1 # Run this TCL script using an SQLite-enabled TCL interpreter to get a report
2 # on how much disk space is used by a particular data to actually store data
3 # versus how much space is unused.
5 # The dbstat virtual table is required.
10 # Argument $tname is the name of a table within the database opened by
11 # database handle [db]. Return true if it is a WITHOUT ROWID table, or
14 proc is_without_rowid
{tname
} {
15 set t
[string map
{' ''
} $tname]
16 db
eval "PRAGMA index_list = '$t'" o
{
17 if {$o(origin
) == "pk"} {
19 if {0==[db one
{ SELECT count
(*) FROM sqlite_schema WHERE name
=$n }]} {
27 # Read and run TCL commands from standard input. Used to implement
32 while {![eof stdin
]} {
39 append line
[gets stdin
]
40 if {[info complete
$line]} {
41 if {[catch {uplevel #0 $line} result]} {
42 puts stderr
"Error: $result"
43 } elseif
{$result!=""} {
54 # Get the name of the database to analyze
57 set argv0
[file rootname
[file tail
[info nameofexecutable
]]]
58 puts stderr
"Usage: $argv0 ?--pageinfo? ?--stats? database-filename"
60 Analyze the SQLite3 database
file specified by the
"database-filename"
61 argument and output a report detailing size and storage efficiency
62 information
for the database and its constituent tables and indexes.
66 --pageinfo Show how each page of the database-file is used
68 --stats Output SQL
text that creates a new database containing
69 statistics about the database that was analyzed
71 --tclsh Run the built-in TCL interpreter interactively
(for debugging
)
73 --version Show the version number of SQLite
77 set file_to_analyze
{}
78 set flags
(-pageinfo) 0
83 if {[regexp {^
-+pageinfo
$} $arg]} {
84 set flags
(-pageinfo) 1
85 } elseif
{[regexp {^
-+stats
$} $arg]} {
87 } elseif
{[regexp {^
-+debug
$} $arg]} {
89 } elseif
{[regexp {^
-+tclsh
$} $arg]} {
92 } elseif
{[regexp {^
-+version
$} $arg]} {
94 puts [mem one
{SELECT sqlite_version
()||' '||sqlite_source_id
()}]
97 } elseif
{[regexp {^
-} $arg]} {
98 puts stderr
"Unknown option: $arg"
100 } elseif
{$file_to_analyze!=""} {
103 set file_to_analyze
$arg
106 if {$file_to_analyze==""} usage
107 set root_filename
$file_to_analyze
108 regexp {^
file:(//)?
([^?
]*)} $file_to_analyze all x1 root_filename
109 if {![file exists
$root_filename]} {
110 puts stderr
"No such file: $root_filename"
113 if {![file readable
$root_filename]} {
114 puts stderr
"File is not readable: $root_filename"
117 set true_file_size
[file size
$root_filename]
118 if {$true_file_size<512} {
119 puts stderr
"Empty or malformed database: $root_filename"
123 # Compute the total file size assuming test_multiplexor is being used.
124 # Assume that SQLITE_ENABLE_8_3_NAMES might be enabled
126 set extension
[file extension
$root_filename]
127 set pattern
$root_filename
128 append pattern
{[0-3][0-9][0-9]}
129 foreach f
[glob -nocomplain $pattern] {
130 incr true_file_size
[file size
$f]
133 if {[string length
$extension]>=2 && [string length
$extension]<=4} {
134 set pattern
[file rootname
$root_filename]
135 append pattern
{.
