/tags/2.3.0/Drivers/Postgre7.1/parse.c

/* Module:          parse.c
 *
 * Description:     This module contains routines related to parsing SQL statements.
 *                  This can be useful for two reasons:
 *
 *                  1. So the query does not actually have to be executed to return data about it
 *
 *                  2. To be able to return information about precision, nullability, aliases, etc.
 *                     in the functions SQLDescribeCol and SQLColAttributes.  Currently, Postgres
 *                     doesn't return any information about these things in a query.
 *
 * Classes:         none
 *
 * API functions:   none
 *
 * Comments:        See "notice.txt" for copyright and license information.
 *
 */

#include <config.h>

#include <stdio.h>
#include <stdio.h>
#include <string.h>
#include <ctype.h>

#include "statement.h"
#include "connection.h"
#include "qresult.h"
#include "pgtypes.h"

#ifndef WIN32
#ifndef HAVE_STRICMP
#define stricmp(s1,s2) 		strcasecmp(s1,s2)
#define strnicmp(s1,s2,n)	strncasecmp(s1,s2,n)
#endif
#endif

#define FLD_INCR	32
#define TAB_INCR	8
#define COL_INCR	16

char *getNextToken(char *s, char *token, int smax, char *delim, char *quote, char *dquote, char *numeric);
void getColInfo(COL_INFO *col_info, FIELD_INFO *fi, int k);
char searchColInfo(COL_INFO *col_info, FIELD_INFO *fi);

char *
getNextToken(char *s, char *token, int smax, char *delim, char *quote, char *dquote, char *numeric)
{
int i = 0;
int out = 0;
char qc, in_escape = FALSE;

	if (smax <= 1)
		return NULL;

	smax--;

	/* skip leading delimiters */
	while (isspace((unsigned char) s[i]) || s[i] == ',') {
		/* mylog("skipping '%c'\n", s[i]); */
		i++;
	}

	if (s[0] == '\0') {
		token[0] = '\0';
		return NULL;
	}

	if (quote) *quote = FALSE;
	if (dquote) *dquote = FALSE;
	if (numeric) *numeric = FALSE;

	/* get the next token */
	while ( ! isspace((unsigned char) s[i]) && s[i] != ',' &&
			s[i] != '\0' && out != smax) {

		/*	Handle quoted stuff */
		if ( out == 0 && (s[i] == '\"' || s[i] == '\'')) {
			qc = s[i];
			if (qc == '\"') {
				if (dquote) *dquote = TRUE;
			}
			if (qc == '\'') {
				if (quote) *quote = TRUE;
			}

			i++;		/* dont return the quote */
			while (s[i] != '\0' && out != smax) {  
				if (s[i] == qc && ! in_escape) {
					break;
				}
				if (s[i] == '\\' && ! in_escape) {
					in_escape = TRUE;
				}
				else {
					in_escape = FALSE;	
					token[out++] = s[i];
				}
				i++;
			}
			if (s[i] == qc)
				i++;
			break;
		}

		/*	Check for numeric literals */
		if ( out == 0 && isdigit((unsigned char) s[i])) {
			if (numeric) *numeric = TRUE;
			token[out++] = s[i++];
			while ( isalnum((unsigned char) s[i]) || s[i] == '.')
				token[out++] = s[i++];

			break;
		}

		if ( ispunct((unsigned char) s[i]) && s[i] != '_') {
			mylog("got ispunct: s[%d] = '%c'\n", i, s[i]);

			if (out == 0) {
				token[out++] = s[i++];
				break;
			}
			else
				break;
		}

		if (out != smax)
			token[out++] = s[i];

		i++;
	}

	/* mylog("done -- s[%d] = '%c'\n", i, s[i]); */

	token[out] = '\0';

	/*	find the delimiter  */
	while ( isspace((unsigned char) s[i]))
		i++;

	/*	return the most priority delimiter */
	if (s[i] == ',')  {
		if (delim) *delim = s[i];
	}
	else if (s[i] == '\0') {
		if (delim) *delim = '\0';
	}
	else  {
		if (delim) *delim = ' ';
	}

