varlena.c

postgresql8.3.4源码,开源数据库

/*-------------------------------------------------------------------------
 *
 * varlena.c
 *	  Functions for the variable-length built-in types.
 *
 * Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
 * Portions Copyright (c) 1994, Regents of the University of California
 *
 *
 * IDENTIFICATION
 *	  $PostgreSQL: pgsql/src/backend/utils/adt/varlena.c,v 1.162.2.1 2008/03/13 18:32:02 tgl Exp $
 *
 *-------------------------------------------------------------------------
 */
#include "postgres.h"

#include <ctype.h>

#include "access/tupmacs.h"
#include "access/tuptoaster.h"
#include "catalog/pg_type.h"
#include "libpq/md5.h"
#include "libpq/pqformat.h"
#include "miscadmin.h"
#include "parser/scansup.h"
#include "regex/regex.h"
#include "utils/builtins.h"
#include "utils/lsyscache.h"
#include "utils/pg_locale.h"


typedef struct varlena unknown;

typedef struct
{
	bool		use_wchar;		/* T if multibyte encoding */
	char	   *str1;			/* use these if not use_wchar */
	char	   *str2;			/* note: these point to original texts */
	pg_wchar   *wstr1;			/* use these if use_wchar */
	pg_wchar   *wstr2;			/* note: these are palloc'd */
	int			len1;			/* string lengths in logical characters */
	int			len2;
} TextPositionState;

#define DatumGetUnknownP(X)			((unknown *) PG_DETOAST_DATUM(X))
#define DatumGetUnknownPCopy(X)		((unknown *) PG_DETOAST_DATUM_COPY(X))
#define PG_GETARG_UNKNOWN_P(n)		DatumGetUnknownP(PG_GETARG_DATUM(n))
#define PG_GETARG_UNKNOWN_P_COPY(n) DatumGetUnknownPCopy(PG_GETARG_DATUM(n))
#define PG_RETURN_UNKNOWN_P(x)		PG_RETURN_POINTER(x)

#define PG_TEXTARG_GET_STR(arg_) \
	DatumGetCString(DirectFunctionCall1(textout, PG_GETARG_DATUM(arg_)))
#define PG_TEXT_GET_STR(textp_) \
	DatumGetCString(DirectFunctionCall1(textout, PointerGetDatum(textp_)))
#define PG_STR_GET_TEXT(str_) \
	DatumGetTextP(DirectFunctionCall1(textin, CStringGetDatum(str_)))
#define TEXTLEN(textp) \
	text_length(PointerGetDatum(textp))

static int	text_cmp(text *arg1, text *arg2);
static int32 text_length(Datum str);
static int	text_position(text *t1, text *t2);
static void text_position_setup(text *t1, text *t2, TextPositionState *state);
static int	text_position_next(int start_pos, TextPositionState *state);
static void text_position_cleanup(TextPositionState *state);
static text *text_substring(Datum str,
			   int32 start,
			   int32 length,
			   bool length_not_specified);

static void appendStringInfoText(StringInfo str, const text *t);


/*****************************************************************************
 *	 USER I/O ROUTINES														 *
 *****************************************************************************/


#define VAL(CH)			((CH) - '0')
#define DIG(VAL)		((VAL) + '0')

/*
 *		byteain			- converts from printable representation of byte array
 *
 *		Non-printable characters must be passed as '\nnn' (octal) and are
 *		converted to internal form.  '\' must be passed as '\\'.
 *		ereport(ERROR, ...) if bad form.
 *
 *		BUGS:
 *				The input is scanned twice.
 *				The error checking of input is minimal.
 */
Datum
byteain(PG_FUNCTION_ARGS)
{
	char	   *inputText = PG_GETARG_CSTRING(0);
	char	   *tp;
	char	   *rp;
	int			byte;
	bytea	   *result;

	for (byte = 0, tp = inputText; *tp != '\0'; byte++)
	{
		if (tp[0] != '\\')
			tp++;
		else if ((tp[0] == '\\') &&
				 (tp[1] >= '0' && tp[1] <= '3') &&
				 (tp[2] >= '0' && tp[2] <= '7') &&
				 (tp[3] >= '0' && tp[3] <= '7'))
			tp += 4;
		else if ((tp[0] == '\\') &&
				 (tp[1] == '\\'))
			tp += 2;
		else
		{
			/*
			 * one backslash, not followed by 0 or ### valid octal
			 */
			ereport(ERROR,
					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
					 errmsg("invalid input syntax for type bytea")));
		}
	}

