varbit.c

PostgreSQL 8.1.4的源码 适用于Linux下的开源数据库系统

/*-------------------------------------------------------------------------
 *
 * varbit.c
 *	  Functions for the SQL datatypes BIT() and BIT VARYING().
 *
 * Code originally contributed by Adriaan Joubert.
 *
 * Portions Copyright (c) 1996-2005, PostgreSQL Global Development Group
 * Portions Copyright (c) 1994, Regents of the University of California
 *
 * IDENTIFICATION
 *	  $PostgreSQL: pgsql/src/backend/utils/adt/varbit.c,v 1.47 2005/10/15 02:49:30 momjian Exp $
 *
 *-------------------------------------------------------------------------
 */

#include "postgres.h"

#include "catalog/pg_type.h"
#include "libpq/pqformat.h"
#include "utils/array.h"
#include "utils/fmgroids.h"
#include "utils/memutils.h"
#include "utils/varbit.h"

#define HEXDIG(z)	 ((z)<10 ? ((z)+'0') : ((z)-10+'A'))


/*----------
 *	attypmod -- contains the length of the bit string in bits, or for
 *			   varying bits the maximum length.
 *
 *	The data structure contains the following elements:
 *	  header  -- length of the whole data structure (incl header)
 *				 in bytes. (as with all varying length datatypes)
 *	  data section -- private data section for the bits data structures
 *		bitlength -- length of the bit string in bits
 *		bitdata   -- bit string, most significant byte first
 *
 *	The length of the bitdata vector should always be exactly as many
 *	bytes as are needed for the given bitlength.  If the bitlength is
 *	not a multiple of 8, the extra low-order padding bits of the last
 *	byte must be zeroes.
 *----------
 */

/*
 * bit_in -
 *	  converts a char string to the internal representation of a bitstring.
 *		  The length is determined by the number of bits required plus
 *		  VARHDRSZ bytes or from atttypmod.
 */
Datum
bit_in(PG_FUNCTION_ARGS)
{
	char	   *input_string = PG_GETARG_CSTRING(0);

#ifdef NOT_USED
	Oid			typelem = PG_GETARG_OID(1);
#endif
	int32		atttypmod = PG_GETARG_INT32(2);
	VarBit	   *result;			/* The resulting bit string			  */
	char	   *sp;				/* pointer into the character string  */
	bits8	   *r;				/* pointer into the result */
	int			len,			/* Length of the whole data structure */
				bitlen,			/* Number of bits in the bit string   */
				slen;			/* Length of the input string		  */
	bool		bit_not_hex;	/* false = hex string  true = bit string */
	int			bc;
	bits8		x = 0;

	/* Check that the first character is a b or an x */
	if (input_string[0] == 'b' || input_string[0] == 'B')
	{
		bit_not_hex = true;
		sp = input_string + 1;
	}
	else if (input_string[0] == 'x' || input_string[0] == 'X')
	{
		bit_not_hex = false;
		sp = input_string + 1;
	}
	else
	{
		/*
		 * Otherwise it's binary.  This allows things like cast('1001' as bit)
		 * to work transparently.
		 */
		bit_not_hex = true;
		sp = input_string;
	}

	slen = strlen(sp);
	/* Determine bitlength from input string */
	if (bit_not_hex)
		bitlen = slen;
	else
		bitlen = slen * 4;

	/*
	 * Sometimes atttypmod is not supplied. If it is supplied we need to make
	 * sure that the bitstring fits.
	 */
	if (atttypmod <= 0)
		atttypmod = bitlen;
	else if (bitlen != atttypmod)
		ereport(ERROR,
				(errcode(ERRCODE_STRING_DATA_LENGTH_MISMATCH),
				 errmsg("bit string length %d does not match type bit(%d)",
						bitlen, atttypmod)));

	len = VARBITTOTALLEN(atttypmod);
	/* set to 0 so that *r is always initialised and string is zero-padded */
	result = (VarBit *) palloc0(len);
	VARATT_SIZEP(result) = len;
	VARBITLEN(result) = atttypmod;

	r = VARBITS(result);
	if (bit_not_hex)
	{
		/* Parse the bit representation of the string */
		/* We know it fits, as bitlen was compared to atttypmod */
		x = BITHIGH;
		for (; *sp; sp++)
		{
			if (*sp == '1')
				*r |= x;
			else if (*sp != '0')
				ereport(ERROR,
						(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
						 errmsg("\"%c\" is not a valid binary digit",
								*sp)));

			x >>= 1;
			if (x == 0)
			{
				x = BITHIGH;
				r++;
			}
		}
	}
	else
	{
		/* Parse the hex representation of the string */
		for (bc = 0; *sp; sp++)
		{
			if (*sp >= '0' && *sp <= '9')
				x = (bits8) (*sp - '0');
			else if (*sp >= 'A' && *sp <= 'F')
				x = (bits8) (*sp - 'A') + 10;
			else if (*sp >= 'a' && *sp <= 'f')
				x = (bits8) (*sp - 'a') + 10;
			else
				ereport(ERROR,
						(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
						 errmsg("\"%c\" is not a valid hexadecimal digit",
								*sp)));

			if (bc)
			{
				*r++ |= x;
				bc = 0;
			}
			else
			{
				*r = x << 4;
				bc = 1;
			}
		}
	}

