Support empty string handling for date and time datatypes (#3807)

Add TSQL-compatible handling of empty string ('') for date and time types, defaulting to '1900-01-01' for date and time, similar to datetime, datetime2, datetimeoffset, smalldatetime.

Changes:
1. Add helper functions to initialize the default date and time variables
   and to check for strings with whitespaces.
2. Modify varchar2date() and varchar2time() to handle empty/whitespace
   casts from varchar types to date and time types.
3. Update tsql_coerce_string_literal_hook to handle empty/whitespace
   casts from string literals to date and time datatypes.
4. Default empty/whitespace strings to '1900-01-01' for date and
   '00:00:00.0000000' for time.
5. Updated existing date-time related test cases. Added more tests cases
   for wider test coverage.

Task: BABEL-3433
Signed-off-by: Manisha Deshpande <mmdeshp@amazon.com>

Author: manisha-deshpande

contrib/babelfishpg_common/src/babelfishpg_common.c

@@ -216,6 +216,8 @@ get_common_utility_plugin(void)
 		common_utility_plugin_var.TsqlUTF8LengthInUTF16 = &TsqlUTF8LengthInUTF16;
 		common_utility_plugin_var.TsqlUTF8toUTF16StringInfo = &TsqlUTF8toUTF16StringInfo;
 		common_utility_plugin_var.tsql_numeric_get_typmod = &tsql_numeric_get_typmod;
+		common_utility_plugin_var.initializeToDefaultDate = &initializeToDefaultDate;
+		common_utility_plugin_var.initializeToDefaultTime = &initializeToDefaultTime;
 	}
 	return &common_utility_plugin_var;
 }

contrib/babelfishpg_common/src/babelfishpg_common.h

@@ -2,6 +2,7 @@
 
 #include "fmgr.h"
 #include "numeric.h"
+#include "utils/date.h"
 
 /*
  * Casting float < -1.0 to unsigned integer could cause issues on ARM.
@@ -95,4 +96,6 @@ typedef struct common_utility_plugin
 	int			(*TsqlUTF8LengthInUTF16) (const void *vin, int len);
 	void		(*TsqlUTF8toUTF16StringInfo) (StringInfo utf16_data, const void *data, size_t len);
 	int32_t		(*tsql_numeric_get_typmod) (Numeric num);
+	DateADT		(*initializeToDefaultDate) (void);
+	TimeADT		(*initializeToDefaultTime) (int32 typmod);
 } common_utility_plugin;

contrib/babelfishpg_common/src/datetime.c

@@ -544,7 +544,7 @@ datetime_in_str(char *str, Node *escontext)
 	 * Set input to default '1900-01-01 00:00:00.000' if empty string
 	 * encountered
 	 */
-	if (*str == '\0')
+	if (isEmptyOrWhitespace(str))
 	{
 		result = initializeToDefaultDatetime();
 		PG_RETURN_TIMESTAMP(result);
@@ -1201,6 +1201,62 @@ initializeToDefaultDatetime(void)
 	return result;
 }
 
+/*
+ * Set input to default '1900-01-01' if empty string encountered
+ */
+DateADT
+initializeToDefaultDate(void)
+{
+	DateADT date;
+	struct pg_tm tt,
+				 *tm = &tt;
+	tm->tm_year = 1900;
+	tm->tm_mon = 1;
+	tm->tm_mday = 1;
+	date = date2j(tm->tm_year, tm->tm_mon, tm->tm_mday) - POSTGRES_EPOCH_JDATE;
+
+	return date;
+}
+
+/*
+ * Set input to default '00:00:00.0000000' if empty string encountered
+ */
+TimeADT
+initializeToDefaultTime(int32 typmod)
+{
+	TimeADT	time;
+	struct pg_tm tt,
+				 *tm = &tt;
+	tm->tm_hour = tm->tm_min = tm->tm_sec = 0;
+	tm2time(tm, 0, &time);
+	AdjustTimeForTypmod(&time, typmod);
+
+	return time;
+}
+
+
+/*
+ * Check if string is empty or contains only whitespace
+ * Returns true if string is empty or contains only whitespace characters
+ */
+bool
+isEmptyOrWhitespace(const char *str)
+{
+	size_t i;
+	size_t len;
+
+	if (!str)
+		return true;
+
+	len = strlen(str);
+	for (i = 0; i < len; i++)
+	{
+		if (!isspace((unsigned char)str[i]))
+			return false;
+	}
+	return true;
+}
+
 Datum
 datetime_pl_datetime(PG_FUNCTION_ARGS)
 {