	byte += VARHDRSZ;
	result = (bytea *) palloc(byte);
	SET_VARSIZE(result, byte);

	tp = inputText;
	rp = VARDATA(result);
	while (*tp != '\0')
	{
		if (tp[0] != '\\')
			*rp++ = *tp++;
		else if ((tp[0] == '\\') &&
				 (tp[1] >= '0' && tp[1] <= '3') &&
				 (tp[2] >= '0' && tp[2] <= '7') &&
				 (tp[3] >= '0' && tp[3] <= '7'))
		{
			byte = VAL(tp[1]);
			byte <<= 3;
			byte += VAL(tp[2]);
			byte <<= 3;
			*rp++ = byte + VAL(tp[3]);
			tp += 4;
		}
		else if ((tp[0] == '\\') &&
				 (tp[1] == '\\'))
		{
			*rp++ = '\\';
			tp += 2;
		}
		else
		{
			/*
			 * We should never get here. The first pass should not allow it.
			 */
			ereport(ERROR,
					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
					 errmsg("invalid input syntax for type bytea")));
		}
	}

	PG_RETURN_BYTEA_P(result);
}

/*
 *		byteaout		- converts to printable representation of byte array
 *
 *		Non-printable characters are inserted as '\nnn' (octal) and '\' as
 *		'\\'.
 *
 *		NULL vlena should be an error--returning string with NULL for now.
 */
Datum
byteaout(PG_FUNCTION_ARGS)
{
	bytea	   *vlena = PG_GETARG_BYTEA_PP(0);
	char	   *result;
	char	   *vp;
	char	   *rp;
	int			val;			/* holds unprintable chars */
	int			i;
	int			len;

	len = 1;					/* empty string has 1 char */
	vp = VARDATA_ANY(vlena);
	for (i = VARSIZE_ANY_EXHDR(vlena); i != 0; i--, vp++)
	{
		if (*vp == '\\')
			len += 2;
		else if ((unsigned char) *vp < 0x20 || (unsigned char) *vp > 0x7e)
			len += 4;
		else
			len++;
	}
	rp = result = (char *) palloc(len);
	vp = VARDATA_ANY(vlena);
	for (i = VARSIZE_ANY_EXHDR(vlena); i != 0; i--, vp++)
	{
		if (*vp == '\\')
		{
			*rp++ = '\\';
			*rp++ = '\\';
		}
		else if ((unsigned char) *vp < 0x20 || (unsigned char) *vp > 0x7e)
		{
			val = *vp;
			rp[0] = '\\';
			rp[3] = DIG(val & 07);
			val >>= 3;
			rp[2] = DIG(val & 07);
			val >>= 3;
			rp[1] = DIG(val & 03);
			rp += 4;
		}
		else
			*rp++ = *vp;
	}
	*rp = '\0';
	PG_RETURN_CSTRING(result);
}

/*
 *		bytearecv			- converts external binary format to bytea
 */
Datum
bytearecv(PG_FUNCTION_ARGS)
{
	StringInfo	buf = (StringInfo) PG_GETARG_POINTER(0);
	bytea	   *result;
	int			nbytes;

	nbytes = buf->len - buf->cursor;
	result = (bytea *) palloc(nbytes + VARHDRSZ);
	SET_VARSIZE(result, nbytes + VARHDRSZ);
	pq_copymsgbytes(buf, VARDATA(result), nbytes);
	PG_RETURN_BYTEA_P(result);
}

/*
 *		byteasend			- converts bytea to binary format
 *
 * This is a special case: just copy the input...
 */
Datum
byteasend(PG_FUNCTION_ARGS)
{
	bytea	   *vlena = PG_GETARG_BYTEA_P_COPY(0);

	PG_RETURN_BYTEA_P(vlena);
}


/*
 *		textin			- converts "..." to internal representation
 */
Datum
textin(PG_FUNCTION_ARGS)
{
	char	   *inputText = PG_GETARG_CSTRING(0);
	text	   *result;
	int			len;

	len = strlen(inputText);
	result = (text *) palloc(len + VARHDRSZ);
	SET_VARSIZE(result, len + VARHDRSZ);

	memcpy(VARDATA(result), inputText, len);