	/* skip trailing blanks  */
	while ( isspace((unsigned char) s[i])) {
		i++;
	}

	return &s[i];
}


#if 0
QR_set_num_fields(stmt->result, 14);
QR_set_field_info(stmt->result, 0, "TABLE_QUALIFIER", PG_TYPE_TEXT, MAX_INFO_STRING);
QR_set_field_info(stmt->result, 1, "TABLE_OWNER", PG_TYPE_TEXT, MAX_INFO_STRING);
QR_set_field_info(stmt->result, 2, "TABLE_NAME", PG_TYPE_TEXT, MAX_INFO_STRING);
QR_set_field_info(stmt->result, 3, "COLUMN_NAME", PG_TYPE_TEXT, MAX_INFO_STRING);
QR_set_field_info(stmt->result, 4, "DATA_TYPE", PG_TYPE_INT2, 2);
QR_set_field_info(stmt->result, 5, "TYPE_NAME", PG_TYPE_TEXT, MAX_INFO_STRING);
QR_set_field_info(stmt->result, 6, "PRECISION", PG_TYPE_INT4, 4);
QR_set_field_info(stmt->result, 7, "LENGTH", PG_TYPE_INT4, 4);
QR_set_field_info(stmt->result, 8, "SCALE", PG_TYPE_INT2, 2);
QR_set_field_info(stmt->result, 9, "RADIX", PG_TYPE_INT2, 2);
QR_set_field_info(stmt->result, 10, "NULLABLE", PG_TYPE_INT2, 2);
QR_set_field_info(stmt->result, 11, "REMARKS", PG_TYPE_TEXT, 254);
/*	User defined fields */
QR_set_field_info(stmt->result, 12, "DISPLAY_SIZE", PG_TYPE_INT4, 4);
QR_set_field_info(stmt->result, 13, "FIELD_TYPE", PG_TYPE_INT4, 4);
#endif

void
getColInfo(COL_INFO *col_info, FIELD_INFO *fi, int k)
{
	if (fi->name[0] == '\0')
		strcpy(fi->name, QR_get_value_manual(col_info->result, k, 3));

	fi->type = atoi( QR_get_value_manual(col_info->result, k, 13));
	fi->precision = atoi( QR_get_value_manual(col_info->result, k, 6));
	fi->length = atoi( QR_get_value_manual(col_info->result, k, 7));
	fi->nullable = atoi( QR_get_value_manual(col_info->result, k, 10));
	fi->display_size = atoi( QR_get_value_manual(col_info->result, k, 12));
}

char
searchColInfo(COL_INFO *col_info, FIELD_INFO *fi)
{
int k;
char *col;


	for (k = 0; k < QR_get_num_tuples(col_info->result); k++) {
		col = QR_get_value_manual(col_info->result, k, 3);
		if ( ! strcmp(col, fi->name)) {
			getColInfo(col_info, fi, k);

			mylog("PARSE: searchColInfo: \n");
			return TRUE;
		}
	}

	return FALSE;
}


char
parse_statement(StatementClass *stmt)
{
static char* const func="parse_statement";
char token[256];
char delim, quote, dquote, numeric, unquoted;
char *ptr;
char in_select = FALSE, in_distinct = FALSE, in_on = FALSE, in_from = FALSE, in_where = FALSE, in_table = FALSE;
char in_field = FALSE, in_expr = FALSE, in_func = FALSE, in_dot = FALSE, in_as = FALSE;
int j, i, k = 0, n, blevel = 0;
FIELD_INFO **fi;
TABLE_INFO **ti;
char parse;
ConnectionClass *conn = stmt->hdbc;
HSTMT hcol_stmt;
StatementClass *col_stmt;
RETCODE result;


	mylog("%s: entering...\n", func);

	ptr = stmt->statement;
	fi = stmt->fi;
	ti = stmt->ti;

	stmt->nfld = 0;
	stmt->ntab = 0;

	while ((ptr = getNextToken(ptr, token, sizeof(token), &delim, &quote, &dquote, &numeric)) != NULL) {

		unquoted = ! ( quote || dquote );

		mylog("unquoted=%d, quote=%d, dquote=%d, numeric=%d, delim='%c', token='%s', ptr='%s'\n", unquoted, quote, dquote, numeric, delim, token, ptr);

		if ( unquoted && ! stricmp(token, "select")) {
			in_select = TRUE;

