Added another single byte oriented decompressor, useful for

comparision functions.

Added all lztext comparision functions, operators and a default
operator class for nbtree on lztext.

Jan
This commit is contained in:
Jan Wieck 1999-11-25 01:28:07 +00:00
parent 4ae43c84c3
commit 4069d48aa7
9 changed files with 435 additions and 9 deletions

View File

@ -1,7 +1,7 @@
/* ----------
* lztext.c -
*
* $Header: /cvsroot/pgsql/src/backend/utils/adt/Attic/lztext.c,v 1.3 1999/11/24 03:45:12 ishii Exp $
* $Header: /cvsroot/pgsql/src/backend/utils/adt/Attic/lztext.c,v 1.4 1999/11/25 01:28:04 wieck Exp $
*
* Text type with internal LZ compressed representation. Uses the
* standard PostgreSQL compression method.
@ -290,3 +290,146 @@ lztext_text(lztext *lz)
}
/* ----------
* lztext_cmp -
*
* Comparision function for two lztext datum's.
*
* Returns -1, 0 or 1.
* ----------
*/
int32
lztext_cmp(lztext *lz1, lztext *lz2)
{
#ifdef USE_LOCALE
char *cp1;
char *cp2;
int result;
if (lz1 == NULL || lz2 == NULL)
return (int32)0;
cp1 = lztextout(lz1);
cp2 = lztextout(lz2);
result = strcoll(cp1, cp2);
pfree(cp1);
pfree(cp2);
return result;
#else /* !USE_LOCALE */
PGLZ_DecompState ds1;
PGLZ_DecompState ds2;
int c1;
int c2;
int32 result = (int32)0;
if (lz1 == NULL || lz2 == NULL)
return (int32)0;
pglz_decomp_init(&ds1, lz1);
pglz_decomp_init(&ds2, lz2);
for(;;)
{
c1 = pglz_decomp_getchar(&ds1);
c2 = pglz_decomp_getchar(&ds2);
if (c1 == EOF)
{
if (c2 != EOF)
result = (int32)-1;
break;
} else {
if (c2 == EOF)
{
result = (int32)1;
}
}
if (c1 != c2)
{
result = (int32)(c1 - c2);
break;
}
}
pglz_decomp_end(&ds1);
pglz_decomp_end(&ds2);
return result;
#endif /* USE_LOCALE */
}
/* ----------
* lztext_eq ... -
*
* =, !=, >, >=, < and <= operator functions for two
* lztext datums.
* ----------
*/
bool
lztext_eq(lztext *lz1, lztext *lz2)
{
if (lz1 == NULL || lz2 == NULL)
return false;
return (bool)(lztext_cmp(lz1, lz2) == 0);
}
bool
lztext_ne(lztext *lz1, lztext *lz2)
{
if (lz1 == NULL || lz2 == NULL)
return false;
return (bool)(lztext_cmp(lz1, lz2) != 0);
}
bool
lztext_gt(lztext *lz1, lztext *lz2)
{
if (lz1 == NULL || lz2 == NULL)
return false;
return (bool)(lztext_cmp(lz1, lz2) > 0);
}
bool
lztext_ge(lztext *lz1, lztext *lz2)
{
if (lz1 == NULL || lz2 == NULL)
return false;
return (bool)(lztext_cmp(lz1, lz2) >= 0);
}
bool
lztext_lt(lztext *lz1, lztext *lz2)
{
if (lz1 == NULL || lz2 == NULL)
return false;
return (bool)(lztext_cmp(lz1, lz2) < 0);
}
bool
lztext_le(lztext *lz1, lztext *lz2)
{
if (lz1 == NULL || lz2 == NULL)
return false;
return (bool)(lztext_cmp(lz1, lz2) <= 0);
}

