libdb/lang/sql/sqlite/test/e_vacuum.test

# 2010 September 24
#
# The author disclaims copyright to this source code.  In place of
# a legal notice, here is a blessing:
#
#    May you do good and not evil.
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# This file implements tests to verify that the "testable statements" in 
# the lang_vacuum.html document are correct.
#

set testdir [file dirname $argv0]
source $testdir/tester.tcl

sqlite3_test_control_pending_byte 0x1000000

proc create_db {{sql ""}} {
  catch { db close }
  forcedelete test.db
  sqlite3 db test.db

  db transaction {
    execsql { PRAGMA page_size = 1024; }
    execsql $sql
    execsql {
      CREATE TABLE t1(a PRIMARY KEY, b UNIQUE);
      INSERT INTO t1 VALUES(1, randomblob(400));
      INSERT INTO t1 SELECT a+1,  randomblob(400) FROM t1;
      INSERT INTO t1 SELECT a+2,  randomblob(400) FROM t1;
      INSERT INTO t1 SELECT a+4,  randomblob(400) FROM t1;
      INSERT INTO t1 SELECT a+8,  randomblob(400) FROM t1;
      INSERT INTO t1 SELECT a+16, randomblob(400) FROM t1;
      INSERT INTO t1 SELECT a+32, randomblob(400) FROM t1;
      INSERT INTO t1 SELECT a+64, randomblob(400) FROM t1;

      CREATE TABLE t2(a PRIMARY KEY, b UNIQUE);
      INSERT INTO t2 SELECT * FROM t1;
    }
  }

  return [expr {[file size test.db] / 1024}]
}

# This proc returns the number of contiguous blocks of pages that make up
# the table or index named by the only argument. For example, if the table
# occupies database pages 3, 4, 8 and 9, then this command returns 2 (there
# are 2 fragments - one consisting of pages 3 and 4, the other of fragments
# 8 and 9).
#
proc fragment_count {name} {
  execsql { CREATE VIRTUAL TABLE temp.stat USING dbstat }
  set nFrag 1
  db eval {SELECT pageno FROM stat WHERE name = 't1' ORDER BY pageno} {
    if {[info exists prevpageno] && $prevpageno != $pageno-1} {
      incr nFrag
    }
    set prevpageno $pageno
  }
  execsql { DROP TABLE temp.stat }
  set nFrag
}


# EVIDENCE-OF: R-63707-33375 -- syntax diagram vacuum-stmt
#
do_execsql_test e_vacuum-0.1 { VACUUM } {}

# EVIDENCE-OF: R-51469-36013 Unless SQLite is running in
# "auto_vacuum=FULL" mode, when a large amount of data is deleted from
# the database file it leaves behind empty space, or "free" database
# pages.
#
# EVIDENCE-OF: R-60541-63059 Running VACUUM to rebuild the database
# reclaims this space and reduces the size of the database file.
#
foreach {tn avmode sz} {
  1 none        7 
  2 full        8 
  3 incremental 8
} {
  set nPage [create_db "PRAGMA auto_vacuum = $avmode"]

  do_execsql_test e_vacuum-1.1.$tn.1 {
    DELETE FROM t1;
    DELETE FROM t2;
  } {}

  if {$avmode == "full"} {
    # This branch tests the "unless ... auto_vacuum=FULL" in the requirement
    # above. If auto_vacuum is set to FULL, then no empty space is left in
    # the database file.
    do_execsql_test e_vacuum-1.1.$tn.2 {PRAGMA freelist_count} 0
  } else {
    set freelist [expr {$nPage - $sz}]
    if {$avmode == "incremental"} { 
      # The page size is 1024 bytes. Therefore, assuming the database contains
      # somewhere between 207 and 411 pages (it does), there are 2 pointer-map
      # pages.
      incr freelist -2
    }
    do_execsql_test e_vacuum-1.1.$tn.3 {PRAGMA freelist_count} $freelist
    do_execsql_test e_vacuum-1.1.$tn.4 {VACUUM} {}
  }