[0-3][0-9][0-9]}
136 foreach f
[glob -nocomplain $pattern] {
137 incr true_file_size
[file size
$f]
143 if {[catch {sqlite3 db
$file_to_analyze -uri 1} msg
]} {
144 puts stderr
"error trying to open $file_to_analyze: $msg"
147 if {$flags(-debug)} {
148 proc dbtrace
{txt
} {puts $txt; flush stdout
;}
152 # Make sure all required compile-time options are available
154 if {![db exists
{SELECT
1 FROM pragma_compile_options
155 WHERE compile_options
='ENABLE_DBSTAT_VTAB'
}]} {
156 puts "The SQLite database engine linked with this application\
157 lacks required capabilities. Recompile using the\
158 -DSQLITE_ENABLE_DBSTAT_VTAB compile-time option to fix\
163 db
eval {SELECT count
(*) FROM sqlite_schema
}
164 set pageSize
[expr {wide
([db one
{PRAGMA page_size
}])}]
166 if {$flags(-pageinfo)} {
167 db
eval {CREATE VIRTUAL TABLE temp.stat USING dbstat
}
168 db
eval {SELECT name
, path
, pageno FROM temp.stat ORDER BY pageno
} {
169 puts "$pageno $name $path"
173 if {$flags(-stats)} {
174 db
eval {CREATE VIRTUAL TABLE temp.stat USING dbstat
}
176 puts "CREATE TABLE stats("
177 puts " name STRING, /* Name of table or index */"
178 puts " path INTEGER, /* Path to page from root */"
179 puts " pageno INTEGER, /* Page number */"
180 puts " pagetype STRING, /* 'internal', 'leaf' or 'overflow' */"
181 puts " ncell INTEGER, /* Cells on page (0 for overflow) */"
182 puts " payload INTEGER, /* Bytes of payload on this page */"
183 puts " unused INTEGER, /* Bytes of unused space on this page */"
184 puts " mx_payload INTEGER, /* Largest payload size of all cells */"
185 puts " pgoffset INTEGER, /* Offset of page in file */"
186 puts " pgsize INTEGER /* Size of the page */"
188 db
eval {SELECT quote
(name
) || '
,' ||
189 quote
(path
) || '
,' ||
190 quote
(pageno
) || '
,' ||
191 quote
(pagetype
) || '
,' ||
192 quote
(ncell
) || '
,' ||
193 quote
(payload
) || '
,' ||
194 quote
(unused
) || '
,' ||
195 quote
(mx_payload
) || '
,' ||
196 quote
(pgoffset
) || '
,' ||
197 quote
(pgsize
) AS x FROM stat
} {
198 puts "INSERT INTO stats VALUES($x);"
205 # In-memory database for collecting statistics. This script loops through
206 # the tables and indices in the database being analyzed, adding a row for each
207 # to an in-memory database (for which the schema is shown below). It then
208 # queries the in-memory db to produce the space-analysis report.
211 if {$flags(-debug)} {
212 proc dbtrace
{txt
} {puts $txt; flush stdout
;}
215 set tabledef
{CREATE TABLE space_used
(
216 name clob
, -- Name of a table or index in the database
file
217 tblname clob
, -- Name of associated table
218 is_index boolean
, -- TRUE
if it is an index
, false
for a table
219 is_without_rowid boolean
, -- TRUE
if WITHOUT ROWID table
220 nentry int
, -- Number of entries in the BTree
221 leaf_entries int
, -- Number of leaf entries
222 depth int
, -- Depth of the b-tree
223 payload int
, -- Total amount of data stored in this table or index
224 ovfl_payload int
, -- Total amount of data stored on overflow pages
225 ovfl_cnt int
, -- Number of entries that use overflow
226 mx_payload int
, -- Maximum payload size
227 int_pages int
, -- Number of interior pages used
228 leaf_pages int
, -- Number of leaf pages used
229 ovfl_pages int
, -- Number of overflow pages used
230 int_unused int
, -- Number of unused bytes on interior pages
231 leaf_unused int
, -- Number of unused bytes on primary pages
232 ovfl_unused int
, -- Number of unused bytes on overflow pages
233 gap_cnt int
, -- Number of gaps in the page layout
234 compressed_size int
-- Total bytes stored on disk
238 # Create a temporary "dbstat" virtual table.
