Convert a few more datatype input functions to report errors softly.

Convert the remaining string-category input functions
(bpcharin, varcharin, byteain) to the new style.

Discussion: https://postgr.es/m/3038346.1671060258@sss.pgh.pa.us

src/backend/utils/adt/encode.c

@@ -171,8 +171,8 @@ hex_encode(const char *src, size_t len, char *dst)
 	return (uint64) len * 2;
 }
 
-static inline char
+static inline bool
-get_hex(const char *cp)
+get_hex(const char *cp, char *out)
 {
 	unsigned char c = (unsigned char) *cp;
 	int			res = -1;
@@ -180,17 +180,19 @@ get_hex(const char *cp)
 	if (c < 127)
 		res = hexlookup[c];
 
-	if (res < 0)
+	*out = (char) res;
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
-				 errmsg("invalid hexadecimal digit: \"%.*s\"",
-						pg_mblen(cp), cp)));
 
-	return (char) res;
+	return (res >= 0);
 }
 
 uint64
 hex_decode(const char *src, size_t len, char *dst)
+{
+	return hex_decode_safe(src, len, dst, NULL);
+}
+
+uint64
+hex_decode_safe(const char *src, size_t len, char *dst, Node *escontext)
 {
 	const char *s,
 			   *srcend;
@@ -208,16 +210,23 @@ hex_decode(const char *src, size_t len, char *dst)
 			s++;
 			continue;
 		}
-		v1 = get_hex(s) << 4;
+		if (!get_hex(s, &v1))
+			ereturn(escontext, 0,
+					(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+					 errmsg("invalid hexadecimal digit: \"%.*s\"",
+							pg_mblen(s), s)));
 		s++;
 		if (s >= srcend)
-			ereport(ERROR,
+			ereturn(escontext, 0,
 					(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
 					 errmsg("invalid hexadecimal data: odd number of digits")));
-
+		if (!get_hex(s, &v2))
-		v2 = get_hex(s);
+			ereturn(escontext, 0,
+					(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+					 errmsg("invalid hexadecimal digit: \"%.*s\"",
+							pg_mblen(s), s)));
 		s++;
-		*p++ = v1 | v2;
+		*p++ = (v1 << 4) | v2;
 	}
 
 	return p - dst;

src/backend/utils/adt/varchar.c

@@ -122,9 +122,13 @@ anychar_typmodout(int32 typmod)
  *
  * If the input string is too long, raise an error, unless the extra
  * characters are spaces, in which case they're truncated.  (per SQL)
+ *
+ * If escontext points to an ErrorSaveContext node, that is filled instead
+ * of throwing an error; the caller must check SOFT_ERROR_OCCURRED()
+ * to detect errors.
  */
 static BpChar *
-bpchar_input(const char *s, size_t len, int32 atttypmod)
+bpchar_input(const char *s, size_t len, int32 atttypmod, Node *escontext)
 {
 	BpChar	   *result;
 	char	   *r;
@@ -153,7 +157,7 @@ bpchar_input(const char *s, size_t len, int32 atttypmod)
 			for (j = mbmaxlen; j < len; j++)
 			{
 				if (s[j] != ' ')
-					ereport(ERROR,
+					ereturn(escontext, NULL,
 							(errcode(ERRCODE_STRING_DATA_RIGHT_TRUNCATION),
 							 errmsg("value too long for type character(%d)",
 									(int) maxlen)));
@@ -195,14 +199,13 @@ Datum
 bpcharin(PG_FUNCTION_ARGS)
 {
 	char	   *s = PG_GETARG_CSTRING(0);
-
 #ifdef NOT_USED
 	Oid			typelem = PG_GETARG_OID(1);
 #endif
 	int32		atttypmod = PG_GETARG_INT32(2);
 	BpChar	   *result;
 
-	result = bpchar_input(s, strlen(s), atttypmod);
+	result = bpchar_input(s, strlen(s), atttypmod, fcinfo->context);
 	PG_RETURN_BPCHAR_P(result);
 }
 