View File

@ -1,7 +1,7 @@
/* ----------
* pg_lzcompress.c -
*
* $Header: /cvsroot/pgsql/src/backend/utils/adt/pg_lzcompress.c,v 1.2 1999/11/17 22:18:45 wieck Exp $
* $Header: /cvsroot/pgsql/src/backend/utils/adt/pg_lzcompress.c,v 1.3 1999/11/25 01:28:04 wieck Exp $
*
* This is an implementation of LZ compression for PostgreSQL.
* It uses a simple history table and generates 2-3 byte tags
@ -671,3 +671,167 @@ pglz_decompress (PGLZ_Header *source, char *dest)
}
/* ----------
* pglz_get_next_decomp_char_from_lzdata -
*
* Reads the next character from a decompression state if the
* input data to pglz_decomp_init() was in compressed format.
* ----------
*/
int
pglz_get_next_decomp_char_from_lzdata(PGLZ_DecompState *dstate)
{
unsigned char retval;
if (dstate->tocopy > 0)
{
/* ----------
* Copy one byte from output to output until we did it
* for the length specified by the last tag. Return that
* byte.
* ----------
*/
dstate->tocopy--;
return (*(dstate->cp_out++) = *(dstate->cp_copy++));
}
if (dstate->ctrl_count == 0)
{
/* ----------
* Get the next control byte if we need to, but check
* for EOF before.
* ----------
*/
if (dstate->cp_in == dstate->cp_end)
{
return EOF;
}
/* ----------
* This decompression method saves time only, if we stop near
* the beginning of the data (maybe because we're called by a
* comparision function and a difference occurs early). Otherwise,
* all the checks, needed here, cause too much overhead.
*
* Thus we decompress the entire rest at once into the temporary
* buffer and change the decomp state to return the prepared
* data from the buffer by the more simple calls to
* pglz_get_next_decomp_char_from_plain().
* ----------
*/
if (dstate->cp_out - dstate->temp_buf >= 256)
{
unsigned char *cp_in = dstate->cp_in;
unsigned char *cp_out = dstate->cp_out;
unsigned char *cp_end = dstate->cp_end;
unsigned char *cp_copy;
unsigned char ctrl;
int off;
int len;
int i;
while (cp_in < cp_end)
{
ctrl = *cp_in++;
for (i = 0; i < 8; i++)
{
if (cp_in == cp_end)
break;
if (ctrl & 0x01)
{
len = (cp_in[0] & 0x0f) + 3;
off = ((cp_in[0] & 0xf0) << 4) | cp_in[1];
cp_in += 2;
if (len == 18)
len += *cp_in++;
cp_copy = cp_out - off;
while(len--)
*cp_out++ = *cp_copy++;
} else {
*cp_out++ = *cp_in++;
}
ctrl >>= 1;
}
}
dstate->cp_in = dstate->cp_out;
dstate->cp_end = cp_out;
dstate->next_char = pglz_get_next_decomp_char_from_plain;
return (int)(*(dstate->cp_in++));
}
/* ----------
* Not yet, get next control byte into decomp state.
* ----------
*/
dstate->ctrl = (unsigned char)(*(dstate->cp_in++));
dstate->ctrl_count = 8;
}
/* ----------
* Check for EOF in tag/literal byte data.
* ----------
*/
if (dstate->cp_in == dstate->cp_end)
{
return EOF;
}
/* ----------
* Handle next control bit.
* ----------
*/
dstate->ctrl_count--;
if (dstate->ctrl & 0x01)
{
/* ----------
* Bit is set, so tag is following. Setup copy information
* and do the copy for the first byte as above.
* ----------
*/
int off;
dstate->tocopy = (dstate->cp_in[0] & 0x0f) + 3;
off = ((dstate->cp_in[0] & 0xf0) << 4) | dstate->cp_in[1];
dstate->cp_in += 2;
if (dstate->tocopy == 18)
dstate->tocopy += *(dstate->cp_in++);
dstate->cp_copy = dstate->cp_out - off;
dstate->tocopy--;
retval = (*(dstate->cp_out++) = *(dstate->cp_copy++));
} else {
/* ----------
* Bit is unset, so literal byte follows.
* ----------
*/
retval = (int)(*(dstate->cp_out++) = *(dstate->cp_in++));
}
dstate->ctrl >>= 1;
return (int)retval;
}
/* ----------
* pglz_get_next_decomp_char_from_plain -
*
* The input data to pglz_decomp_init() was stored in uncompressed
* format. So we don't have a temporary output buffer and simply
* return bytes from the input until EOF.
* ----------
*/
int
pglz_get_next_decomp_char_from_plain(PGLZ_DecompState *dstate)
{
if (dstate->cp_in >= dstate->cp_end)
return EOF;
return (int)(*(dstate->cp_in++));
}

