From: drh <>
Date: Thu, 9 Mar 2023 16:04:34 +0000 (+0000)
Subject: In the Bloom filter optimization, hash all strings and blobs into the same
X-Git-Tag: version-3.41.1~1
X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=10e44bf373df584a31f3a1011720b10f00429010;p=thirdparty%2Fsqlite.git

In the Bloom filter optimization, hash all strings and blobs into the same
value, because we do not know if two different strings might compare equal
even if they have different byte sequences, due to collating functions.
Formerly, the hash of a string or blob was just its length.  This could
all be improved.

FossilOrigin-Name: cc8a0ee40cfc83ab42a0ff452de0a53fe17784aefb944ea7ac2cb078a8310730
---

diff --git a/manifest b/manifest
index 14cf93a8aa..23ec9ec89b 100644
--- a/manifest
+++ b/manifest
@@ -1,5 +1,5 @@
-C Fix\scountofview.test\sso\sthat\sit\sworks\swith\sSQLITE_OMIT_PROGRESS_CALLBACK\sbuilds.
-D 2023-03-09T15:08:56.667
+C In\sthe\sBloom\sfilter\soptimization,\shash\sall\sstrings\sand\sblobs\sinto\sthe\ssame\nvalue,\sbecause\swe\sdo\snot\sknow\sif\stwo\sdifferent\sstrings\smight\scompare\sequal\neven\sif\sthey\shave\sdifferent\sbyte\ssequences,\sdue\sto\scollating\sfunctions.\nFormerly,\sthe\shash\sof\sa\sstring\sor\sblob\swas\sjust\sits\slength.\s\sThis\scould\nall\sbe\simproved.
+D 2023-03-09T16:04:34.291
 F .fossil-settings/empty-dirs dbb81e8fc0401ac46a1491ab34a7f2c7c0452f2f06b54ebb845d024ca8283ef1
 F .fossil-settings/ignore-glob 35175cdfcf539b2318cb04a9901442804be81cd677d8b889fcc9149c21f239ea
 F LICENSE.md df5091916dbb40e6e9686186587125e1b2ff51f022cc334e886c19a0e9982724
@@ -691,7 +691,7 @@ F src/upsert.c 5303dc6c518fa7d4b280ec65170f465c7a70b7ac2b22491598f6d0b4875b3145
 F src/utf.c ee39565f0843775cc2c81135751ddd93eceb91a673ea2c57f61c76f288b041a0
 F src/util.c 3ff7bc2b48dd425b1448304bb86273b05da1621f136d51dbb9789f8803559a1f
 F src/vacuum.c 84ce7f01f8a7a08748e107a441db83bcec13970190ddcb0c9ff522adbc1c23fd
-F src/vdbe.c 0bf4804f6e9cae872502aa4533a62ba3f6e97ee28523302deb27f9c990b5f44f
+F src/vdbe.c b3fd04b0643edd7e0a4356aff6d2cf50f04d0e182e292c3a330d1afffe3100e1
 F src/vdbe.h 73b904a6b3bb27f308c6cc287a5751ebc7f1f89456be0ed068a12b92844c6e8c
 F src/vdbeInt.h a4147a4ddf613cb1bcb555ace9e9e74a9c099d65facd88155f191b1fb4d74cfb
 F src/vdbeapi.c 40c47b1528d308a322203de21d2e0d711753257ed9771771b6129214b1d65932
@@ -814,7 +814,7 @@ F test/bind2.test 918bc35135f4141809ead7585909cde57d44db90a7a62aef540127148f91aa
 F test/bindxfer.test efecd12c580c14df5f4ad3b3e83c667744a4f7e0
 F test/bitvec.test 75894a880520164d73b1305c1c3f96882615e142
 F test/blob.test e7ac6c7d3a985cc4678c64f325292529a69ae252
