
[--stack_top
] = tab_suffix
[code
];
stack
[--stack_top
] = finchar
= tab_suffix
[code
];
* And put them out in forward order
for ( ; stack_top
< MAXSTACK
; stack_top
++ )
putchar(stack
[stack_top
]);
* Generate the new entry.
if ( (code
=free_ent
) < maxmaxcode
) {
tab_prefix
[code
] = (unsigned short)oldcode
;
tab_suffix
[code
] = finchar
;
* Remember previous code.
/*****************************************************************
* Read one code from the standard input. If EOF, return -1.
* code or -1 is returned.
* On the VAX, it is important to have the register declarations
* in exactly the order given, or the asm will break.
static int offset
= 0, size
= 0;
static char_type buf
[BITS
];
register int r_off
, bits
;
register char_type
*bp
= buf
;
if ( clear_flg
> 0 || offset
>= size
|| free_ent
> maxcode
) {
* If the next entry will be too big for the current code
* size, then we must increase the size. This implies reading
* a new buffer full, too.
if ( free_ent
> maxcode
) {
maxcode
= maxmaxcode
; /* won't get any bigger now */
maxcode
= MAXCODE(n_bits
);
maxcode
= MAXCODE (n_bits
= INIT_BITS
);
size
= fread( buf
, 1, n_bits
, stdin
);
return -1; /* end of file */
/* Round size down to integral number of codes */
size
= (size
<< 3) - (n_bits
- 1);
asm( "extzv r10,r9,(r8),r11" );
/* Get first part (low order bits) */
code
= ((*bp
++ >> r_off
) & rmask
[8 - r_off
]) & 0xff;
r_off
= 8 - r_off
; /* now, offset into code word */
/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
code
|= (*bp
++ & 0xff) << r_off
;
code
|= (*bp
& rmask
[bits
]) << r_off
;
rindex(s
, c
) /* For those who don't have it in libc.a */
* Just print out codes from input file. Mostly for debugging.
bits
= n_bits
= INIT_BITS
;
maxcode
= MAXCODE(n_bits
);
free_ent
= ((block_compress
) ? FIRST
: 256 );
while ( ( code
= getcode() ) >= 0 ) {
if ( (code
== CLEAR
) && block_compress
) {
else if ( free_ent
< maxmaxcode
)
fprintf(stderr
, "\nChange to %d bits\n", n_bits
);
fprintf(stderr
, "%5d%c", code
, (col
+=6) >= 74 ? (col
= 0, '\n') : ' ' );
dump_tab() /* dump string table */
char stack
[4 * MAXSTACK
]; /* \nnn makes it 4 times bigger */
int stack_top
= 4 * MAXSTACK
;
for ( i
= 0; i
< free_ent
; i
++ ) {
if ( isascii(tab_suffix
[ent
]) && isprint(tab_suffix
[ent
]) )
fprintf( stderr
, "%5d: %5d/'%c' \"",
ent
, tab_prefix
[ent
], tab_suffix
[ent
] );
fprintf( stderr
, "%5d: %5d/\\%03o \"",
ent
, tab_prefix
[ent
], tab_suffix
[ent
] );
stack
[--stack_top
] = '\n';
stack
[--stack_top
] = '"';
ent
= (ent
>= FIRST
? tab_prefix
[ent
] : NULL
) ) {
if ( isascii(tab_suffix
[ent
]) && isprint(tab_suffix
[ent
]) )
stack
[--stack_top
] = tab_suffix
[ent
];
switch( tab_suffix
[ent
] ) {
case '\n': stack
[--stack_top
] = 'n'; break;
case '\t': stack
[--stack_top
] = 't'; break;
case '\b': stack
[--stack_top
] = 'b'; break;
case '\f': stack
[--stack_top
] = 'f'; break;
case '\r': stack
[--stack_top
] = 'r'; break;
stack
[--stack_top
] = '0' + tab_suffix
[ent
] % 8;
stack
[--stack_top
] = '0' + (tab_suffix
[ent
] / 8) % 8;
stack
[--stack_top
] = '0' + tab_suffix
[ent
] / 64;
stack
[--stack_top
] = '\\';
fwrite( &stack
[stack_top
], 1, 4 * MAXSTACK
- stack_top
, stderr
);
stack_top
= 4 * MAXSTACK
;
/*****************************************************************
* Exits with a message. We only check for write errors often enough
* to avoid a lot of "file system full" messages, not on every write.
* ferror() check after fflush will catch any others (I trust).
if (stat(ifname
, &statbuf
)) { /* Get stat on input file */
if ((statbuf
.st_mode
& S_IFMT
/*0170000*/) != S_IFREG
/*0100000*/) {
fprintf(stderr
, "%s: ", ifname
);
fprintf(stderr
, " -- not a regular file: unchanged");
} else if (statbuf
.st_nlink
> 1) {
fprintf(stderr
, "%s: ", ifname
);
fprintf(stderr
, " -- has %d other links: unchanged",
} else if (exit_stat
== 2 && (!force
)) { /* No compression: remove file.Z */
fprintf(stderr
, " -- file unchanged");
} else { /* ***** Successful Compression ***** */
mode
= statbuf
.st_mode
& 07777;
if (chmod(ofname
, mode
)) /* Copy modes */
chown(ofname
, statbuf
.st_uid
, statbuf
.st_gid
); /* Copy ownership */
timep
[0] = statbuf
.st_atime
;
timep
[1] = statbuf
.st_mtime
;
utime(ofname
, timep
); /* Update last accessed and modified times */
if (unlink(ifname
)) /* Remove input file */
fprintf(stderr
, " -- replaced with %s", ofname
);
return; /* Successful return */
/* Unsuccessful return -- one of the tests failed */
* This routine returns 1 if we are running in the foreground and stderr
if(bgnd_flag
) { /* background? */
} else { /* foreground */
if(isatty(2)) { /* and stderr is a tty */
clear () /* table clear for block compress */
register count_int
*p
, *endp
;
register unsigned short *q
;
fprintf ( stderr
, "count: %ld ratio: %f\n", in_count
,
(double) in_count
/ (double) bytes_out
);
checkpoint
= in_count
+ CHECK_GAP
;
if ( (double) in_count
/ (double) bytes_out
> ratio
)
ratio
= (double) in_count
/ (double) bytes_out
;
if ( maxbits
<= FBITS
) /* sparse array clear */
for ( i
= (1 << maxbits
) - 1; i
>= 0; i
-- )
ftable
[fcodemem
[i
]] = 0; /* indirect thru "shadow" */
#endif USERMEM /* hash table clear */
for ( p
= &htab
[0], q
= &codetab
[0]; p
< endp
; ) {
output ( (code_int
) CLEAR
);
fprintf ( stderr
, "clear\n" );
creset ( n
) /* clear hash cache */
register count_int n
; /* clear at least this many entries */
register unsigned short *hash_p
;
register unsigned short zero
= 0;
if ( nfiles
++ == 0 ) /* No clear needed if first time */
for ( i
= n
; i
> 0; i
-=16 ) {
hogtally () /* compute character code hog */
for ( i
= most
= 0; i
< 256; i
++ )
if ( cfreq
[i
] >= cfreq
[most
] )
register count_int
*htab_p
= htab
+hsize
;
#define min(a,b) ((a>b) ? b : a)
creset( min((count_int
)hsize
, MAX_CACHE
) );
} while ((i
-= 16) >= 0);
for ( i
+= 16; i
> 0; i
-- )