240 db
eval {CREATE VIRTUAL TABLE temp.stat USING dbstat
}
241 db
eval {CREATE TEMP TABLE dbstat AS SELECT
* FROM temp.stat
243 db
eval {DROP TABLE temp.stat
}
246 set compressOverhead
0
248 set sql
{ SELECT name
, tbl_name FROM sqlite_schema WHERE rootpage
>0 }
249 foreach {name tblname
} [concat sqlite_schema sqlite_schema
[db
eval $sql]] {
251 set is_index
[expr {$name!=$tblname}]
252 set is_without_rowid
[is_without_rowid
$name]
255 sum
(ncell
) AS nentry
,
256 sum
((pagetype
=='leaf'
)*ncell
) AS leaf_entries
,
257 sum
(payload
) AS payload
,
258 sum
((pagetype
=='overflow'
) * payload
) AS ovfl_payload
,
259 sum
(path LIKE '
%+000000'
) AS ovfl_cnt
,
260 max
(mx_payload
) AS mx_payload
,
261 sum
(pagetype
=='internal'
) AS int_pages
,
262 sum
(pagetype
=='leaf'
) AS leaf_pages
,
263 sum
(pagetype
=='overflow'
) AS ovfl_pages
,
264 sum
((pagetype
=='internal'
) * unused
) AS int_unused
,
265 sum
((pagetype
=='leaf'
) * unused
) AS leaf_unused
,
266 sum
((pagetype
=='overflow'
) * unused
) AS ovfl_unused
,
267 sum
(pgsize
) AS compressed_size
,
268 max
((length
(CASE WHEN path LIKE '
%+%' THEN '' ELSE path END
)+3)/4)
270 FROM temp.dbstat WHERE name
= $name
273 set total_pages
[expr {$leaf_pages+$int_pages+$ovfl_pages}]
274 set storage
[expr {$total_pages*$pageSize}]
275 if {!$isCompressed && $storage>$compressed_size} {
277 set compressOverhead
14
280 # Column 'gap_cnt' is set to the number of non-contiguous entries in the
281 # list of pages visited if the b-tree structure is traversed in a top-down
282 # fashion (each node visited before its child-tree is passed). Any overflow
283 # chains present are traversed from start to finish before any child-tree
289 SELECT pageno
, pagetype FROM temp.dbstat
293 if {$prev>0 && $pagetype=="leaf" && $pageno!=$prev+1} {
299 INSERT INTO space_used VALUES
(
323 proc integerify
{real
} {
324 if {[string is double
-strict $real]} {
325 return [expr {wide
($real)}]
330 mem function int integerify
332 # Quote a string for use in an SQL query. Examples:
334 # [quote {hello world}] == {'hello world'}
335 # [quote {hello world's}] == {'hello world''s'}
338 return [string map
{' ''
} $txt]
341 # Output a title line
343 proc titleline
{title
} {
345 puts [string repeat
* 79]
347 set len
[string length
$title]
348 set stars
[string repeat
* [expr 79-$len-5]]
349 puts "*** $title $stars"
353 # Generate a single line of output in the statistics section of the
356 proc statline
{title value
{extra
{}}} {
357 set len
[string length
$title]
358 set dots
[string repeat .
[expr 50-$len]]
359 set len
[string length
$value]
360 set sp2
[string range
{ } $len end
]
364 puts "$title$dots $value$sp2$extra"
367 # Generate a formatted percentage value for $num/$denom
369 proc percent
{num denom
{of
{}}} {
370 if {$denom==0.0} {return ""}
371 set v
[expr {$num*100.0/$denom}]
373 if {$v==100.0 ||
$v<0.001 ||
($v>1.0 && $v<99.0)} {
374 return [format {%5.1f
%% %s
} $v $of]
375 } elseif
{$v<0.1 ||
$v>99.9} {
376 return [format {%7.3f
%% %s
} $v $of]
378 return [format {%6.2f
%% %s
} $v $of]
382 proc divide
{num denom
} {
383 if {$denom==0} {return 0.0}
384 return [format %.2f
[expr double
($num)/double
($denom)]]
387 # Generate a subreport that covers some subset of the database.
388 # the $where clause determines which subset to analyze.
390 proc subreport
{title where showFrag
} {
391 global pageSize file_pgcnt compressOverhead
393 # Query the in-memory database for the sum of various statistics
394 # for the subset of tables/indices identified by the WHERE clause in
395 # $where. Note that even if the WHERE clause matches no rows, the
396 # following query returns exactly one row (because it is an aggregate).
398 # The results of the query are stored directly by SQLite into local
399 # variables (i.e. $nentry, $payload etc.).
404 CASE WHEN (is_without_rowid OR is_index) THEN nentry
408 int(sum(payload)) AS payload,
409 int(sum(ovfl_payload)) AS ovfl_payload,
410 max(mx_payload) AS mx_payload,
411 int(sum(ovfl_cnt)) as ovfl_cnt,
412 int(sum(leaf_pages)) AS leaf_pages,
413 int(sum(int_pages)) AS int_pages,
414 int(sum(ovfl_pages)) AS ovfl_pages,
415 int(sum(leaf_unused)) AS leaf_unused,
416 int(sum(int_unused)) AS int_unused,
417 int(sum(ovfl_unused)) AS ovfl_unused,
418 int(sum(gap_cnt)) AS gap_cnt,
419 int(sum(compressed_size)) AS compressed_size,
420 int(max(depth)) AS depth,
422 FROM space_used WHERE $where" {} {}
424 # Output the sub-report title, nicely decorated with * characters.