View File

@ -7,7 +7,7 @@
*
* Copyright (c) 1994, Regents of the University of California
*
* $Id: pg_amop.h,v 1.24 1999/09/29 21:13:30 wieck Exp $
* $Id: pg_amop.h,v 1.25 1999/11/25 01:28:05 wieck Exp $
*
* NOTES
* the genbki.sh script reads this file and generates .bki
@ -348,6 +348,16 @@ DATA(insert OID = 0 ( 403 1768 1752 3 btreesel btreenpage ));
DATA(insert OID = 0 ( 403 1768 1757 4 btreesel btreenpage ));
DATA(insert OID = 0 ( 403 1768 1756 5 btreesel btreenpage ));
/*
* nbtree lztext
*/
DATA(insert OID = 0 ( 403 1663 1659 1 btreesel btreenpage ));
DATA(insert OID = 0 ( 403 1663 1660 2 btreesel btreenpage ));
DATA(insert OID = 0 ( 403 1663 1657 3 btreesel btreenpage ));
DATA(insert OID = 0 ( 403 1663 1662 4 btreesel btreenpage ));
DATA(insert OID = 0 ( 403 1663 1661 5 btreesel btreenpage ));
/*
* hash table _ops
*/

View File

@ -9,7 +9,7 @@
*
* Copyright (c) 1994, Regents of the University of California
*
* $Id: pg_amproc.h,v 1.15 1999/09/29 21:13:30 wieck Exp $
* $Id: pg_amproc.h,v 1.16 1999/11/25 01:28:05 wieck Exp $
*
* NOTES
* the genbki.sh script reads this file and generates .bki
@ -97,6 +97,7 @@ DATA(insert OID = 0 (403 1313 1315 1));
DATA(insert OID = 0 (403 810 836 1));
DATA(insert OID = 0 (403 935 926 1));
DATA(insert OID = 0 (403 1768 1769 1));
DATA(insert OID = 0 (403 1663 1636 1));
/* hash */

View File

@ -7,7 +7,7 @@
*
* Copyright (c) 1994, Regents of the University of California
*
* $Id: pg_opclass.h,v 1.22 1999/11/23 04:47:39 momjian Exp $
* $Id: pg_opclass.h,v 1.23 1999/11/25 01:28:05 wieck Exp $
*
* NOTES
* the genbki.sh script reads this file and generates .bki
@ -117,5 +117,7 @@ DATA(insert OID = 652 ( cidr_ops 650 ));
DESCR("");
DATA(insert OID = 1768 ( numeric_ops 1700 ));
DESCR("");
DATA(insert OID = 1663 ( lztext_ops 1625 ));
DESCR("");
#endif /* PG_OPCLASS_H */

