arrayfuncs.c

关系型数据库 Postgresql 6.5.2

/*-------------------------------------------------------------------------
 *
 * arrayfuncs.c
 *	  Special functions for arrays.
 *
 * Copyright (c) 1994, Regents of the University of California
 *
 *
 * IDENTIFICATION
 *	  $Header: /usr/local/cvsroot/pgsql/src/backend/utils/adt/arrayfuncs.c,v 1.43.2.1 1999/08/02 05:24:49 scrappy Exp $
 *
 *-------------------------------------------------------------------------
 */

#include <ctype.h>

#include "postgres.h"


#include "catalog/catalog.h"
#include "catalog/pg_type.h"
#include "fmgr.h"
#include "libpq/be-fsstubs.h"
#include "libpq/libpq-fs.h"
#include "storage/fd.h"
#include "utils/array.h"
#include "utils/memutils.h"
#include "utils/syscache.h"

#define ASSGN	 "="

/* An array has the following internal structure:
 *	  <nbytes>		- total number of bytes
 *	  <ndim>		- number of dimensions of the array
 *	  <flags>		- bit mask of flags
 *	  <dim>			- size of each array axis
 *	  <dim_lower>	- lower boundary of each dimension
 *	  <actual data> - whatever is the stored data
 */

/*-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-*/
static int	_ArrayCount(char *str, int *dim, int typdelim);
static char *_ReadArrayStr(char *arrayStr, int nitems, int ndim, int *dim,
			  FmgrInfo *inputproc, Oid typelem, int32 typmod,
			  char typdelim, int typlen, bool typbyval,
			  char typalign, int *nbytes);

#ifdef LOARRAY
static char *_ReadLOArray(char *str, int *nbytes, int *fd, bool *chunkFlag,
			 int ndim, int *dim, int baseSize);

#endif
static void _CopyArrayEls(char **values, char *p, int nitems, int typlen,
			  char typalign, bool typbyval);
static void system_cache_lookup(Oid element_type, bool input, int *typlen,
				 bool *typbyval, char *typdelim, Oid *typelem, Oid *proc,
					char *typalign);
static Datum _ArrayCast(char *value, bool byval, int len);

#ifdef LOARRAY
static char *_AdvanceBy1word(char *str, char **word);

#endif
static void _ArrayRange(int *st, int *endp, int bsize, char *destPtr,
			ArrayType *array, int from);
static int	_ArrayClipCount(int *stI, int *endpI, ArrayType *array);
static void _LOArrayRange(int *st, int *endp, int bsize, int srcfd,
			  int destfd, ArrayType *array, int isSrcLO, bool *isNull);
static void _ReadArray(int *st, int *endp, int bsize, int srcfd, int destfd,
		   ArrayType *array, int isDestLO, bool *isNull);
static int	ArrayCastAndSet(char *src, bool typbyval, int typlen, char *dest);
static int	SanityCheckInput(int ndim, int n, int *dim, int *lb, int *indx);
static int	array_read(char *destptr, int eltsize, int nitems, char *srcptr);
static char *array_seek(char *ptr, int eltsize, int nitems);

/*---------------------------------------------------------------------
 * array_in :
 *		  converts an array from the external format in "string" to
 *		  it internal format.
 * return value :
 *		  the internal representation of the input array
 *--------------------------------------------------------------------
 */
char *
array_in(char *string,			/* input array in external form */
		 Oid element_type,		/* type OID of an array element */
		 int32 typmod)
{
	int			typlen;
	bool		typbyval,
				done;
	char		typdelim;
	Oid			typinput;
	Oid			typelem;
	char	   *string_save,
			   *p,
			   *q,
			   *r;
	FmgrInfo	inputproc;
	int			i,
				nitems;
	int32		nbytes;
	char	   *dataPtr;
	ArrayType  *retval = NULL;
	int			ndim,
				dim[MAXDIM],
				lBound[MAXDIM];
	char		typalign;

	system_cache_lookup(element_type, true, &typlen, &typbyval, &typdelim,
						&typelem, &typinput, &typalign);

	fmgr_info(typinput, &inputproc);

	string_save = (char *) palloc(strlen(string) + 3);
	strcpy(string_save, string);