430 # Calculate statistics and store the results in TCL variables, as follows:
432 # total_pages: Database pages consumed.
433 # total_pages_percent: Pages consumed as a percentage of the file.
434 # storage: Bytes consumed.
435 # payload_percent: Payload bytes used as a percentage of $storage.
436 # total_unused: Unused bytes on pages.
437 # avg_payload: Average payload per btree entry.
438 # avg_fanout: Average fanout for internal pages.
439 # avg_unused: Average unused bytes per btree entry.
440 # avg_meta: Average metadata overhead per entry.
441 # ovfl_cnt_percent: Percentage of btree entries that use overflow pages.
443 set total_pages
[expr {$leaf_pages+$int_pages+$ovfl_pages}]
444 set total_pages_percent
[percent
$total_pages $file_pgcnt]
445 set storage
[expr {$total_pages*$pageSize}]
446 set payload_percent
[percent
$payload $storage {of storage consumed
}]
447 set total_unused
[expr {$ovfl_unused+$int_unused+$leaf_unused}]
448 set avg_payload
[divide
$payload $nentry]
449 set avg_unused
[divide
$total_unused $nentry]
450 set total_meta
[expr {$storage - $payload - $total_unused}]
451 set total_meta
[expr {$total_meta + 4*($ovfl_pages - $ovfl_cnt)}]
452 set meta_percent
[percent
$total_meta $storage {of metadata
}]
453 set avg_meta
[divide
$total_meta $nentry]
455 # TODO: Is this formula correct?
457 SELECT count(*) FROM (
458 SELECT DISTINCT tblname FROM space_used WHERE $where AND is_index=0
461 set avg_fanout
[mem
eval "
462 SELECT (sum(leaf_pages+int_pages)-$nTab)/sum(int_pages) FROM space_used
465 set avg_fanout
[format %.2f
$avg_fanout]
467 set ovfl_cnt_percent
[percent
$ovfl_cnt $nentry {of all entries
}]
469 # Print out the sub-report statistics.
471 statline
{Percentage of total database
} $total_pages_percent
472 statline
{Number of entries
} $nentry
473 statline
{Bytes of storage consumed
} $storage
474 if {$compressed_size!=$storage} {
475 set compressed_size
[expr {$compressed_size+$compressOverhead*$total_pages}]
476 set pct
[expr {$compressed_size*100.0/$storage}]
477 set pct
[format {%5.1f
%%} $pct]
478 statline
{Bytes used
after compression
} $compressed_size $pct
480 statline
{Bytes of payload
} $payload $payload_percent
481 statline
{Bytes of metadata
} $total_meta $meta_percent
482 if {$cnt==1} {statline
{B-tree depth
} $depth}
483 statline
{Average payload per
entry} $avg_payload
484 statline
{Average unused bytes per
entry} $avg_unused
485 statline
{Average metadata per
entry} $avg_meta
486 if {[info exists avg_fanout
]} {
487 statline
{Average fanout
} $avg_fanout
489 if {$showFrag && $total_pages>1} {
490 set fragmentation
[percent
$gap_cnt [expr {$total_pages-1}]]
491 statline
{Non-sequential pages
} $gap_cnt $fragmentation
493 statline
{Maximum payload per
entry} $mx_payload
494 statline
{Entries that use overflow
} $ovfl_cnt $ovfl_cnt_percent
496 statline
{Index pages used
} $int_pages
498 statline
{Primary pages used
} $leaf_pages
499 statline
{Overflow pages used
} $ovfl_pages
500 statline
{Total pages used
} $total_pages
502 set int_unused_percent
[
503 percent
$int_unused [expr {$int_pages*$pageSize}] {of index space
}]
504 statline
"Unused bytes on index pages" $int_unused $int_unused_percent
506 statline
"Unused bytes on primary pages" $leaf_unused [
507 percent
$leaf_unused [expr {$leaf_pages*$pageSize}] {of primary space
}]
508 statline
"Unused bytes on overflow pages" $ovfl_unused [
509 percent
$ovfl_unused [expr {$ovfl_pages*$pageSize}] {of overflow space
}]
510 statline
"Unused bytes on all pages" $total_unused [
511 percent
$total_unused $storage {of all space
}]
515 # Calculate the overhead in pages caused by auto-vacuum.