			mylog("SELECT\n");
			continue;
		}

		if ( unquoted && in_select && ! stricmp(token, "distinct")) {
			in_distinct = TRUE;

			mylog("DISTINCT\n");
			continue;
		}

		if ( unquoted && ! stricmp(token, "into")) {
			in_select = FALSE;

			mylog("INTO\n");
			continue;
		}

		if ( unquoted && ! stricmp(token, "from")) {
			in_select = FALSE;
			in_from = TRUE;

			mylog("FROM\n");
			continue;
		}

		if ( unquoted && (! stricmp(token, "where") ||
			! stricmp(token, "union") || 
			! stricmp(token, "order") || 
			! stricmp(token, "group") || 
			! stricmp(token, "having"))) {

			in_select = FALSE;
			in_from = FALSE;
			in_where = TRUE;

			mylog("WHERE...\n");
			break;
		}

		if (in_select) {

			if ( in_distinct) {
				mylog("in distinct\n");

				if (unquoted && ! stricmp(token, "on")) {
					in_on = TRUE;
					mylog("got on\n");
					continue;
				}
				if (in_on) {
					in_distinct = FALSE;
					in_on = FALSE;
					continue;		/*	just skip the unique on field */
				}
				mylog("done distinct\n");
				in_distinct = FALSE;
			}

			if ( in_expr || in_func) {		/* just eat the expression */
				mylog("in_expr=%d or func=%d\n", in_expr, in_func);
				if (quote || dquote)
					continue;

				if (in_expr && blevel == 0 && delim == ',') {
					mylog("**** in_expr and Got comma\n");
					in_expr = FALSE;
					in_field = FALSE;
				}

				else if (token[0] == '(') {
					blevel++;
					mylog("blevel++ = %d\n", blevel);
				}
				else if (token[0] == ')') {
					blevel--;
					mylog("blevel-- = %d\n", blevel);
					if (delim==',') {
						in_func = FALSE;
						in_expr = FALSE;
						in_field = FALSE;
					}
				}
				continue;
			}

			if ( ! in_field) {

				if ( ! token[0])
					continue;

				if ( ! (stmt->nfld % FLD_INCR)) {
					mylog("reallocing at nfld=%d\n", stmt->nfld);
					fi = (FIELD_INFO **) realloc(fi, (stmt->nfld + FLD_INCR) * sizeof(FIELD_INFO *));
					if ( ! fi) {
						stmt->parse_status = STMT_PARSE_FATAL;
						return FALSE;
					}
					stmt->fi = fi;
				}

				fi[stmt->nfld] = (FIELD_INFO *) malloc( sizeof(FIELD_INFO));
				if (fi[stmt->nfld] == NULL) {
					stmt->parse_status = STMT_PARSE_FATAL;
					return FALSE;
				}

				/*	Initialize the field info */
				memset(fi[stmt->nfld], 0, sizeof(FIELD_INFO));

				/*	double quotes are for qualifiers */
				if (dquote)
					fi[stmt->nfld]->dquote = TRUE;

				if (quote) {
					fi[stmt->nfld++]->quote = TRUE;
					continue;
				}
				else if (numeric) {
					mylog("**** got numeric: nfld = %d\n", stmt->nfld);
					fi[stmt->nfld]->numeric = TRUE;
				}
				else if (token[0] == '(') {		/* expression */
					mylog("got EXPRESSION\n");
					fi[stmt->nfld++]->expr = TRUE;
					in_expr = TRUE;
					blevel = 1;
					continue;
				}

				else {
					strcpy(fi[stmt->nfld]->name, token);
					fi[stmt->nfld]->dot[0] = '\0';
				}
				mylog("got field='%s', dot='%s'\n", fi[stmt->nfld]->name, fi[stmt->nfld]->dot);

				if (delim == ',') {
					mylog("comma (1)\n");
				}
				else {
					in_field = TRUE;
				}
				stmt->nfld++;
				continue;
			}