	/* --- read array dimensions  ---------- */
	p = q = string_save;
	done = false;
	for (ndim = 0; !done;)
	{
		while (isspace(*p))
			p++;
		if (*p == '[')
		{
			p++;
			if ((r = (char *) strchr(p, ':')) == (char *) NULL)
				lBound[ndim] = 1;
			else
			{
				*r = '\0';
				lBound[ndim] = atoi(p);
				p = r + 1;
			}
			for (q = p; isdigit(*q); q++);
			if (*q != ']')
				elog(ERROR, "array_in: missing ']' in array declaration");
			*q = '\0';
			dim[ndim] = atoi(p);
			if ((dim[ndim] < 0) || (lBound[ndim] < 0))
				elog(ERROR, "array_in: array dimensions need to be positive");
			dim[ndim] = dim[ndim] - lBound[ndim] + 1;
			if (dim[ndim] < 0)
				elog(ERROR, "array_in: upper_bound cannot be < lower_bound");
			p = q + 1;
			ndim++;
		}
		else
			done = true;
	}

	if (ndim == 0)
	{
		if (*p == '{')
		{
			ndim = _ArrayCount(p, dim, typdelim);
			for (i = 0; i < ndim; lBound[i++] = 1);
		}
		else
			elog(ERROR, "array_in: Need to specify dimension");
	}
	else
	{
		while (isspace(*p))
			p++;
		if (strncmp(p, ASSGN, strlen(ASSGN)))
			elog(ERROR, "array_in: missing assignment operator");
		p += strlen(ASSGN);
		while (isspace(*p))
			p++;
	}

#ifdef ARRAYDEBUG
	printf("array_in- ndim %d (", ndim);
	for (i = 0; i < ndim; i++)
	{
		printf(" %d", dim[i]);
	};
	printf(") for %s\n", string);
#endif

	nitems = getNitems(ndim, dim);
	if (nitems == 0)
	{
		char	   *emptyArray = palloc(sizeof(ArrayType));

		MemSet(emptyArray, 0, sizeof(ArrayType));
		*(int32 *) emptyArray = sizeof(ArrayType);
		return emptyArray;
	}

	if (*p == '{')
	{
		/* array not a large object */
		dataPtr = (char *) _ReadArrayStr(p, nitems, ndim, dim, &inputproc, typelem,
							typmod, typdelim, typlen, typbyval, typalign,
										 &nbytes);
		nbytes += ARR_OVERHEAD(ndim);
		retval = (ArrayType *) palloc(nbytes);
		MemSet(retval, 0, nbytes);
		memmove(retval, (char *) &nbytes, sizeof(int));
		memmove((char *) ARR_NDIM_PTR(retval), (char *) &ndim, sizeof(int));
		SET_LO_FLAG(false, retval);
		memmove((char *) ARR_DIMS(retval), (char *) dim, ndim * sizeof(int));
		memmove((char *) ARR_LBOUND(retval), (char *) lBound,
				ndim * sizeof(int));

		/*
		 * dataPtr is an array of arbitraystuff even though its type is
		 * char* cast to char** to pass to _CopyArrayEls for now  - jolly
		 */
		_CopyArrayEls((char **) dataPtr,
					  ARR_DATA_PTR(retval), nitems,
					  typlen, typalign, typbyval);
	}
	else
	{
#ifdef LOARRAY
		int			dummy,
					bytes;
		bool		chunked = false;

		dataPtr = _ReadLOArray(p, &bytes, &dummy, &chunked, ndim,
							   dim, typlen);
		nbytes = bytes + ARR_OVERHEAD(ndim);
		retval = (ArrayType *) palloc(nbytes);
		MemSet(retval, 0, nbytes);
		memmove(retval, (char *) &nbytes, sizeof(int));
		memmove((char *) ARR_NDIM_PTR(retval), (char *) &ndim, sizeof(int));
		SET_LO_FLAG(true, retval);
		SET_CHUNK_FLAG(chunked, retval);
		memmove((char *) ARR_DIMS(retval), (char *) dim, ndim * sizeof(int));
		memmove((char *) ARR_LBOUND(retval), (char *) lBound, ndim * sizeof(int));
		memmove(ARR_DATA_PTR(retval), dataPtr, bytes);
#endif
		elog(ERROR, "large object arrays not supported");
	}
	pfree(string_save);
	return (char *) retval;
}