View File

@ -7,7 +7,7 @@
*
* Copyright (c) 1994, Regents of the University of California
*
* $Id: pg_operator.h,v 1.61 1999/11/22 17:56:37 momjian Exp $
* $Id: pg_operator.h,v 1.62 1999/11/25 01:28:05 wieck Exp $
*
* NOTES
* the genbki.sh script reads this file and generates .bki
@ -687,6 +687,14 @@ DATA(insert OID = 1761 ( "/" PGUID 0 b t f 1700 1700 1700 0 0 0 0 numeric
DATA(insert OID = 1762 ( "%" PGUID 0 b t f 1700 1700 1700 0 0 0 0 numeric_mod - - ));
DATA(insert OID = 1763 ( "@" PGUID 0 l t f 0 1700 1700 0 0 0 0 numeric_abs - - ));
/* LZTEXT type */
DATA(insert OID = 1657 ( "=" PGUID 0 b t f 1625 1625 16 1657 1658 1659 1659 lztext_eq eqsel eqjoinsel ));
DATA(insert OID = 1658 ( "<>" PGUID 0 b t f 1625 1625 16 1658 1657 0 0 lztext_ne neqsel neqjoinsel ));
DATA(insert OID = 1659 ( "<" PGUID 0 b t f 1625 1625 16 1661 1662 0 0 lztext_lt intltsel intltjoinsel ));
DATA(insert OID = 1660 ( "<=" PGUID 0 b t f 1625 1625 16 1662 1661 0 0 lztext_le intltsel intltjoinsel ));
DATA(insert OID = 1661 ( ">" PGUID 0 b t f 1625 1625 16 1659 1660 0 0 lztext_gt intgtsel intgtjoinsel ));
DATA(insert OID = 1662 ( ">=" PGUID 0 b t f 1625 1625 16 1660 1659 0 0 lztext_ge intgtsel intgtjoinsel ));
/*

View File

@ -6,7 +6,7 @@
*
* Copyright (c) 1994, Regents of the University of California
*
* $Id: pg_proc.h,v 1.106 1999/11/17 21:21:50 wieck Exp $
* $Id: pg_proc.h,v 1.107 1999/11/25 01:28:05 wieck Exp $
*
* NOTES
* The script catalog/genbki.sh reads this file and generates .bki
@ -2359,6 +2359,20 @@ DATA(insert OID = 1634 ( lztextoctetlen PGUID 11 f t t 1 f 23 "1625" 100 0
DESCR("octet length");
DATA(insert OID = 1635 ( octet_length PGUID 11 f t t 1 f 23 "1625" 100 0 1 0 lztextoctetlen - ));
DESCR("octet length");
DATA(insert OID = 1636 ( lztext_cmp PGUID 11 f t t 2 f 23 "1625 1625" 100 0 1 0 lztext_cmp - ));
DESCR("compare lztext vs. lztext");
DATA(insert OID = 1637 ( lztext_eq PGUID 11 f t t 2 f 16 "1625 1625" 100 0 1 0 lztext_eq - ));
DESCR("equal");
DATA(insert OID = 1638 ( lztext_ne PGUID 11 f t t 2 f 16 "1625 1625" 100 0 1 0 lztext_ne - ));
DESCR("not equal");
DATA(insert OID = 1639 ( lztext_gt PGUID 11 f t t 2 f 16 "1625 1625" 100 0 1 0 lztext_gt - ));
DESCR("greater-than");
DATA(insert OID = 1654 ( lztext_ge PGUID 11 f t t 2 f 16 "1625 1625" 100 0 1 0 lztext_ge - ));
DESCR("greater-than-or-equal");
DATA(insert OID = 1655 ( lztext_lt PGUID 11 f t t 2 f 16 "1625 1625" 100 0 1 0 lztext_lt - ));
DESCR("lower-than");
DATA(insert OID = 1656 ( lztext_le PGUID 11 f t t 2 f 16 "1625 1625" 100 0 1 0 lztext_le - ));
DESCR("lower-than-or-equal");
/*

View File

@ -6,7 +6,7 @@
*
* Copyright (c) 1994, Regents of the University of California
*
* $Id: builtins.h,v 1.90 1999/11/17 21:21:51 wieck Exp $
* $Id: builtins.h,v 1.91 1999/11/25 01:28:07 wieck Exp $
*
* NOTES
* This should normally only be included by fmgr.h.
@ -635,5 +635,12 @@ text *lztext_text(lztext *lz);
lztext *text_lztext(text *txt);
int32 lztextlen(lztext *lz);
int32 lztextoctetlen(lztext *lz);
int32 lztext_cmp(lztext *lz1, lztext *lz2);
bool lztext_eq(lztext *lz1, lztext *lz2);
bool lztext_ne(lztext *lz1, lztext *lz2);
bool lztext_gt(lztext *lz1, lztext *lz2);
bool lztext_ge(lztext *lz1, lztext *lz2);
bool lztext_lt(lztext *lz1, lztext *lz2);
bool lztext_le(lztext *lz1, lztext *lz2);
#endif /* BUILTINS_H */