-F test/bloom1.test 2785a190fcc2a5e170e5d38b08aca8ff0f3e3b4a74d47453d6ac1bd355180a6a
+F test/bloom1.test 589361c1f20158a8583863738c883f0e73e82d18422c9b4ed9c7068c13c2d310
 F test/boundary1.tcl 6421b2d920d8b09539503a8673339d32f7609eb1
 F test/boundary1.test 66d7f4706ccdb42d58eafdb081de07b0eb42d77b
 F test/boundary2.tcl e34ef4e930cf1083150d4d2c603e146bd3b76bcb
@@ -2045,9 +2045,9 @@ F vsixtest/vsixtest.tcl 6a9a6ab600c25a91a7acc6293828957a386a8a93
 F vsixtest/vsixtest.vcxproj.data 2ed517e100c66dc455b492e1a33350c1b20fbcdc
 F vsixtest/vsixtest.vcxproj.filters 37e51ffedcdb064aad6ff33b6148725226cd608e
 F vsixtest/vsixtest_TemporaryKey.pfx e5b1b036facdb453873e7084e1cae9102ccc67a0
-P e4e2e64725a196ee4f287649db33f49fd7329c95cef3ce743cd177c998a6a343
-Q +2fc7c3fcee05c2a251ceb3666f3f6e9014cfe6e2f8570b72c43f251067e6b252
-R 53da1bf93f2f919c47eb8511c942a273
-U dan
-Z ceb9f98a8348af253cc47acc64f81e23
+P d55a7742c9447eba20a7fe252fb3edcabba3793030a55976d5a5e6c65bce7706
+Q +090304b870419acb5b05205a07fc75830b556928149f76a843cda526f77a6fc0
+R 36a17de70fdfc70a02115f278d762e2d
+U drh
+Z 997ae489aebc30f3e5f62ab576efe0df
 # Remove this line to create a well-formed Fossil manifest.
diff --git a/manifest.uuid b/manifest.uuid
index 4f5ea50ac3..d24bb35f8b 100644
--- a/manifest.uuid
+++ b/manifest.uuid
@@ -1 +1 @@
-d55a7742c9447eba20a7fe252fb3edcabba3793030a55976d5a5e6c65bce7706
\ No newline at end of file
+cc8a0ee40cfc83ab42a0ff452de0a53fe17784aefb944ea7ac2cb078a8310730
\ No newline at end of file
diff --git a/src/vdbe.c b/src/vdbe.c
index f3853e7fd2..a9febbb6e2 100644
--- a/src/vdbe.c
+++ b/src/vdbe.c
@@ -683,8 +683,7 @@ static u64 filterHash(const Mem *aMem, const Op *pOp){
     }else if( p->flags & MEM_Real ){
       h += sqlite3VdbeIntValue(p);
     }else if( p->flags & (MEM_Str|MEM_Blob) ){
-      h += p->n;
-      if( p->flags & MEM_Zero ) h += p->u.nZero;
+      /* no-op */
     }
   }
   return h;
diff --git a/test/bloom1.test b/test/bloom1.test
index 1846e4d63f..87a7c8632c 100644
--- a/test/bloom1.test
+++ b/test/bloom1.test
@@ -99,6 +99,20 @@ do_eqp_test 2.1 {
   |     |--SCAN transit
   |     `--SEARCH objs USING COVERING INDEX objs_cspo (o=? AND p=?)
   `--SCAN transit
-} 
+}
+
+# 2023-02-28
+# https://sqlite.org/forum/forumpost/0846211821
+#
+# Bloom filter gives an incorrect result if the collating sequence is
+# anything other than binary.
+#
+reset_db
+do_execsql_test 3.1 {
+  CREATE TABLE t0(x TEXT COLLATE rtrim);
+  INSERT INTO t0(x) VALUES ('a'), ('b'), ('c');
+  CREATE VIEW v0(y) AS SELECT DISTINCT x FROM t0;
+  SELECT count(*) FROM t0, v0 WHERE x='b ';
+} 3
 
 finish_test