			/**************************/
			/*	We are in a field now */
			/**************************/
			if (in_dot) {
				stmt->nfld--;
				strcpy(fi[stmt->nfld]->dot, fi[stmt->nfld]->name);
				strcpy(fi[stmt->nfld]->name, token);
				stmt->nfld++;
				in_dot = FALSE;

				if (delim == ',') {
					mylog("in_dot: got comma\n");
					in_field = FALSE;
				}

				continue;
			}

			if (in_as) {
				stmt->nfld--;
				strcpy(fi[stmt->nfld]->alias, token);
				mylog("alias for field '%s' is '%s'\n", fi[stmt->nfld]->name, fi[stmt->nfld]->alias);
				in_as = FALSE;
				in_field = FALSE;

				stmt->nfld++;

				if (delim == ',') {
					mylog("comma(2)\n");
				}
				continue;
			}

			/*	Function */
			if (token[0] == '(') {
				in_func = TRUE;
				blevel = 1;
				fi[stmt->nfld-1]->func = TRUE;
				/*	name will have the function name -- maybe useful some day */
				mylog("**** got function = '%s'\n", fi[stmt->nfld-1]->name);
				continue;
			}

			if (token[0] == '.') {
				in_dot = TRUE;	
				mylog("got dot\n");
				continue;
			}

			if ( ! stricmp(token, "as")) {
				in_as = TRUE;
				mylog("got AS\n");
				continue;
			}

			/*	otherwise, it's probably an expression */
			in_expr = TRUE;
			fi[stmt->nfld-1]->expr = TRUE;
			fi[stmt->nfld-1]->name[0] = '\0';
			mylog("*** setting expression\n");

		}

		if (in_from) {

			if ( ! in_table) {
				if ( ! token[0])
					continue;

				if ( ! (stmt->ntab % TAB_INCR)) {
					ti = (TABLE_INFO **) realloc(ti, (stmt->ntab + TAB_INCR) * sizeof(TABLE_INFO *));
					if ( ! ti) {
						stmt->parse_status = STMT_PARSE_FATAL;
						return FALSE;
					}
					stmt->ti = ti;
				}
				ti[stmt->ntab] = (TABLE_INFO *) malloc(sizeof(TABLE_INFO));
				if (ti[stmt->ntab] == NULL) {
					stmt->parse_status = STMT_PARSE_FATAL;
					return FALSE;
				}

				ti[stmt->ntab]->alias[0] = '\0';

				strcpy(ti[stmt->ntab]->name, token);
				mylog("got table = '%s'\n", ti[stmt->ntab]->name);

				if (delim == ',') {
					mylog("more than 1 tables\n");
				}
				else {
					in_table = TRUE;
				}
				stmt->ntab++;
				continue;
			}

			strcpy(ti[stmt->ntab-1]->alias, token);
			mylog("alias for table '%s' is '%s'\n", ti[stmt->ntab-1]->name, ti[stmt->ntab-1]->alias);
			in_table = FALSE;
			if (delim == ',') {
				mylog("more than 1 tables\n");
			}
		}
	}


	/*************************************************/
	/*	Resolve any possible field names with tables */
	/*************************************************/

	parse = TRUE;

	/*	Resolve field names with tables */
	for (i = 0; i < stmt->nfld; i++) {

		if (fi[i]->func || fi[i]->expr || fi[i]->numeric) {
			fi[i]->ti = NULL;
			fi[i]->type = -1;
			parse = FALSE;
			continue;
		}

		else if (fi[i]->quote) {		/* handle as text */
			fi[i]->ti = NULL;
			fi[i]->type = PG_TYPE_TEXT;
			fi[i]->precision = 0;
			continue;
		} 

		/*	it's a dot, resolve to table or alias */
		else if (fi[i]->dot[0]) {
			for (k = 0; k < stmt->ntab; k++) {
				if ( ! stricmp(ti[k]->name, fi[i]->dot)) {
					fi[i]->ti = ti[k];
					break;
				}
				else if ( ! stricmp(ti[k]->alias, fi[i]->dot)) {
					fi[i]->ti = ti[k];
					break;
				}
			}
		}
		else if (stmt->ntab == 1)
			fi[i]->ti = ti[0];
	}

	mylog("--------------------------------------------\n");
	mylog("nfld=%d, ntab=%d\n", stmt->nfld, stmt->ntab);