  do_test e_vacuum-1.1.$tn.5 { expr {[file size test.db] / 1024} } $sz
}

# EVIDENCE-OF: R-50943-18433 Frequent inserts, updates, and deletes can
# cause the database file to become fragmented - where data for a single
# table or index is scattered around the database file.
#
# EVIDENCE-OF: R-05791-54928 Running VACUUM ensures that each table and
# index is largely stored contiguously within the database file.
#
#   e_vacuum-1.2.1 - Perform many INSERT, UPDATE and DELETE ops on table t1.
#   e_vacuum-1.2.2 - Verify that t1 and its indexes are now quite fragmented.
#   e_vacuum-1.2.3 - Run VACUUM.
#   e_vacuum-1.2.4 - Verify that t1 and its indexes are now much 
#                    less fragmented.
#
ifcapable vtab {
  create_db 
  register_dbstat_vtab db
  do_execsql_test e_vacuum-1.2.1 {
    DELETE FROM t1 WHERE a%2;
    INSERT INTO t1 SELECT b, a FROM t2 WHERE a%2;
    UPDATE t1 SET b=randomblob(600) WHERE (a%2)==0;
  } {}
  
  do_test e_vacuum-1.2.2.1 { expr [fragment_count t1]>100 } 1
  do_test e_vacuum-1.2.2.2 { expr [fragment_count sqlite_autoindex_t1_1]>100 } 1
  do_test e_vacuum-1.2.2.3 { expr [fragment_count sqlite_autoindex_t1_2]>100 } 1
  
  do_execsql_test e_vacuum-1.2.3 { VACUUM } {}
  
  # In practice, the tables and indexes each end up stored as two fragments -
  # one containing the root page and another containing all other pages.
  #
  do_test e_vacuum-1.2.4.1 { fragment_count t1 }                    2
  do_test e_vacuum-1.2.4.2 { fragment_count sqlite_autoindex_t1_1 } 2
  do_test e_vacuum-1.2.4.3 { fragment_count sqlite_autoindex_t1_2 } 2
}

# EVIDENCE-OF: R-20474-44465 Normally, the database page_size and
# whether or not the database supports auto_vacuum must be configured
# before the database file is actually created.
#
do_test e_vacuum-1.3.1.1 {
  create_db "PRAGMA page_size = 1024 ; PRAGMA auto_vacuum = FULL"
  execsql { PRAGMA page_size ; PRAGMA auto_vacuum }
} {1024 1}
do_test e_vacuum-1.3.1.2 {
  execsql { PRAGMA page_size = 2048 }
  execsql { PRAGMA auto_vacuum = NONE }
  execsql { PRAGMA page_size ; PRAGMA auto_vacuum }
} {1024 1}

# EVIDENCE-OF: R-08570-19916 However, when not in write-ahead log mode,
# the page_size and/or auto_vacuum properties of an existing database
# may be changed by using the page_size and/or pragma auto_vacuum
# pragmas and then immediately VACUUMing the database.
#
do_test e_vacuum-1.3.2.1 {
  execsql { PRAGMA journal_mode = delete }
  execsql { PRAGMA page_size = 2048 }
  execsql { PRAGMA auto_vacuum = NONE }
  execsql VACUUM
  execsql { PRAGMA page_size ; PRAGMA auto_vacuum }
} {2048 0}

# EVIDENCE-OF: R-48521-51450 When in write-ahead log mode, only the
# auto_vacuum support property can be changed using VACUUM.
#
ifcapable wal {
do_test e_vacuum-1.3.3.1 {
  execsql { PRAGMA journal_mode = wal }
  execsql { PRAGMA page_size ; PRAGMA auto_vacuum }
} {2048 0}
do_test e_vacuum-1.3.3.2 {
  execsql { PRAGMA page_size = 1024 }
  execsql { PRAGMA auto_vacuum = FULL }
  execsql VACUUM
  execsql { PRAGMA page_size ; PRAGMA auto_vacuum }
} {2048 1}
}

# EVIDENCE-OF: R-38001-03952 VACUUM only works on the main database. It
# is not possible to VACUUM an attached database file.
forcedelete test.db2
create_db { PRAGMA auto_vacuum = NONE }
do_execsql_test e_vacuum-2.1.1 {
  ATTACH 'test.db2' AS aux;
  PRAGMA aux.page_size = 1024;
  CREATE TABLE aux.t3 AS SELECT * FROM t1;
  DELETE FROM t3;
} {}
set original_size [file size test.db2]