/*-----------------------------------------------------------------------------
 * _ArrayCount
 *	 Counts the number of dimensions and the *dim array for an array string.
 *		 The syntax for array input is C-like nested curly braces
 *-----------------------------------------------------------------------------
 */
static int
_ArrayCount(char *str, int *dim, int typdelim)
{
	int			nest_level = 0,
				i;
	int			ndim = 0,
				temp[MAXDIM];
	bool		scanning_string = false;
	bool		eoArray = false;
	char	   *q;

	for (i = 0; i < MAXDIM; ++i)
		temp[i] = dim[i] = 0;

	if (strncmp(str, "{}", 2) == 0)
		return 0;

	q = str;
	while (eoArray != true)
	{
		bool		done = false;

		while (!done)
		{
			switch (*q)
			{
				case '\\':
					/* skip escaped characters (\ and ") inside strings */
					if (scanning_string && *(q + 1))
						q++;
					break;
				case '\0':

					/*
					 * Signal a premature end of the string.  DZ -
					 * 2-9-1996
					 */
					elog(ERROR, "malformed array constant: %s", str);
					break;
				case '\"':
					scanning_string = !scanning_string;
					break;
				case '{':
					if (!scanning_string)
					{
						temp[nest_level] = 0;
						nest_level++;
					}
					break;
				case '}':
					if (!scanning_string)
					{
						if (!ndim)
							ndim = nest_level;
						nest_level--;
						if (nest_level)
							temp[nest_level - 1]++;
						if (nest_level == 0)
							eoArray = done = true;
					}
					break;
				default:
					if (!ndim)
						ndim = nest_level;
					if (*q == typdelim && !scanning_string)
						done = true;
					break;
			}
			if (!done)
				q++;
		}
		temp[ndim - 1]++;
		q++;
		if (!eoArray)
			while (isspace(*q))
				q++;
	}
	for (i = 0; i < ndim; ++i)
		dim[i] = temp[i];

	return ndim;
}

/*---------------------------------------------------------------------------
 * _ReadArrayStr :
 *	 parses the array string pointed by "arrayStr" and converts it in the
 *	 internal format. The external format expected is like C array
 *	 declaration. Unspecified elements are initialized to zero for fixed length
 *	 base types and to empty varlena structures for variable length base
 *	 types.
 * result :
 *	 returns the internal representation of the array elements
 *	 nbytes is set to the size of the array in its internal representation.
 *---------------------------------------------------------------------------
 */
static char *
_ReadArrayStr(char *arrayStr,
			  int nitems,
			  int ndim,
			  int *dim,
			  FmgrInfo *inputproc,		/* function used for the
										 * conversion */
			  Oid typelem,
			  int32 typmod,
			  char typdelim,
			  int typlen,
			  bool typbyval,
			  char typalign,
			  int *nbytes)
{
	int			i,
				nest_level = 0;
	char	   *p,
			   *q,
			   *r,
			  **values;
	bool		scanning_string = false;
	int			indx[MAXDIM],
				prod[MAXDIM];
	bool		eoArray = false;

	mda_get_prod(ndim, dim, prod);
	for (i = 0; i < ndim; indx[i++] = 0);
	/* read array enclosed within {} */
	values = (char **) palloc(nitems * sizeof(char *));
	MemSet(values, 0, nitems * sizeof(char *));
	q = p = arrayStr;

	while (!eoArray)
	{
		bool		done = false;
		int			i = -1;

		while (!done)
		{
			switch (*q)
			{
				case '\\':
					/* Crunch the string on top of the backslash. */
					for (r = q; *r != '\0'; r++)
						*r = *(r + 1);
					break;
				case '\"':
					if (!scanning_string)
					{
						while (p != q)
							p++;
						p++;	/* get p past first doublequote */
					}
					else
						*q = '\0';
					scanning_string = !scanning_string;
					break;
				case '{':
					if (!scanning_string)
					{
						p++;
						nest_level++;
						if (nest_level > ndim)
							elog(ERROR, "array_in: illformed array constant");
						indx[nest_level - 1] = 0;
						indx[ndim - 1] = 0;
					}
					break;
				case '}':
					if (!scanning_string)
					{
						if (i == -1)
							i = tuple2linear(ndim, indx, prod);
						nest_level--;
						if (nest_level == 0)
							eoArray = done = true;
						else
						{
							*q = '\0';
							indx[nest_level - 1]++;
						}
					}
					break;
				default:
					if (*q == typdelim && !scanning_string)
					{
						if (i == -1)
							i = tuple2linear(ndim, indx, prod);
						done = true;
						indx[ndim - 1]++;
					}
					break;
			}
			if (!done)
				q++;
		}
		*q = '\0';
		if (i >= nitems)
			elog(ERROR, "array_in: illformed array constant");
		values[i] = (*fmgr_faddr(inputproc)) (p, typelem, typmod);
		p = ++q;
		if (!eoArray)