	PG_RETURN_TEXT_P(result);
}

/*
 *		textout			- converts internal representation to "..."
 */
Datum
textout(PG_FUNCTION_ARGS)
{
	text	   *t = PG_GETARG_TEXT_PP(0);
	int			len;
	char	   *result;

	len = VARSIZE_ANY_EXHDR(t);
	result = (char *) palloc(len + 1);
	memcpy(result, VARDATA_ANY(t), len);
	result[len] = '\0';

	PG_RETURN_CSTRING(result);
}

/*
 *		textrecv			- converts external binary format to text
 */
Datum
textrecv(PG_FUNCTION_ARGS)
{
	StringInfo	buf = (StringInfo) PG_GETARG_POINTER(0);
	text	   *result;
	char	   *str;
	int			nbytes;

	str = pq_getmsgtext(buf, buf->len - buf->cursor, &nbytes);

	result = (text *) palloc(nbytes + VARHDRSZ);
	SET_VARSIZE(result, nbytes + VARHDRSZ);
	memcpy(VARDATA(result), str, nbytes);
	pfree(str);
	PG_RETURN_TEXT_P(result);
}

/*
 *		textsend			- converts text to binary format
 */
Datum
textsend(PG_FUNCTION_ARGS)
{
	text	   *t = PG_GETARG_TEXT_PP(0);
	StringInfoData buf;

	pq_begintypsend(&buf);
	pq_sendtext(&buf, VARDATA_ANY(t), VARSIZE_ANY_EXHDR(t));
	PG_RETURN_BYTEA_P(pq_endtypsend(&buf));
}


/*
 *		unknownin			- converts "..." to internal representation
 */
Datum
unknownin(PG_FUNCTION_ARGS)
{
	char	   *str = PG_GETARG_CSTRING(0);

	/* representation is same as cstring */
	PG_RETURN_CSTRING(pstrdup(str));
}

/*
 *		unknownout			- converts internal representation to "..."
 */
Datum
unknownout(PG_FUNCTION_ARGS)
{
	/* representation is same as cstring */
	char	   *str = PG_GETARG_CSTRING(0);

	PG_RETURN_CSTRING(pstrdup(str));
}

/*
 *		unknownrecv			- converts external binary format to unknown
 */
Datum
unknownrecv(PG_FUNCTION_ARGS)
{
	StringInfo	buf = (StringInfo) PG_GETARG_POINTER(0);
	char	   *str;
	int			nbytes;

	str = pq_getmsgtext(buf, buf->len - buf->cursor, &nbytes);
	/* representation is same as cstring */
	PG_RETURN_CSTRING(str);
}

/*
 *		unknownsend			- converts unknown to binary format
 */
Datum
unknownsend(PG_FUNCTION_ARGS)
{
	/* representation is same as cstring */
	char	   *str = PG_GETARG_CSTRING(0);
	StringInfoData buf;

	pq_begintypsend(&buf);
	pq_sendtext(&buf, str, strlen(str));
	PG_RETURN_BYTEA_P(pq_endtypsend(&buf));
}


/* ========== PUBLIC ROUTINES ========== */

/*
 * textlen -
 *	  returns the logical length of a text*
 *	   (which is less than the VARSIZE of the text*)
 */
Datum
textlen(PG_FUNCTION_ARGS)
{
	Datum		str = PG_GETARG_DATUM(0);

	/* try to avoid decompressing argument */
	PG_RETURN_INT32(text_length(str));
}

/*
 * text_length -
 *	Does the real work for textlen()
 *
 *	This is broken out so it can be called directly by other string processing
 *	functions.	Note that the argument is passed as a Datum, to indicate that
 *	it may still be in compressed form.  We can avoid decompressing it at all
 *	in some cases.
 */
static int32
text_length(Datum str)
{
	/* fastpath when max encoding length is one */
	if (pg_database_encoding_max_length() == 1)
		PG_RETURN_INT32(toast_raw_datum_size(str) - VARHDRSZ);
	else
	{
		text	   *t = DatumGetTextPP(str);

		PG_RETURN_INT32(pg_mbstrlen_with_len(VARDATA_ANY(t),
											 VARSIZE_ANY_EXHDR(t)));
	}
}

/*
 * textoctetlen -
 *	  returns the physical length of a text*
 *	   (which is less than the VARSIZE of the text*)
 */
Datum
textoctetlen(PG_FUNCTION_ARGS)
{
	Datum		str = PG_GETARG_DATUM(0);

	/* We need not detoast the input at all */
	PG_RETURN_INT32(toast_raw_datum_size(str) - VARHDRSZ);
}