# Try everything we can think of to get the aux database vacuumed:
do_execsql_test e_vacuum-2.1.3 { VACUUM } {}
do_execsql_test e_vacuum-2.1.4 { VACUUM aux } {}
do_execsql_test e_vacuum-2.1.5 { VACUUM 'test.db2' } {}

# Despite our efforts, space in the aux database has not been reclaimed:
do_test e_vacuum-2.1.6 { expr {[file size test.db2]==$::original_size} } 1

# EVIDENCE-OF: R-17495-17419 The VACUUM command may change the ROWIDs of
# entries in any tables that do not have an explicit INTEGER PRIMARY
# KEY.
#
#   Tests e_vacuum-3.1.1 - 3.1.2 demonstrate that rowids can change when
#   a database is VACUUMed. Tests e_vacuum-3.1.3 - 3.1.4 show that adding
#   an INTEGER PRIMARY KEY column to a table stops this from happening.
#
do_execsql_test e_vacuum-3.1.1 {
  CREATE TABLE t4(x);
  INSERT INTO t4(x) VALUES('x');
  INSERT INTO t4(x) VALUES('y');
  INSERT INTO t4(x) VALUES('z');
  DELETE FROM t4 WHERE x = 'y';
  SELECT rowid, x FROM t4;
} {1 x 3 z}
do_execsql_test e_vacuum-3.1.2 {
  VACUUM;
  SELECT rowid, x FROM t4;
} {1 x 2 z}

do_execsql_test e_vacuum-3.1.3 {
  CREATE TABLE t5(x, y INTEGER PRIMARY KEY);
  INSERT INTO t5(x) VALUES('x');
  INSERT INTO t5(x) VALUES('y');
  INSERT INTO t5(x) VALUES('z');
  DELETE FROM t5 WHERE x = 'y';
  SELECT rowid, x FROM t5;
} {1 x 3 z}
do_execsql_test e_vacuum-3.1.4 {
  VACUUM;
  SELECT rowid, x FROM t5;
} {1 x 3 z}

# EVIDENCE-OF: R-49563-33883 A VACUUM will fail if there is an open
# transaction, or if there are one or more active SQL statements when it
# is run.
#
do_execsql_test  e_vacuum-3.2.1.1 { BEGIN } {}
do_catchsql_test e_vacuum-3.2.1.2 { 
  VACUUM 
} {1 {cannot VACUUM from within a transaction}}
do_execsql_test  e_vacuum-3.2.1.3 { COMMIT } {}
do_execsql_test  e_vacuum-3.2.1.4 { VACUUM } {}
do_execsql_test  e_vacuum-3.2.1.5 { SAVEPOINT x } {}
do_catchsql_test e_vacuum-3.2.1.6 { 
  VACUUM 
} {1 {cannot VACUUM from within a transaction}}
do_execsql_test  e_vacuum-3.2.1.7 { COMMIT } {}
do_execsql_test  e_vacuum-3.2.1.8 { VACUUM } {}

create_db
do_test e_vacuum-3.2.2.1 {
  set res ""
  db eval { SELECT a FROM t1 } {
    if {$a == 10} { set res [catchsql VACUUM] }
  }
  set res
} {1 {cannot VACUUM - SQL statements in progress}}


# EVIDENCE-OF: R-38735-12540 As of SQLite version 3.1, an alternative to
# using the VACUUM command to reclaim space after data has been deleted
# is auto-vacuum mode, enabled using the auto_vacuum pragma.
#
do_test e_vacuum-3.3.1 {
  create_db { PRAGMA auto_vacuum = FULL }
  execsql { PRAGMA auto_vacuum }
} {1}

# EVIDENCE-OF: R-64844-34873 When auto_vacuum is enabled for a database
# free pages may be reclaimed after deleting data, causing the file to
# shrink, without rebuilding the entire database using VACUUM.
#
do_test e_vacuum-3.3.2.1 {
  create_db { PRAGMA auto_vacuum = FULL }
  execsql {
    DELETE FROM t1;
    DELETE FROM t2;
  }
  expr {[file size test.db] / 1024}
} {8}
do_test e_vacuum-3.3.2.2 {
  create_db { PRAGMA auto_vacuum = INCREMENTAL }
  execsql {
    DELETE FROM t1;
    DELETE FROM t2;
    PRAGMA incremental_vacuum;
  }
  expr {[file size test.db] / 1024}
} {8}

finish_test