xref: /sqlite-3.40.0/test/pageropt.test (revision 82f52540)
1538f570cSdrh# 2007 April 12
2538f570cSdrh#
3538f570cSdrh# The author disclaims copyright to this source code.  In place of
4538f570cSdrh# a legal notice, here is a blessing:
5538f570cSdrh#
6538f570cSdrh#    May you do good and not evil.
7538f570cSdrh#    May you find forgiveness for yourself and forgive others.
8538f570cSdrh#    May you share freely, never taking more than you give.
9538f570cSdrh#
10538f570cSdrh#***********************************************************************
11538f570cSdrh# This file implements regression tests for SQLite library.
12538f570cSdrh# The focus of the tests in this file are to verify that the
13538f570cSdrh# pager optimizations implemented in version 3.3.14 work.
14538f570cSdrh#
158c0a791aSdanielk1977# $Id: pageropt.test,v 1.5 2008/08/20 14:49:25 danielk1977 Exp $
16538f570cSdrh
17538f570cSdrhset testdir [file dirname $argv0]
18538f570cSdrhsource $testdir/tester.tcl
197da56b4fSdrhdo_not_use_codec
20538f570cSdrh
21f4ba1093Sdanifcapable {!pager_pragmas||secure_delete||direct_read} {
22538f570cSdrh  finish_test
23538f570cSdrh  return
24538f570cSdrh}
25538f570cSdrh
26*82f52540Sdrh# A non-zero reserved_bytes value changes the number of pages in the
27*82f52540Sdrh# database file, which messes up the results in this test.
28*82f52540Sdrhif {[nonzero_reserved_bytes]} {finish_test; return;}
29*82f52540Sdrh
30538f570cSdrh# Run the SQL statement supplied by the argument and return
31538f570cSdrh# the results.  Prepend four integers to the beginning of the
32538f570cSdrh# result which are
33538f570cSdrh#
34538f570cSdrh#     (1)  The number of page reads from the database
35538f570cSdrh#     (2)  The number of page writes to the database
36538f570cSdrh#     (3)  The number of page writes to the journal
37538f570cSdrh#     (4)  The number of cache pages freed
38538f570cSdrh#
39538f570cSdrhproc pagercount_sql {sql {db db}} {
40538f570cSdrh  global sqlite3_pager_readdb_count
41538f570cSdrh  global sqlite3_pager_writedb_count
42538f570cSdrh  global sqlite3_pager_writej_count
43538f570cSdrh  global sqlite3_pager_pgfree_count
44538f570cSdrh  set sqlite3_pager_readdb_count 0
45538f570cSdrh  set sqlite3_pager_writedb_count 0
46538f570cSdrh  set sqlite3_pager_writej_count 0
47538f570cSdrh  set r [$db eval $sql]
48538f570cSdrh  set cnt [list $sqlite3_pager_readdb_count \
49538f570cSdrh                $sqlite3_pager_writedb_count \
508c0a791aSdanielk1977                $sqlite3_pager_writej_count ]
51538f570cSdrh  return [concat $cnt $r]
52538f570cSdrh}
53538f570cSdrh
54538f570cSdrh# Setup the test database
55538f570cSdrh#
56538f570cSdrhdo_test pageropt-1.1 {
573aefabafSdrh  sqlite3_soft_heap_limit 0
58538f570cSdrh  execsql {
59538f570cSdrh    PRAGMA auto_vacuum = OFF;
60538f570cSdrh    PRAGMA page_size = 1024;
61538f570cSdrh  }
62538f570cSdrh  pagercount_sql {
63538f570cSdrh    CREATE TABLE t1(x);
64538f570cSdrh  }
658c0a791aSdanielk1977} {0 2 0}
66538f570cSdrhdo_test pageropt-1.2 {
67538f570cSdrh  pagercount_sql {
68538f570cSdrh    INSERT INTO t1 VALUES(randomblob(5000));
69538f570cSdrh  }
708c0a791aSdanielk1977} {0 6 2}
71538f570cSdrh
72538f570cSdrh# Verify that values remain in cache on for subsequent reads.
73538f570cSdrh# We should not have to go back to disk.
74538f570cSdrh#
75538f570cSdrhdo_test pageropt-1.3 {
76538f570cSdrh  pagercount_sql {
77538f570cSdrh    SELECT length(x) FROM t1
78538f570cSdrh  }
798c0a791aSdanielk1977} {0 0 0 5000}
80538f570cSdrh
81538f570cSdrh# If another thread reads the database, the original cache
82538f570cSdrh# remains valid.
83538f570cSdrh#
84538f570cSdrhsqlite3 db2 test.db
85538f570cSdrhset blobcontent [db2 one {SELECT hex(x) FROM t1}]
86538f570cSdrhdo_test pageropt-1.4 {
87538f570cSdrh  pagercount_sql {
88538f570cSdrh    SELECT hex(x) FROM t1
89538f570cSdrh  }
908c0a791aSdanielk1977} [list 0 0 0 $blobcontent]
91538f570cSdrh
92538f570cSdrh# But if the other thread modifies the database, then the cache
93538f570cSdrh# must refill.