/*
 * textcat -
 *	  takes two text* and returns a text* that is the concatenation of
 *	  the two.
 *
 * Rewritten by Sapa, [e-mail]. 8-Jul-96.
 * Updated by Thomas, [e-mail] 1997-07-10.
 * Allocate space for output in all cases.
 * XXX - thomas 1997-07-10
 */
Datum
textcat(PG_FUNCTION_ARGS)
{
	text	   *t1 = PG_GETARG_TEXT_PP(0);
	text	   *t2 = PG_GETARG_TEXT_PP(1);
	int			len1,
				len2,
				len;
	text	   *result;
	char	   *ptr;

	len1 = VARSIZE_ANY_EXHDR(t1);
	if (len1 < 0)
		len1 = 0;

	len2 = VARSIZE_ANY_EXHDR(t2);
	if (len2 < 0)
		len2 = 0;

	len = len1 + len2 + VARHDRSZ;
	result = (text *) palloc(len);

	/* Set size of result string... */
	SET_VARSIZE(result, len);

	/* Fill data field of result string... */
	ptr = VARDATA(result);
	if (len1 > 0)
		memcpy(ptr, VARDATA_ANY(t1), len1);
	if (len2 > 0)
		memcpy(ptr + len1, VARDATA_ANY(t2), len2);

	PG_RETURN_TEXT_P(result);
}

/*
 * charlen_to_bytelen()
 *	Compute the number of bytes occupied by n characters starting at *p
 *
 * It is caller's responsibility that there actually are n characters;
 * the string need not be null-terminated.
 */
static int
charlen_to_bytelen(const char *p, int n)
{
	if (pg_database_encoding_max_length() == 1)
	{
		/* Optimization for single-byte encodings */
		return n;
	}
	else
	{
		const char *s;

		for (s = p; n > 0; n--)
			s += pg_mblen(s);

		return s - p;
	}
}

/*
 * text_substr()
 * Return a substring starting at the specified position.
 * - thomas 1997-12-31
 *
 * Input:
 *	- string
 *	- starting position (is one-based)
 *	- string length
 *
 * If the starting position is zero or less, then return from the start of the string
 *	adjusting the length to be consistent with the "negative start" per SQL92.
 * If the length is less than zero, return the remaining string.
 *
 * Added multibyte support.
 * - Tatsuo Ishii 1998-4-21
 * Changed behavior if starting position is less than one to conform to SQL92 behavior.
 * Formerly returned the entire string; now returns a portion.
 * - Thomas Lockhart 1998-12-10
 * Now uses faster TOAST-slicing interface
 * - John Gray 2002-02-22
 * Remove "#ifdef MULTIBYTE" and test for encoding_max_length instead. Change
 * behaviors conflicting with SQL92 to meet SQL92 (if E = S + L < S throw
 * error; if E < 1, return '', not entire string). Fixed MB related bug when
 * S > LC and < LC + 4 sometimes garbage characters are returned.
 * - Joe Conway 2002-08-10
 */
Datum
text_substr(PG_FUNCTION_ARGS)
{
	PG_RETURN_TEXT_P(text_substring(PG_GETARG_DATUM(0),
									PG_GETARG_INT32(1),
									PG_GETARG_INT32(2),
									false));
}

/*
 * text_substr_no_len -
 *	  Wrapper to avoid opr_sanity failure due to
 *	  one function accepting a different number of args.
 */
Datum
text_substr_no_len(PG_FUNCTION_ARGS)
{
	PG_RETURN_TEXT_P(text_substring(PG_GETARG_DATUM(0),
									PG_GETARG_INT32(1),
									-1, true));
}

/*
 * text_substring -
 *	Does the real work for text_substr() and text_substr_no_len()
 *
 *	This is broken out so it can be called directly by other string processing
 *	functions.	Note that the argument is passed as a Datum, to indicate that
 *	it may still be in compressed/toasted form.  We can avoid detoasting all
 *	of it in some cases.
 *
 *	The result is always a freshly palloc'd datum.
 */
static text *
text_substring(Datum str, int32 start, int32 length, bool length_not_specified)
{
	int32		eml = pg_database_encoding_max_length();
	int32		S = start;		/* start position */
	int32		S1;				/* adjusted start position */
	int32		L1;				/* adjusted substring length */

	/* life is easy if the encoding max length is 1 */
	if (eml == 1)