contrib/babelfishpg_common/src/datetime.h

@@ -9,6 +9,7 @@
 #define PLTSQL_DATETIME_H
 
 #include "datatype/timestamp.h"
+#include "utils/date.h"
 
 /*	Round off to MAX_DATETIME_PRECISION decimal places. */
 #define DT_PREC_INV 1000
@@ -33,6 +34,10 @@
 #define TSQL_DEFAULT_DATETIME	INT64CONST(-3155673600000000)
 
 extern Timestamp initializeToDefaultDatetime(void);
+extern DateADT initializeToDefaultDate(void);
+extern TimeADT initializeToDefaultTime(int32 typmod);
+extern bool isEmptyOrWhitespace(const char *str);
+
 /** Utility function to calculate days from '1900-01-01 00:00:00' */
 extern double calculateDaysFromDefaultDatetime(Timestamp timestamp_left); 
 extern int roundFractionalSeconds(int fractseconds); 

contrib/babelfishpg_common/src/datetime2.c

@@ -274,7 +274,7 @@ datetime2_in_str(char *str, int32 typmod, Node *escontext)
 	tm->tm_mday = 0;
 
 	/* Set input to default '1900-01-01 00:00:00.* if empty string encountered */
-	if (*str == '\0')
+	if (isEmptyOrWhitespace(str))
 	{
 		result = initializeToDefaultDatetime();
 		AdjustDatetime2ForTypmod(&result, typmod);

contrib/babelfishpg_common/src/datetimeoffset.c

@@ -114,7 +114,7 @@ datetimeoffset_in(PG_FUNCTION_ARGS)
 	 * Set input to default '1900-01-01 00:00:00.* 00:00' if empty string
 	 * encountered
 	 */
-	if (*str == '\0')
+	if (isEmptyOrWhitespace(str))
 	{
 		tsql_ts = initializeToDefaultDatetime();
 		AdjustDatetimeoffsetForTypmod(&tsql_ts, typmod);

contrib/babelfishpg_common/src/smalldatetime.c

@@ -79,7 +79,7 @@ smalldatetime_in_str(char *str, Node *escontext)
 	DateTimeErrorExtra extra;
 
 	/* Set input to default '1900-01-01 00:00:00' if empty string encountered */
-	if (*str == '\0')
+	if (isEmptyOrWhitespace(str))
 	{
 		result = initializeToDefaultDatetime();
 		AdjustTimestampForSmallDatetime(&result);

contrib/babelfishpg_common/src/varchar.c

@@ -40,6 +40,7 @@
 #include "utils/uuid.h"
 #include "utils/timestamp.h"
 #include "utils/numeric.h"
+#include "datetime.h"
 #include "typecode.h"
 #include "varchar.h"
 
