Speed up byteain by not parsing traditional-style input twice.

Instead of laboriously computing the exact output length, use strlen
to get an upper bound cheaply.  (This is still O(N) of course, but
the constant factor is a lot less.)  This will typically result in
overallocating the output datum, but that's of little concern since
it's a short-lived allocation in just about all use-cases.

A simple microbenchmark showed about 40% speedup for long input
strings.

While here, make some cosmetic cleanups and add a test case that
covers the double-backslash code path in byteain and byteaout.

Author: Steven Niu <niushiji@gmail.com>
Reviewed-by: Kirill Reshke <reshkekirill@gmail.com>
Reviewed-by: Stepan Neretin <slpmcf@gmail.com>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/ca315729-140b-426e-81a6-6cd5cfe7ecc5@gmail.com

3 files changed, 30 insertions, 45 deletions

diff --git a/src/backend/utils/adt/bytea.c b/src/backend/utils/adt/bytea.c
index 2e539c2504e..6e7b914c563 100644
--- a/src/backend/utils/adt/bytea.c
+++ b/src/backend/utils/adt/bytea.c
@@ -182,27 +182,21 @@ bytea_overlay(bytea *t1, bytea *t2, int sp, int sl)
  *
  *		Non-printable characters must be passed as '\nnn' (octal) and are
  *		converted to internal form.  '\' must be passed as '\\'.
- *		ereport(ERROR, ...) if bad form.
- *
- *		BUGS:
- *				The input is scanned twice.
- *				The error checking of input is minimal.
  */
 Datum
 byteain(PG_FUNCTION_ARGS)
 {
 	char	   *inputText = PG_GETARG_CSTRING(0);
 	Node	   *escontext = fcinfo->context;
+	size_t		len = strlen(inputText);
+	size_t		bc;
 	char	   *tp;
 	char	   *rp;
-	int			bc;
 	bytea	   *result;
 
 	/* Recognize hex input */
 	if (inputText[0] == '\\' && inputText[1] == 'x')
 	{
-		size_t		len = strlen(inputText);
-
 		bc = (len - 2) / 2 + VARHDRSZ;	/* maximum possible length */
 		result = palloc(bc);
 		bc = hex_decode_safe(inputText + 2, len - 2, VARDATA(result),
@@ -213,33 +207,7 @@ byteain(PG_FUNCTION_ARGS)
 	}
 
 	/* Else, it's the traditional escaped style */
-	for (bc = 0, tp = inputText; *tp != '\0'; bc++)
-	{
-		if (tp[0] != '\\')
-			tp++;
-		else if ((tp[0] == '\\') &&
-				 (tp[1] >= '0' && tp[1] <= '3') &&
-				 (tp[2] >= '0' && tp[2] <= '7') &&
-				 (tp[3] >= '0' && tp[3] <= '7'))
-			tp += 4;
-		else if ((tp[0] == '\\') &&
-				 (tp[1] == '\\'))
-			tp += 2;
-		else
-		{
-			/*
-			 * one backslash, not followed by another or ### valid octal
-			 */
-			ereturn(escontext, (Datum) 0,
-					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-					 errmsg("invalid input syntax for type %s", "bytea")));
-		}
-	}
-
-	bc += VARHDRSZ;
-
-	result = (bytea *) palloc(bc);
-	SET_VARSIZE(result, bc);
+	result = (bytea *) palloc(len + VARHDRSZ);	/* maximum possible length */
 
 	tp = inputText;
 	rp = VARDATA(result);
@@ -247,21 +215,21 @@ byteain(PG_FUNCTION_ARGS)
 	{
 		if (tp[0] != '\\')
 			*rp++ = *tp++;
-		else if ((tp[0] == '\\') &&
-				 (tp[1] >= '0' && tp[1] <= '3') &&
+		else if ((tp[1] >= '0' && tp[1] <= '3') &&
 				 (tp[2] >= '0' && tp[2] <= '7') &&
 				 (tp[3] >= '0' && tp[3] <= '7'))
 		{
-			bc = VAL(tp[1]);
-			bc <<= 3;
-			bc += VAL(tp[2]);
-			bc <<= 3;
-			*rp++ = bc + VAL(tp[3]);
+			int			v;
+
+			v = VAL(tp[1]);
+			v <<= 3;
+			v += VAL(tp[2]);
+			v <<= 3;
+			*rp++ = v + VAL(tp[3]);
 
 			tp += 4;
 		}
-		else if ((tp[0] == '\\') &&
-				 (tp[1] == '\\'))
+		else if (tp[1] == '\\')
 		{
 			*rp++ = '\\';
 			tp += 2;
@@ -269,7 +237,7 @@ byteain(PG_FUNCTION_ARGS)
 		else
 		{
 			/*
-			 * We should never get here. The first pass should not allow it.
+			 * one backslash, not followed by another or ### valid octal
 			 */
 			ereturn(escontext, (Datum) 0,
 					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
@@ -277,6 +245,9 @@ byteain(PG_FUNCTION_ARGS)
 		}
 	}
 
+	bc = rp - VARDATA(result);	/* actual length */
+	SET_VARSIZE(result, bc + VARHDRSZ);
+
 	PG_RETURN_BYTEA_P(result);
 }
 
diff --git a/src/test/regress/expected/strings.out b/src/test/regress/expected/strings.out
index 788844abd20..1bfd33de3f3 100644
--- a/src/test/regress/expected/strings.out
+++ b/src/test/regress/expected/strings.out
@@ -236,6 +236,12 @@ SELECT E'De\\678dBeEf'::bytea;
 ERROR:  invalid input syntax for type bytea
 LINE 1: SELECT E'De\\678dBeEf'::bytea;
                ^
+SELECT E'DeAd\\\\BeEf'::bytea;
+        bytea         
+----------------------
+ \x446541645c42654566
+(1 row)
+
 SELECT reverse(''::bytea);
  reverse 
 ---------
@@ -291,6 +297,12 @@ SELECT E'De\\123dBeEf'::bytea;
  DeSdBeEf
 (1 row)
 
+SELECT E'DeAd\\\\BeEf'::bytea;
+   bytea    
+------------
+ DeAd\\BeEf
+(1 row)
+
 -- Test non-error-throwing API too
 SELECT pg_input_is_valid(E'\\xDeAdBeE', 'bytea');
  pg_input_is_valid 
diff --git a/src/test/regress/sql/strings.sql b/src/test/regress/sql/strings.sql
index 2577a42987d..92c445c2439 100644
--- a/src/test/regress/sql/strings.sql
+++ b/src/test/regress/sql/strings.sql
@@ -76,6 +76,7 @@ SELECT E'De\\000dBeEf'::bytea;
 SELECT E'De\123dBeEf'::bytea;
 SELECT E'De\\123dBeEf'::bytea;
 SELECT E'De\\678dBeEf'::bytea;
+SELECT E'DeAd\\\\BeEf'::bytea;
 
 SELECT reverse(''::bytea);
 SELECT reverse('\xaa'::bytea);
@@ -88,6 +89,7 @@ SELECT E'\\xDe00BeEf'::bytea;
 SELECT E'DeAdBeEf'::bytea;
 SELECT E'De\\000dBeEf'::bytea;
 SELECT E'De\\123dBeEf'::bytea;
+SELECT E'DeAd\\\\BeEf'::bytea;
 
 -- Test non-error-throwing API too
 SELECT pg_input_is_valid(E'\\xDeAdBeE', 'bytea');