	for (i=0; i < stmt->nfld; i++) {
		mylog("Field %d:  expr=%d, func=%d, quote=%d, dquote=%d, numeric=%d, name='%s', alias='%s', dot='%s'\n", i, fi[i]->expr, fi[i]->func, fi[i]->quote, fi[i]->dquote, fi[i]->numeric, fi[i]->name, fi[i]->alias, fi[i]->dot);
		if (fi[i]->ti)
		mylog("     ----> table_name='%s', table_alias='%s'\n", fi[i]->ti->name, fi[i]->ti->alias);
	}

	for (i=0; i < stmt->ntab; i++) {
		mylog("Table %d: name='%s', alias='%s'\n", i, ti[i]->name, ti[i]->alias);
	}


	/******************************************************/
	/*	Now save the SQLColumns Info for the parse tables */
	/******************************************************/


	/*	Call SQLColumns for each table and store the result */
	for (i = 0; i < stmt->ntab; i++) {

		/*	See if already got it */
		char found = FALSE;
		for (k = 0; k < conn->ntables; k++) {
			if ( ! stricmp(conn->col_info[k]->name, ti[i]->name)) {
				mylog("FOUND col_info table='%s'\n", ti[i]->name);
				found = TRUE;
				break;
			}
		}
				
		if ( ! found) {

			mylog("PARSE: Getting SQLColumns for table[%d]='%s'\n", i, ti[i]->name);

			result = PG_SQLAllocStmt( stmt->hdbc, &hcol_stmt);
			if((result != SQL_SUCCESS) && (result != SQL_SUCCESS_WITH_INFO)) {
				SC_set_error(stmt, STMT_NO_MEMORY_ERROR, "SQLAllocStmt failed in parse_statement for columns.");
				stmt->parse_status = STMT_PARSE_FATAL;
				return FALSE;
			}

			col_stmt = (StatementClass *) hcol_stmt;
			col_stmt->internal = TRUE;

			result = PG_SQLColumns(hcol_stmt, "", 0, "", 0, 
						ti[i]->name, (SWORD) strlen(ti[i]->name), "", 0);
		
			mylog("        Past SQLColumns\n");
			if (result == SQL_SUCCESS) {
				mylog("      Success\n");
				if ( ! (conn->ntables % COL_INCR)) {

					mylog("PARSE: Allocing col_info at ntables=%d\n", conn->ntables);

					conn->col_info = (COL_INFO **) realloc(conn->col_info, (conn->ntables + COL_INCR) * sizeof(COL_INFO *));
					if ( ! conn->col_info) {
						stmt->parse_status = STMT_PARSE_FATAL;
						return FALSE;
					}
				}

				mylog("PARSE: malloc at conn->col_info[%d]\n", conn->ntables);
				conn->col_info[conn->ntables] = (COL_INFO *) malloc(sizeof(COL_INFO));
				if ( ! conn->col_info[conn->ntables]) {
					stmt->parse_status = STMT_PARSE_FATAL;
					return FALSE;
				}

				/*	Store the table name and the SQLColumns result structure */
				strcpy(conn->col_info[conn->ntables]->name, ti[i]->name);
				conn->col_info[conn->ntables]->result = col_stmt->result;

				/*	The connection will now free the result structures, so make
					sure that the statement doesn't free it 
				*/
				col_stmt->result = NULL;

				conn->ntables++;

				PG_SQLFreeStmt(hcol_stmt, SQL_DROP);
				mylog("Created col_info table='%s', ntables=%d\n", ti[i]->name, conn->ntables);
			}
			else {
				PG_SQLFreeStmt(hcol_stmt, SQL_DROP);
				break;
			}
		}

		/*	Associate a table from the statement with a SQLColumn info */
		ti[i]->col_info = conn->col_info[k];
		mylog("associate col_info: i=%d, k=%d\n", i, k);
	}


	mylog("Done SQLColumns\n");

	/******************************************************/
	/*	Now resolve the fields to point to column info    */
	/******************************************************/



	for (i = 0; i < stmt->nfld;) {

		/*	Dont worry about functions or quotes */
		if (fi[i]->func || fi[i]->quote || fi[i]->numeric) {
			i++;
			continue;
		}