	PG_RETURN_VARBIT_P(result);
}


Datum
bit_out(PG_FUNCTION_ARGS)
{
#if 1
	/* same as varbit output */
	return varbit_out(fcinfo);
#else
/* This is how one would print a hex string, in case someone wants to
   write a formatting function. */
	VarBit	   *s = PG_GETARG_VARBIT_P(0);
	char	   *result,
			   *r;
	bits8	   *sp;
	int			i,
				len,
				bitlen;

	bitlen = VARBITLEN(s);
	len = (bitlen + 3) / 4;
	result = (char *) palloc(len + 2);
	sp = VARBITS(s);
	r = result;
	*r++ = 'X';
	/* we cheat by knowing that we store full bytes zero padded */
	for (i = 0; i < len; i += 2, sp++)
	{
		*r++ = HEXDIG((*sp) >> 4);
		*r++ = HEXDIG((*sp) & 0xF);
	}

	/*
	 * Go back one step if we printed a hex number that was not part of the
	 * bitstring anymore
	 */
	if (i > len)
		r--;
	*r = '\0';

	PG_RETURN_CSTRING(result);
#endif
}

/*
 *		bit_recv			- converts external binary format to bit
 */
Datum
bit_recv(PG_FUNCTION_ARGS)
{
	StringInfo	buf = (StringInfo) PG_GETARG_POINTER(0);

#ifdef NOT_USED
	Oid			typelem = PG_GETARG_OID(1);
#endif
	int32		atttypmod = PG_GETARG_INT32(2);
	VarBit	   *result;
	int			len,
				bitlen;
	int			ipad;
	bits8		mask;

	bitlen = pq_getmsgint(buf, sizeof(int32));
	if (bitlen < 0)
		ereport(ERROR,
				(errcode(ERRCODE_INVALID_BINARY_REPRESENTATION),
				 errmsg("invalid length in external bit string")));

	/*
	 * Sometimes atttypmod is not supplied. If it is supplied we need to make
	 * sure that the bitstring fits.
	 */
	if (atttypmod > 0 && bitlen != atttypmod)
		ereport(ERROR,
				(errcode(ERRCODE_STRING_DATA_LENGTH_MISMATCH),
				 errmsg("bit string length %d does not match type bit(%d)",
						bitlen, atttypmod)));

	len = VARBITTOTALLEN(bitlen);
	result = (VarBit *) palloc(len);
	VARATT_SIZEP(result) = len;
	VARBITLEN(result) = bitlen;

	pq_copymsgbytes(buf, (char *) VARBITS(result), VARBITBYTES(result));

	/* Make sure last byte is zero-padded if needed */
	ipad = VARBITPAD(result);
	if (ipad > 0)
	{
		mask = BITMASK << ipad;
		*(VARBITS(result) + VARBITBYTES(result) - 1) &= mask;
	}

	PG_RETURN_VARBIT_P(result);
}

/*
 *		bit_send			- converts bit to binary format
 */
Datum
bit_send(PG_FUNCTION_ARGS)
{
	/* Exactly the same as varbit_send, so share code */
	return varbit_send(fcinfo);
}

/* bit()
 * Converts a bit() type to a specific internal length.
 * len is the bitlength specified in the column definition.
 *
 * If doing implicit cast, raise error when source data is wrong length.
 * If doing explicit cast, silently truncate or zero-pad to specified length.
 */
Datum
bit(PG_FUNCTION_ARGS)
{
	VarBit	   *arg = PG_GETARG_VARBIT_P(0);
	int32		len = PG_GETARG_INT32(1);
	bool		isExplicit = PG_GETARG_BOOL(2);
	VarBit	   *result;
	int			rlen;
	int			ipad;
	bits8		mask;

	/* No work if typmod is invalid or supplied data matches it already */
	if (len <= 0 || len == VARBITLEN(arg))
		PG_RETURN_VARBIT_P(arg);

	if (!isExplicit)
		ereport(ERROR,
				(errcode(ERRCODE_STRING_DATA_LENGTH_MISMATCH),
				 errmsg("bit string length %d does not match type bit(%d)",
						VARBITLEN(arg), len)));

	rlen = VARBITTOTALLEN(len);
	/* set to 0 so that string is zero-padded */
	result = (VarBit *) palloc0(rlen);
	VARATT_SIZEP(result) = rlen;
	VARBITLEN(result) = len;

	memcpy(VARBITS(result), VARBITS(arg),
		   Min(VARBITBYTES(result), VARBITBYTES(arg)));