@@ -228,7 +231,6 @@ Datum
 bpcharrecv(PG_FUNCTION_ARGS)
 {
 	StringInfo	buf = (StringInfo) PG_GETARG_POINTER(0);
-
 #ifdef NOT_USED
 	Oid			typelem = PG_GETARG_OID(1);
 #endif
@@ -238,7 +240,7 @@ bpcharrecv(PG_FUNCTION_ARGS)
 	int			nbytes;
 
 	str = pq_getmsgtext(buf, buf->len - buf->cursor, &nbytes);
-	result = bpchar_input(str, nbytes, atttypmod);
+	result = bpchar_input(str, nbytes, atttypmod, NULL);
 	pfree(str);
 	PG_RETURN_BPCHAR_P(result);
 }
@@ -448,11 +450,12 @@ bpchartypmodout(PG_FUNCTION_ARGS)
  * If the input string is too long, raise an error, unless the extra
  * characters are spaces, in which case they're truncated.  (per SQL)
  *
- * Uses the C string to text conversion function, which is only appropriate
+ * If escontext points to an ErrorSaveContext node, that is filled instead
- * if VarChar and text are equivalent types.
+ * of throwing an error; the caller must check SOFT_ERROR_OCCURRED()
+ * to detect errors.
  */
 static VarChar *
-varchar_input(const char *s, size_t len, int32 atttypmod)
+varchar_input(const char *s, size_t len, int32 atttypmod, Node *escontext)
 {
 	VarChar    *result;
 	size_t		maxlen;
@@ -468,7 +471,7 @@ varchar_input(const char *s, size_t len, int32 atttypmod)
 		for (j = mbmaxlen; j < len; j++)
 		{
 			if (s[j] != ' ')
-				ereport(ERROR,
+				ereturn(escontext, NULL,
 						(errcode(ERRCODE_STRING_DATA_RIGHT_TRUNCATION),
 						 errmsg("value too long for type character varying(%d)",
 								(int) maxlen)));
@@ -477,6 +480,10 @@ varchar_input(const char *s, size_t len, int32 atttypmod)
 		len = mbmaxlen;
 	}
 
+	/*
+	 * We can use cstring_to_text_with_len because VarChar and text are
+	 * binary-compatible types.
+	 */
 	result = (VarChar *) cstring_to_text_with_len(s, len);
 	return result;
 }
@@ -489,14 +496,13 @@ Datum
 varcharin(PG_FUNCTION_ARGS)
 {
 	char	   *s = PG_GETARG_CSTRING(0);
-
 #ifdef NOT_USED
 	Oid			typelem = PG_GETARG_OID(1);
 #endif
 	int32		atttypmod = PG_GETARG_INT32(2);
 	VarChar    *result;
 
-	result = varchar_input(s, strlen(s), atttypmod);
+	result = varchar_input(s, strlen(s), atttypmod, fcinfo->context);
 	PG_RETURN_VARCHAR_P(result);
 }
 
@@ -522,7 +528,6 @@ Datum
 varcharrecv(PG_FUNCTION_ARGS)
 {
 	StringInfo	buf = (StringInfo) PG_GETARG_POINTER(0);
-
 #ifdef NOT_USED
 	Oid			typelem = PG_GETARG_OID(1);
 #endif
@@ -532,7 +537,7 @@ varcharrecv(PG_FUNCTION_ARGS)
 	int			nbytes;
 
 	str = pq_getmsgtext(buf, buf->len - buf->cursor, &nbytes);
-	result = varchar_input(str, nbytes, atttypmod);
+	result = varchar_input(str, nbytes, atttypmod, NULL);
 	pfree(str);
 	PG_RETURN_VARCHAR_P(result);
 }

src/backend/utils/adt/varlena.c

@@ -295,6 +295,7 @@ Datum
 byteain(PG_FUNCTION_ARGS)
 {
 	char	   *inputText = PG_GETARG_CSTRING(0);
+	Node	   *escontext = fcinfo->context;
 	char	   *tp;
 	char	   *rp;
 	int			bc;
@@ -307,7 +308,8 @@ byteain(PG_FUNCTION_ARGS)
 
 		bc = (len - 2) / 2 + VARHDRSZ;	/* maximum possible length */
 		result = palloc(bc);
-		bc = hex_decode(inputText + 2, len - 2, VARDATA(result));
+		bc = hex_decode_safe(inputText + 2, len - 2, VARDATA(result),
+							 escontext);
 		SET_VARSIZE(result, bc + VARHDRSZ); /* actual length */
 