View File

@ -1,7 +1,7 @@
/* ----------
* pg_lzcompress.h -
*
* $Header: /cvsroot/pgsql/src/include/utils/pg_lzcompress.h,v 1.2 1999/11/17 22:18:46 wieck Exp $
* $Header: /cvsroot/pgsql/src/include/utils/pg_lzcompress.h,v 1.3 1999/11/25 01:28:07 wieck Exp $
*
* Definitions for the builtin LZ compressor
* ----------
@ -110,6 +110,26 @@ typedef struct PGLZ_Strategy {
} PGLZ_Strategy;
/* ----------
* PGLZ_DecompState -
*
* Decompression state variable for byte-per-byte decompression
* using pglz_decomp_getchar() macro.
* ----------
*/
typedef struct PGLZ_DecompState {
unsigned char *temp_buf;
unsigned char *cp_in;
unsigned char *cp_end;
unsigned char *cp_out;
unsigned char *cp_copy;
int (*next_char)(struct PGLZ_DecompState *dstate);
int tocopy;
int ctrl_count;
unsigned char ctrl;
} PGLZ_DecompState;
/* ----------
* The standard strategies
*
@ -139,6 +159,55 @@ extern PGLZ_Strategy *PGLZ_strategy_allways;
extern PGLZ_Strategy *PGLZ_strategy_never;
/* ----------
* pglz_decomp_getchar -
*
* Get next character (or EOF) from decompressor.
* The status variable must be initialized before and deinitialized
* after compression with the next two macros below.
* ----------
*/
#define pglz_decomp_getchar(_ds) \
((*((_ds)->next_char))((_ds)))
/* ----------
* pglz_decomp_init -
*
* Initialize a decomp state from a compressed input.
* ----------
*/
#define pglz_decomp_init(_ds,_lz) { \
(_ds)->cp_in = ((unsigned char *)(_lz)) \
+ sizeof(PGLZ_Header); \
(_ds)->cp_end = (_ds)->cp_in + (_lz)->varsize \
- sizeof(PGLZ_Header); \
if (PGLZ_IS_COMPRESSED((_lz))) { \
(_ds)->temp_buf = (unsigned char *) \
palloc(PGLZ_RAW_SIZE((_lz))); \
(_ds)->cp_out = (_ds)->temp_buf; \
(_ds)->next_char = pglz_get_next_decomp_char_from_lzdata; \
(_ds)->tocopy = 0; \
(_ds)->ctrl_count = 0; \
} else { \
(_ds)->temp_buf = NULL; \
(_ds)->next_char = pglz_get_next_decomp_char_from_plain; \
} \
}
/* ----------
* pglz_decomp_end -
*
* Deallocate resources after decompression.
* ----------
*/
#define pglz_decomp_end(_ds) { \
if ((_ds)->temp_buf != NULL) \
pfree((void *)((_ds)->temp_buf)); \
}
/* ----------
* Global function declarations
* ----------
@ -148,5 +217,13 @@ int pglz_compress (char *source, int32 slen, PGLZ_Header *dest,
int pglz_decompress (PGLZ_Header *source, char *dest);
/* ----------
* Functions used by pglz_decomp_getchar().
* Internal use only.
* ----------
*/
extern int pglz_get_next_decomp_char_from_lzdata(PGLZ_DecompState *dstate);
extern int pglz_get_next_decomp_char_from_plain(PGLZ_DecompState *dstate);
#endif /* _PG_LZCOMPRESS_H_ */