	/*
	 * Make sure last byte is zero-padded if needed.  This is useless but safe
	 * if source data was shorter than target length (we assume the last byte
	 * of the source data was itself correctly zero-padded).
	 */
	ipad = VARBITPAD(result);
	if (ipad > 0)
	{
		mask = BITMASK << ipad;
		*(VARBITS(result) + VARBITBYTES(result) - 1) &= mask;
	}

	PG_RETURN_VARBIT_P(result);
}

/*
 * varbit_in -
 *	  converts a string to the internal representation of a bitstring.
 *		This is the same as bit_in except that atttypmod is taken as
 *		the maximum length, not the exact length to force the bitstring to.
 */
Datum
varbit_in(PG_FUNCTION_ARGS)
{
	char	   *input_string = PG_GETARG_CSTRING(0);

#ifdef NOT_USED
	Oid			typelem = PG_GETARG_OID(1);
#endif
	int32		atttypmod = PG_GETARG_INT32(2);
	VarBit	   *result;			/* The resulting bit string			  */
	char	   *sp;				/* pointer into the character string  */
	bits8	   *r;				/* pointer into the result */
	int			len,			/* Length of the whole data structure */
				bitlen,			/* Number of bits in the bit string   */
				slen;			/* Length of the input string		  */
	bool		bit_not_hex;	/* false = hex string  true = bit string */
	int			bc;
	bits8		x = 0;

	/* Check that the first character is a b or an x */
	if (input_string[0] == 'b' || input_string[0] == 'B')
	{
		bit_not_hex = true;
		sp = input_string + 1;
	}
	else if (input_string[0] == 'x' || input_string[0] == 'X')
	{
		bit_not_hex = false;
		sp = input_string + 1;
	}
	else
	{
		bit_not_hex = true;
		sp = input_string;
	}

	slen = strlen(sp);
	/* Determine bitlength from input string */
	if (bit_not_hex)
		bitlen = slen;
	else
		bitlen = slen * 4;

	/*
	 * Sometimes atttypmod is not supplied. If it is supplied we need to make
	 * sure that the bitstring fits.
	 */
	if (atttypmod <= 0)
		atttypmod = bitlen;
	else if (bitlen > atttypmod)
		ereport(ERROR,
				(errcode(ERRCODE_STRING_DATA_RIGHT_TRUNCATION),
				 errmsg("bit string too long for type bit varying(%d)",
						atttypmod)));

	len = VARBITTOTALLEN(bitlen);
	/* set to 0 so that *r is always initialised and string is zero-padded */
	result = (VarBit *) palloc0(len);
	VARATT_SIZEP(result) = len;
	VARBITLEN(result) = Min(bitlen, atttypmod);

	r = VARBITS(result);
	if (bit_not_hex)
	{
		/* Parse the bit representation of the string */
		/* We know it fits, as bitlen was compared to atttypmod */
		x = BITHIGH;
		for (; *sp; sp++)
		{
			if (*sp == '1')
				*r |= x;
			else if (*sp != '0')
				ereport(ERROR,
						(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
						 errmsg("\"%c\" is not a valid binary digit",
								*sp)));

			x >>= 1;
			if (x == 0)
			{
				x = BITHIGH;
				r++;
			}
		}
	}
	else
	{
		/* Parse the hex representation of the string */
		for (bc = 0; *sp; sp++)
		{
			if (*sp >= '0' && *sp <= '9')
				x = (bits8) (*sp - '0');
			else if (*sp >= 'A' && *sp <= 'F')
				x = (bits8) (*sp - 'A') + 10;
			else if (*sp >= 'a' && *sp <= 'f')
				x = (bits8) (*sp - 'a') + 10;
			else
				ereport(ERROR,
						(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
						 errmsg("\"%c\" is not a valid hexadecimal digit",
								*sp)));

			if (bc)
			{
				*r++ |= x;
				bc = 0;
			}
			else
			{
				*r = x << 4;
				bc = 1;
			}
		}
	}

	PG_RETURN_VARBIT_P(result);
}

/* varbit_out -
 *	  Prints the string as bits to preserve length accurately
 */
Datum
varbit_out(PG_FUNCTION_ARGS)
{
	VarBit	   *s = PG_GETARG_VARBIT_P(0);
	char	   *result,
			   *r;
	bits8	   *sp;
	bits8		x;
	int			i,
				k,
				len;

	len = VARBITLEN(s);
	result = (char *) palloc(len + 1);
	sp = VARBITS(s);
	r = result;
	for (i = 0; i < len - BITS_PER_BYTE; i += BITS_PER_BYTE, sp++)
	{
		x = *sp;
		for (k = 0; k < BITS_PER_BYTE; k++)
		{
			*r++ = (x & BITHIGH) ? '1' : '0';
			x <<= 1;
		}
	}
	x = *sp;
	for (k = i; k < len; k++)
	{
		*r++ = (x & BITHIGH) ? '1' : '0';
		x <<= 1;
	}
	*r = '\0';

	PG_RETURN_CSTRING(result);
}

/*
 *		varbit_recv			- converts external binary format to varbit