			/*
			 * if not at the end of the array skip white space
			 */
			while (isspace(*q))
			{
				p++;
				q++;
			}
	}
	if (typlen > 0)
	{
		*nbytes = nitems * typlen;
		if (!typbyval)
			for (i = 0; i < nitems; i++)
				if (!values[i])
				{
					values[i] = palloc(typlen);
					MemSet(values[i], 0, typlen);
				}
	}
	else
	{
		for (i = 0, *nbytes = 0; i < nitems; i++)
		{
			if (values[i])
			{
				if (typalign == 'd')
					*nbytes += MAXALIGN(*(int32 *) values[i]);
				else
					*nbytes += INTALIGN(*(int32 *) values[i]);
			}
			else
			{
				*nbytes += sizeof(int32);
				values[i] = palloc(sizeof(int32));
				*(int32 *) values[i] = sizeof(int32);
			}
		}
	}
	return (char *) values;
}


/*----------------------------------------------------------------------------
 * Read data about an array to be stored as a large object
 *----------------------------------------------------------------------------
 */
#ifdef LOARRAY
static char *
_ReadLOArray(char *str,
			 int *nbytes,
			 int *fd,
			 bool *chunkFlag,
			 int ndim,
			 int *dim,
			 int baseSize)
{
	char	   *inputfile,
			   *accessfile = NULL,
			   *chunkfile = NULL;
	char	   *retStr,
			   *_AdvanceBy1word();
	Oid			lobjId;

	str = _AdvanceBy1word(str, &inputfile);

	while (str != NULL)
	{
		char	   *word;

		str = _AdvanceBy1word(str, &word);

		if (!strcmp(word, "-chunk"))
		{
			if (str == NULL)
				elog(ERROR, "array_in: access pattern file required");
			str = _AdvanceBy1word(str, &accessfile);
		}
		else if (!strcmp(word, "-noreorg"))
		{
			if (str == NULL)
				elog(ERROR, "array_in: chunk file required");
			str = _AdvanceBy1word(str, &chunkfile);
		}
		else
			elog(ERROR, "usage: <input file> -chunk DEFAULT/<access pattern file> -invert/-native [-noreorg <chunk file>]");
	}

	if (inputfile == NULL)
		elog(ERROR, "array_in: missing file name");
	lobjId = lo_creat(0);
	*fd = lo_open(lobjId, INV_READ);
	if (*fd < 0)
		elog(ERROR, "Large object create failed");
	retStr = inputfile;
	*nbytes = strlen(retStr) + 2;

	if (accessfile)
	{
		FILE	   *afd;

#ifndef __CYGWIN32__
		if ((afd = AllocateFile(accessfile, "r")) == NULL)
#else
		if ((afd = AllocateFile(accessfile, "r")) == NULL)
#endif
			elog(ERROR, "unable to open access pattern file");
		*chunkFlag = true;
		retStr = _ChunkArray(*fd, afd, ndim, dim, baseSize, nbytes,
							 chunkfile);
		FreeFile(afd);
	}
	return retStr;
}

#endif

static void
_CopyArrayEls(char **values,
			  char *p,
			  int nitems,
			  int typlen,
			  char typalign,
			  bool typbyval)
{
	int			i;

	for (i = 0; i < nitems; i++)
	{
		int			inc;

		inc = ArrayCastAndSet(values[i], typbyval, typlen, p);
		p += inc;
		if (!typbyval)
			pfree(values[i]);
	}
	pfree(values);
}

/*-------------------------------------------------------------------------
 * array_out :
 *		   takes the internal representation of an array and returns a string
 *		  containing the array in its external format.
 *-------------------------------------------------------------------------
 */
char *
array_out(ArrayType *v, Oid element_type)
{
	int			typlen;
	bool		typbyval;
	char		typdelim;
	Oid			typoutput,
				typelem;
	FmgrInfo	outputproc;
	char		typalign;

	char	   *p,
			   *tmp,
			   *retval,
			  **values,
				delim[2];
	int			nitems,
				overall_length,
				i,
				j,
				k,
#ifndef TCL_ARRAYS
				l,
#endif
				indx[MAXDIM];
	bool		dummy_bool;
	int			ndim,
			   *dim;

	if (v == (ArrayType *) NULL)
		return (char *) NULL;

	if (ARR_IS_LO(v) == true)
	{
		char	   *p,
				   *save_p;
		int			nbytes;

		/* get a wide string to print to */
		p = array_dims(v, &dummy_bool);