94538f570cSdrh#
95188d4884Sdrhifcapable mmap {
969b4c59faSdrh  set x [expr {[permutation]=="mmap" ? 1 : 6}]
97188d4884Sdrh} else {
98188d4884Sdrh  set x 6
99188d4884Sdrh}
100538f570cSdrhdo_test pageropt-1.5 {
101538f570cSdrh  db2 eval {CREATE TABLE t2(y)}
102538f570cSdrh  pagercount_sql {
103538f570cSdrh    SELECT hex(x) FROM t1
104538f570cSdrh  }
105188d4884Sdrh} [list $x 0 0 $blobcontent]
106538f570cSdrhdo_test pageropt-1.6 {
107538f570cSdrh  pagercount_sql {
108538f570cSdrh    SELECT hex(x) FROM t1
109538f570cSdrh  }
1108c0a791aSdanielk1977} [list 0 0 0 $blobcontent]
111538f570cSdrh
112538f570cSdrh# Verify that the last page of an overflow chain is not read from
113538f570cSdrh# disk when deleting a row.  The one row of t1(x) has four pages
114538f570cSdrh# of overflow.  So deleting that row from t1 should involve reading
115538f570cSdrh# the sqlite_master table (1 page) the main page of t1 (1 page) and
116538f570cSdrh# the three overflow pages of t1 for a total of 5 pages.
117538f570cSdrh#
118538f570cSdrh# Pages written are page 1 (for the freelist pointer), the root page
119538f570cSdrh# of the table, and one of the overflow chain pointers because it
120538f570cSdrh# becomes the trunk of the freelist.  Total 3.
121538f570cSdrh#
122538f570cSdrhdo_test pageropt-2.1 {
123538f570cSdrh  db close
124538f570cSdrh  sqlite3 db test.db
125538f570cSdrh  pagercount_sql {
126538f570cSdrh    DELETE FROM t1 WHERE rowid=1
127538f570cSdrh  }
1288c0a791aSdanielk1977} {5 3 3}
129538f570cSdrh
130538f570cSdrh# When pulling pages off of the freelist, there is no reason
131538f570cSdrh# to actually bring in the old content.
132538f570cSdrh#
133538f570cSdrhdo_test pageropt-2.2 {
134538f570cSdrh  db close
135538f570cSdrh  sqlite3 db test.db
136538f570cSdrh  pagercount_sql {
137538f570cSdrh    INSERT INTO t1 VALUES(randomblob(1500));
138538f570cSdrh  }
1398c0a791aSdanielk1977} {3 4 3}
140538f570cSdrhdo_test pageropt-2.3 {
141538f570cSdrh  pagercount_sql {
142538f570cSdrh    INSERT INTO t1 VALUES(randomblob(1500));
143538f570cSdrh  }
1448c0a791aSdanielk1977} {0 4 3}
145538f570cSdrh
1466558db80Sdrh# Note the new optimization that when pulling the very last page off of the
1476558db80Sdrh# freelist we do not read the content of that page.
1486558db80Sdrh#
1496558db80Sdrhdo_test pageropt-2.4 {
1506558db80Sdrh  pagercount_sql {
1516558db80Sdrh    INSERT INTO t1 VALUES(randomblob(1500));
1526558db80Sdrh  }
1538c0a791aSdanielk1977} {0 5 3}
1546558db80Sdrh
1556558db80Sdrh# Appending a large quantity of data does not involve writing much
1566558db80Sdrh# to the journal file.
1576558db80Sdrh#
1586558db80Sdrhdo_test pageropt-3.1 {
1596558db80Sdrh  pagercount_sql {
1606558db80Sdrh    INSERT INTO t2 SELECT * FROM t1;
1616558db80Sdrh  }
1628c0a791aSdanielk1977} {1 7 2}
1636558db80Sdrh
1646558db80Sdrh# Once again, we do not need to read the last page of an overflow chain
1656558db80Sdrh# while deleting.
1666558db80Sdrh#
1676558db80Sdrhdo_test pageropt-3.2 {
1686558db80Sdrh  pagercount_sql {
1696558db80Sdrh    DROP TABLE t2;
1706558db80Sdrh  }
1718c0a791aSdanielk1977} {0 2 3}
1726558db80Sdrhdo_test pageropt-3.3 {
1736558db80Sdrh  pagercount_sql {
1746558db80Sdrh    DELETE FROM t1;
1756558db80Sdrh  }
1768c0a791aSdanielk1977} {0 3 3}
1776558db80Sdrh
1786558db80Sdrh# There are now 11 pages on the freelist.  Move them all into an
1796558db80Sdrh# overflow chain by inserting a single large record.  Starting from
1806558db80Sdrh# a cold cache, only page 1, the root page of table t1, and the trunk
1816558db80Sdrh# of the freelist need to be read (3 pages).  And only those three
1826558db80Sdrh# pages need to be journalled.  But 13 pages need to be written:
1836558db80Sdrh# page1, the root page of table t1, and an 11 page overflow chain.
1846558db80Sdrh#
1856558db80Sdrhdo_test pageropt-4.1 {
1866558db80Sdrh  db close
1876558db80Sdrh  sqlite3 db test.db
1886558db80Sdrh  pagercount_sql {
1896558db80Sdrh    INSERT INTO t1 VALUES(randomblob(11300))
1906558db80Sdrh  }
1918c0a791aSdanielk1977} {3 13 3}
1926558db80Sdrh
1936558db80Sdrh# Now we delete that big entries starting from a cold cache and an
1946558db80Sdrh# empty freelist.  The first 10 of the 11 pages overflow chain have
1956558db80Sdrh# to be read, together with page1 and the root of the t1 table.  12
1966558db80Sdrh# reads total.  But only page1, the t1 root, and the trunk of the
1976558db80Sdrh# freelist need to be journalled and written back.
1986558db80Sdrh#
199c5d0bd90Sdrhdo_test pageropt-4.2 {
2006558db80Sdrh  db close
2016558db80Sdrh  sqlite3 db test.db
2026558db80Sdrh  pagercount_sql {
2036558db80Sdrh    DELETE FROM t1
2046558db80Sdrh  }
2058c0a791aSdanielk1977} {12 3 3}
2066558db80Sdrh
207c1a60c51Sdansqlite3_soft_heap_limit $cmdlinearg(soft-heap-limit)
208538f570cSdrhcatch {db2 close}
209538f570cSdrhfinish_test
210