@@ -1062,7 +1063,14 @@ varchar2date(PG_FUNCTION_ARGS)
 	DateADT		date;
 
 	str = varchar2cstring(source);
-	date = DatumGetDateADT(DirectFunctionCall1(date_in, CStringGetDatum(str)));
+
+	/*
+	 * Set input to default '1900-01-01' if empty string encountered
+	 */
+	if (isEmptyOrWhitespace(str))
+		date = initializeToDefaultDate();
+	else
+		date = DatumGetDateADT(DirectFunctionCall1(date_in, CStringGetDatum(str)));
 	pfree(str);
 	PG_RETURN_DATEADT(date);
 }
@@ -1079,7 +1087,14 @@ varchar2time(PG_FUNCTION_ARGS)
 		typmod = PG_GETARG_INT32(1);
 
 	str = varchar2cstring(source);
-	time = DatumGetTimeADT(DirectFunctionCall3(time_in, CStringGetDatum(str), InvalidOid, typmod));
+
+	/*
+	 * Set input to default '00:00:00.0000000' if empty string encountered
+	 */
+	if (isEmptyOrWhitespace(str))
+		time = initializeToDefaultTime(typmod);
+	else
+		time = DatumGetTimeADT(DirectFunctionCall3(time_in, CStringGetDatum(str), InvalidOid, typmod));
 	pfree(str);
 	PG_RETURN_TIMEADT(time);
 }

contrib/babelfishpg_tsql/src/pltsql_coerce.c

@@ -33,6 +33,8 @@
 #include "utils/fmgroids.h"
 #include "common/int.h"
 #include "utils/numeric.h"
+#include "utils/date.h"
+#include "utils/datetime.h"
 #include "utils/memutils.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
@@ -2656,7 +2658,7 @@ tsql_coerce_string_literal_hook(Oid targetTypeId,
 		 */
 		for (i = strlen(value) - 1; i >= 0; i--)
 		{
-			if (value[i] != ' ')
+			if (!isspace((unsigned char)value[i]))
 				break;
 		}
 
@@ -2759,6 +2761,20 @@ tsql_coerce_string_literal_hook(Oid targetTypeId,
 						newcon->constvalue = stringTypeDatum(baseType, value, inputTypeMod);
 						break;
 					}
+				case DATEOID:
+					{
+						/* Set input to default '1900-01-01' for empty strings */
+						DateADT date = (*common_utility_plugin_ptr->initializeToDefaultDate) ();
+						newcon->constvalue = DateADTGetDatum(date);
+						break;
+					}
+				case TIMEOID:
+					{
+						/* Set input to default '00:00:00' for empty strings */
+						TimeADT time = (*common_utility_plugin_ptr->initializeToDefaultTime) (inputTypeMod);
+						newcon->constvalue = TimeADTGetDatum(time);
+						break;
+					}
 				default:
 					newcon->constvalue = stringTypeDatum(baseType, value, inputTypeMod);
 			}

test/JDBC/expected/Date.out

@@ -14,22 +14,23 @@ GO
 Declare @a date;
 INSERT INTO DateTest (DateCol) VALUES (@a), ('');
 GO
-~~ERROR (Code: 33557097)~~
-
-~~ERROR (Message: invalid input syntax for type date: "")~~
+~~ROW COUNT: 2~~
 
 SELECT * FROM DateTest WHERE DateCol IS NULL;
 GO
 ~~START~~
 int#!#date
 1#!#<NULL>
+2#!#<NULL>
 ~~END~~
 
 SELECT * FROM DateTest;
 GO
 ~~START~~
 int#!#date
 1#!#<NULL>
+2#!#<NULL>
+3#!#1900-01-01
 ~~END~~
 
 
@@ -2049,9 +2050,10 @@ date
 
 SELECT CAST(CAST('' AS varchar) AS DATE);
 GO
-~~ERROR (Code: 33557097)~~
-
-~~ERROR (Message: invalid input syntax for type date: "")~~
+~~START~~
+date
+1900-01-01
+~~END~~
 
 
 -- nchar
@@ -2650,9 +2652,10 @@ date
 
 SELECT dbo.TestDateFunction(CAST('' AS varchar));
 GO
-~~ERROR (Code: 33557097)~~
-
-~~ERROR (Message: invalid input syntax for type date: "")~~
+~~START~~
+date
+1900-01-01
+~~END~~
 
 
 -- nchar