v7-0013-PoC-Get-rid-of-strlen-calls-when-using-HASH_STRIN.patch

From b7ee120e05d48ebacb078e00ffefc5f98052d214 Mon Sep 17 00:00:00 2001
From: John Naylor <john.naylor@postgresql.org>
Date: Sat, 9 Dec 2023 18:06:05 +0700
Subject: [PATCH v7 13/13] PoC: Get rid of strlen() calls when using
 HASH_STRINGS

Add cstring_hash, which uses the chunked incremental interface
of fasthash. That way, we don't need know the length of the
key upfront.

Open questions:
- Is performance better?
- Since we have the total length when we reach the end, should
  well try to use it in the finalization stage?
- Do we need to keep string_hash around?
---
 src/backend/utils/hash/dynahash.c | 49 +++++++++++++++++++++++++++----
 1 file changed, 44 insertions(+), 5 deletions(-)

diff --git a/src/backend/utils/hash/dynahash.c b/src/backend/utils/hash/dynahash.c
index 012d4a0b1f..ba74126e73 100644
--- a/src/backend/utils/hash/dynahash.c
+++ b/src/backend/utils/hash/dynahash.c
@@ -98,6 +98,7 @@
 
 #include "access/xact.h"
 #include "common/hashfn.h"
+#include "common/hashfn_unstable.h"
 #include "port/pg_bitutils.h"
 #include "storage/shmem.h"
 #include "storage/spin.h"
@@ -307,6 +308,44 @@ string_compare(const char *key1, const char *key2, Size keysize)
 	return strncmp(key1, key2, keysize - 1);
 }
 
+/*
+ * cstring_hash: hash function for keys that are NUL-terminated strings.
+ *
+ * NOTE: this is the default hash function if none is specified.
+ */
+static uint32
+cstring_hash(const void *key, Size keysize)
+{
+	fasthash_state hs;
+	int s_len = 0;
+	const unsigned char *k = (const unsigned char *) key;
+
+	/*
+	 * If the string exceeds keysize-1 bytes, we want to hash only that many,
+	 * because when it is copied into the hash table it will be truncated at
+	 * that length.
+	 */
+
+	fasthash_init(&hs, 0, 0, false);
+
+	while (*k && s_len < keysize)
+	{
+		int chunk_len;
+
+		for (chunk_len = 0;
+			chunk_len < FH_SIZEOF_ACCUM && k[chunk_len] != '\0' && s_len < keysize;
+			chunk_len++)
+		{
+			s_len++;
+		}
+
+		fasthash_accum(&hs, k, chunk_len);
+		k += chunk_len;
+	}
+
+	return fasthash_final32(&hs);
+}
+
 
 /************************** CREATE ROUTINES **********************/
 
@@ -419,7 +458,7 @@ hash_create(const char *tabname, long nelem, const HASHCTL *info, int flags)
 	{
 		/*
 		 * string_hash used to be considered the default hash method, and in a
-		 * non-assert build it effectively still is.  But we now consider it
+		 * non-assert build it effectively still was until version 17.  Since version 14 we consider it
 		 * an assertion error to not say HASH_STRINGS explicitly.  To help
 		 * catch mistaken usage of HASH_STRINGS, we also insist on a
 		 * reasonably long string length: if the keysize is only 4 or 8 bytes,
@@ -428,12 +467,12 @@ hash_create(const char *tabname, long nelem, const HASHCTL *info, int flags)
 		Assert(flags & HASH_STRINGS);
 		Assert(info->keysize > 8);
 
-		hashp->hash = string_hash;
+		hashp->hash = cstring_hash;
 	}
 
 	/*
 	 * If you don't specify a match function, it defaults to string_compare if
-	 * you used string_hash, and to memcmp otherwise.
+	 * you used cstring_hash, and to memcmp otherwise.
 	 *
 	 * Note: explicitly specifying string_hash is deprecated, because this
 	 * might not work for callers in loadable modules on some platforms due to
@@ -442,7 +481,7 @@ hash_create(const char *tabname, long nelem, const HASHCTL *info, int flags)
 	 */
 	if (flags & HASH_COMPARE)
 		hashp->match = info->match;
-	else if (hashp->hash == string_hash)
+	else if (hashp->hash == cstring_hash)
 		hashp->match = (HashCompareFunc) string_compare;
 	else
 		hashp->match = memcmp;
@@ -452,7 +491,7 @@ hash_create(const char *tabname, long nelem, const HASHCTL *info, int flags)
 	 */
 	if (flags & HASH_KEYCOPY)
 		hashp->keycopy = info->keycopy;
-	else if (hashp->hash == string_hash)
+	else if (hashp->hash == cstring_hash)
 	{
 		/*
 		 * The signature of keycopy is meant for memcpy(), which returns
-- 
2.43.0