 		PG_RETURN_BYTEA_P(result);
@@ -331,7 +333,7 @@ byteain(PG_FUNCTION_ARGS)
 			/*
 			 * one backslash, not followed by another or ### valid octal
 			 */
-			ereport(ERROR,
+			ereturn(escontext, (Datum) 0,
 					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
 					 errmsg("invalid input syntax for type %s", "bytea")));
 		}
@@ -372,7 +374,7 @@ byteain(PG_FUNCTION_ARGS)
 			/*
 			 * We should never get here. The first pass should not allow it.
 			 */
-			ereport(ERROR,
+			ereturn(escontext, (Datum) 0,
 					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
 					 errmsg("invalid input syntax for type %s", "bytea")));
 		}

src/include/utils/builtins.h

@@ -34,6 +34,8 @@ extern int	errdomainconstraint(Oid datatypeOid, const char *conname);
 /* encode.c */
 extern uint64 hex_encode(const char *src, size_t len, char *dst);
 extern uint64 hex_decode(const char *src, size_t len, char *dst);
+extern uint64 hex_decode_safe(const char *src, size_t len, char *dst,
+							  Node *escontext);
 
 /* int.c */
 extern int2vector *buildint2vector(const int16 *int2s, int n);

src/test/regress/expected/char.out

@@ -119,6 +119,25 @@ SELECT * FROM CHAR_TBL;
  abcd
 (4 rows)
 
+-- Also try it with non-error-throwing API
+SELECT pg_input_is_valid('abcd  ', 'char(4)');
+ pg_input_is_valid 
+-------------------
+ t
+(1 row)
+
+SELECT pg_input_is_valid('abcde', 'char(4)');
+ pg_input_is_valid 
+-------------------
+ f
+(1 row)
+
+SELECT pg_input_error_message('abcde', 'char(4)');
+        pg_input_error_message        
+--------------------------------------
+ value too long for type character(4)
+(1 row)
+
 --
 -- Also test "char", which is an ad-hoc one-byte type.  It can only
 -- really store ASCII characters, but we allow high-bit-set characters

src/test/regress/expected/char_1.out

@@ -119,6 +119,25 @@ SELECT * FROM CHAR_TBL;
  abcd
 (4 rows)
 
+-- Also try it with non-error-throwing API
+SELECT pg_input_is_valid('abcd  ', 'char(4)');
+ pg_input_is_valid 
+-------------------
+ t
+(1 row)
+
+SELECT pg_input_is_valid('abcde', 'char(4)');
+ pg_input_is_valid 
+-------------------
+ f
+(1 row)
+
+SELECT pg_input_error_message('abcde', 'char(4)');
+        pg_input_error_message        
+--------------------------------------
+ value too long for type character(4)
+(1 row)
+
 --
 -- Also test "char", which is an ad-hoc one-byte type.  It can only
 -- really store ASCII characters, but we allow high-bit-set characters

src/test/regress/expected/char_2.out

@@ -119,6 +119,25 @@ SELECT * FROM CHAR_TBL;
  abcd
 (4 rows)
 
+-- Also try it with non-error-throwing API
+SELECT pg_input_is_valid('abcd  ', 'char(4)');
+ pg_input_is_valid 
+-------------------
+ t
+(1 row)
+
+SELECT pg_input_is_valid('abcde', 'char(4)');
+ pg_input_is_valid 
+-------------------
+ f
+(1 row)
+
+SELECT pg_input_error_message('abcde', 'char(4)');
+        pg_input_error_message        
+--------------------------------------
+ value too long for type character(4)
+(1 row)
+
 --
 -- Also test "char", which is an ad-hoc one-byte type.  It can only
 -- really store ASCII characters, but we allow high-bit-set characters

src/test/regress/expected/strings.out

@@ -273,6 +273,31 @@ SELECT E'De\\123dBeEf'::bytea;
  DeSdBeEf
 (1 row)
 
+-- Test non-error-throwing API too
+SELECT pg_input_is_valid(E'\\xDeAdBeE', 'bytea');
+ pg_input_is_valid 
+-------------------
+ f
+(1 row)
+
+SELECT pg_input_error_message(E'\\xDeAdBeE', 'bytea');
+             pg_input_error_message             
+------------------------------------------------
+ invalid hexadecimal data: odd number of digits
+(1 row)
+
+SELECT pg_input_error_message(E'\\xDeAdBeEx', 'bytea');
+     pg_input_error_message     
+--------------------------------
+ invalid hexadecimal digit: "x"
+(1 row)
+
+SELECT pg_input_error_message(E'foo\\99bar', 'bytea');
+       pg_input_error_message        
+-------------------------------------
+ invalid input syntax for type bytea
+(1 row)
+
 --
 -- test conversions between various string types
 -- E021-10 implicit casting among the character data types

src/test/regress/expected/varchar.out

@@ -111,3 +111,22 @@ SELECT * FROM VARCHAR_TBL;
  abcd
 (4 rows)
 