		/*	Stars get expanded to all fields in the table */
		else if (fi[i]->name[0] == '*') {

			char do_all_tables;
			int total_cols, old_alloc, new_size, cols;
			int increased_cols;

			mylog("expanding field %d\n", i);

			total_cols = 0;	

			if (fi[i]->ti)	/*	The star represents only the qualified table */
				total_cols = QR_get_num_tuples(fi[i]->ti->col_info->result);

			else {	/*	The star represents all tables */

				/*	Calculate the total number of columns after expansion */
				for (k = 0; k < stmt->ntab; k++) {
					total_cols += QR_get_num_tuples(ti[k]->col_info->result);
				}
			}
			increased_cols = total_cols - 1;

			/*	Allocate some more field pointers if necessary */
			/*------------------------------------------------------------- */
			old_alloc = ((stmt->nfld  - 1) / FLD_INCR + 1) * FLD_INCR;
			new_size = stmt->nfld + increased_cols;

			mylog("k=%d, increased_cols=%d, old_alloc=%d, new_size=%d\n", k,increased_cols,old_alloc,new_size);

			if (new_size > old_alloc) {
				int new_alloc = ((new_size / FLD_INCR) + 1) * FLD_INCR;
				mylog("need more cols: new_alloc = %d\n", new_alloc);
				fi = (FIELD_INFO **) realloc(fi, new_alloc * sizeof(FIELD_INFO *));
				if ( ! fi) {
					stmt->parse_status = STMT_PARSE_FATAL;
					return FALSE;
				}
				stmt->fi = fi;
			}

			/*------------------------------------------------------------- */
			/*	copy any other fields (if there are any) up past the expansion */
			for (j = stmt->nfld - 1; j > i; j--) {
				mylog("copying field %d to %d\n", j, increased_cols + j);
				fi[increased_cols + j] = fi[j];
			}
			mylog("done copying fields\n");

			/*------------------------------------------------------------- */
			/*	Set the new number of fields */
			stmt->nfld += increased_cols;
			mylog("stmt->nfld now at %d\n", stmt->nfld);


			/*------------------------------------------------------------- */
			/*	copy the new field info */


			do_all_tables = (fi[i]->ti ? FALSE : TRUE);

			for (k = 0; k < (do_all_tables ? stmt->ntab : 1); k++) {

				TABLE_INFO *the_ti = do_all_tables ? ti[k] : fi[i]->ti;

				cols = QR_get_num_tuples(the_ti->col_info->result);

				for (n = 0; n < cols; n++) {
					mylog("creating field info: n=%d\n", n);
					/* skip malloc (already did it for the Star) */
					if (k > 0 || n > 0) {
						mylog("allocating field info at %d\n", n + i);
						fi[n + i] = (FIELD_INFO *) malloc( sizeof(FIELD_INFO));
						if (fi[n + i] == NULL) {
							stmt->parse_status = STMT_PARSE_FATAL;
							return FALSE;
						}
					}
					/*	Initialize the new space (or the * field) */
					memset(fi[n + i], 0, sizeof(FIELD_INFO));
					fi[n + i]->ti = the_ti;

					mylog("about to copy at %d\n", n + i);

					getColInfo(the_ti->col_info, fi[n + i], n);

					mylog("done copying\n");
				}

				i += cols;
				mylog("i now at %d\n", i);
			}

			/*------------------------------------------------------------- */
		}

		/*	We either know which table the field was in because it was qualified 
			with a table name or alias -OR- there was only 1 table.
		*/
		else if (fi[i]->ti) {

			if ( ! searchColInfo(fi[i]->ti->col_info, fi[i]))
				parse = FALSE;

			i++;
		}

		/*	Don't know the table -- search all tables in "from" list */
		else {
			parse = FALSE;
			for (k = 0; k < stmt->ntab; k++) {
				if ( searchColInfo(ti[k]->col_info, fi[i])) {
					fi[i]->ti = ti[k];	/* now know the table */
					parse = TRUE;
					break;
				}
			}
			i++;
		}
	}


	if ( ! parse)
		stmt->parse_status = STMT_PARSE_INCOMPLETE;
	else
		stmt->parse_status = STMT_PARSE_COMPLETE;


	mylog("done parse_statement: parse=%d, parse_status=%d\n", parse, stmt->parse_status);
	return parse;
}