+-- Also try it with non-error-throwing API
+SELECT pg_input_is_valid('abcd  ', 'varchar(4)');
+ pg_input_is_valid 
+-------------------
+ t
+(1 row)
+
+SELECT pg_input_is_valid('abcde', 'varchar(4)');
+ pg_input_is_valid 
+-------------------
+ f
+(1 row)
+
+SELECT pg_input_error_message('abcde', 'varchar(4)');
+            pg_input_error_message            
+----------------------------------------------
+ value too long for type character varying(4)
+(1 row)
+

src/test/regress/expected/varchar_1.out

@@ -111,3 +111,22 @@ SELECT * FROM VARCHAR_TBL;
  abcd
 (4 rows)
 
+-- Also try it with non-error-throwing API
+SELECT pg_input_is_valid('abcd  ', 'varchar(4)');
+ pg_input_is_valid 
+-------------------
+ t
+(1 row)
+
+SELECT pg_input_is_valid('abcde', 'varchar(4)');
+ pg_input_is_valid 
+-------------------
+ f
+(1 row)
+
+SELECT pg_input_error_message('abcde', 'varchar(4)');
+            pg_input_error_message            
+----------------------------------------------
+ value too long for type character varying(4)
+(1 row)
+

src/test/regress/expected/varchar_2.out

@@ -111,3 +111,22 @@ SELECT * FROM VARCHAR_TBL;
  abcd
 (4 rows)
 
+-- Also try it with non-error-throwing API
+SELECT pg_input_is_valid('abcd  ', 'varchar(4)');
+ pg_input_is_valid 
+-------------------
+ t
+(1 row)
+
+SELECT pg_input_is_valid('abcde', 'varchar(4)');
+ pg_input_is_valid 
+-------------------
+ f
+(1 row)
+
+SELECT pg_input_error_message('abcde', 'varchar(4)');
+            pg_input_error_message            
+----------------------------------------------
+ value too long for type character varying(4)
+(1 row)
+

src/test/regress/sql/char.sql

@@ -72,6 +72,11 @@ INSERT INTO CHAR_TBL (f1) VALUES ('abcde');
 
 SELECT * FROM CHAR_TBL;
 
+-- Also try it with non-error-throwing API
+SELECT pg_input_is_valid('abcd  ', 'char(4)');
+SELECT pg_input_is_valid('abcde', 'char(4)');
+SELECT pg_input_error_message('abcde', 'char(4)');
+
 --
 -- Also test "char", which is an ad-hoc one-byte type.  It can only
 -- really store ASCII characters, but we allow high-bit-set characters

src/test/regress/sql/strings.sql

@@ -85,6 +85,12 @@ SELECT E'DeAdBeEf'::bytea;
 SELECT E'De\\000dBeEf'::bytea;
 SELECT E'De\\123dBeEf'::bytea;
 
+-- Test non-error-throwing API too
+SELECT pg_input_is_valid(E'\\xDeAdBeE', 'bytea');
+SELECT pg_input_error_message(E'\\xDeAdBeE', 'bytea');
+SELECT pg_input_error_message(E'\\xDeAdBeEx', 'bytea');
+SELECT pg_input_error_message(E'foo\\99bar', 'bytea');
+
 --
 -- test conversions between various string types
 -- E021-10 implicit casting among the character data types

src/test/regress/sql/varchar.sql

@@ -66,3 +66,8 @@ DROP TABLE VARCHAR_TBL;
 INSERT INTO VARCHAR_TBL (f1) VALUES ('abcde');
 
 SELECT * FROM VARCHAR_TBL;
+
+-- Also try it with non-error-throwing API
+SELECT pg_input_is_valid('abcd  ', 'varchar(4)');
+SELECT pg_input_is_valid('abcde', 'varchar(4)');
+SELECT pg_input_error_message('abcde', 'varchar(4)');