517 # This procedure calculates and returns the number of pages used by the
518 # auto-vacuum 'pointer-map'. If the database does not support auto-vacuum,
519 # then 0 is returned. The two arguments are the size of the database file in
520 # pages and the page size used by the database (in bytes).
521 proc autovacuum_overhead
{filePages pageSize
} {
523 # Set $autovacuum to non-zero for databases that support auto-vacuum.
524 set autovacuum
[db one
{PRAGMA auto_vacuum
}]
526 # If the database is not an auto-vacuum database or the file consists
527 # of one page only then there is no overhead for auto-vacuum. Return zero.
528 if {0==$autovacuum ||
$filePages==1} {
532 # The number of entries on each pointer map page. The layout of the
533 # database file is one pointer-map page, followed by $ptrsPerPage other
534 # pages, followed by a pointer-map page etc. The first pointer-map page
535 # is the second page of the file overall.
536 set ptrsPerPage
[expr double
($pageSize/5)]
538 # Return the number of pointer map pages in the database.
539 return [expr wide
(ceil
( ($filePages-1.0)/($ptrsPerPage+1.0) ))]
543 # Calculate the summary statistics for the database and store the results
544 # in TCL variables. They are output below. Variables are as follows:
546 # pageSize: Size of each page in bytes.
547 # file_bytes: File size in bytes.
548 # file_pgcnt: Number of pages in the file.
549 # file_pgcnt2: Number of pages in the file (calculated).
550 # av_pgcnt: Pages consumed by the auto-vacuum pointer-map.
551 # av_percent: Percentage of the file consumed by auto-vacuum pointer-map.
552 # inuse_pgcnt: Data pages in the file.
553 # inuse_percent: Percentage of pages used to store data.
554 # free_pgcnt: Free pages calculated as (<total pages> - <in-use pages>)
555 # free_pgcnt2: Free pages in the file according to the file header.
556 # free_percent: Percentage of file consumed by free pages (calculated).
557 # free_percent2: Percentage of file consumed by free pages (header).
558 # ntable: Number of tables in the db.
559 # nindex: Number of indices in the db.
560 # nautoindex: Number of indices created automatically.
561 # nmanindex: Number of indices created manually.
562 # user_payload: Number of bytes of payload in table btrees
563 # (not including sqlite_schema)
564 # user_percent: $user_payload as a percentage of total file size.
566 ### The following, setting $file_bytes based on the actual size of the file
567 ### on disk, causes this tool to choke on zipvfs databases. So set it based
568 ### on the return of [PRAGMA page_count] instead.
570 set file_bytes
[file size
$file_to_analyze]
571 set file_pgcnt
[expr {$file_bytes/$pageSize}]
573 set file_pgcnt
[db one
{PRAGMA page_count
}]
574 set file_bytes
[expr {$file_pgcnt * $pageSize}]
576 set av_pgcnt
[autovacuum_overhead
$file_pgcnt $pageSize]
577 set av_percent
[percent
$av_pgcnt $file_pgcnt]
579 set sql
{SELECT sum
(leaf_pages
+int_pages
+ovfl_pages
) FROM space_used
}
580 set inuse_pgcnt
[expr wide
([mem
eval $sql])]
581 set inuse_percent
[percent
$inuse_pgcnt $file_pgcnt]
583 set free_pgcnt
[expr {$file_pgcnt-$inuse_pgcnt-$av_pgcnt}]
584 if {$file_bytes>1073741824 && $free_pgcnt>0} {incr free_pgcnt
-1}
585 set free_percent
[percent
$free_pgcnt $file_pgcnt]
586 set free_pgcnt2
[db one
{PRAGMA freelist_count
}]
587 set free_percent2
[percent
$free_pgcnt2 $file_pgcnt]
589 set file_pgcnt2
[expr {$inuse_pgcnt+$free_pgcnt2+$av_pgcnt}]
591 # Account for the lockbyte page
592 if {$file_pgcnt2*$pageSize>1073742335} {incr file_pgcnt2
}
594 set ntable
[db
eval {SELECT count
(*)+1 FROM sqlite_schema WHERE type
='table'
}]
595 set nindex
[db
eval {SELECT count
(*) FROM sqlite_schema WHERE type
='index'
}]
596 set sql
{SELECT count
(*) FROM sqlite_schema WHERE name LIKE 'sqlite_autoindex
%'
}
597 set nautoindex
[db
eval $sql]
598 set nmanindex
[expr {$nindex-$nautoindex}]
600 # set total_payload [mem eval "SELECT sum(payload) FROM space_used"]
601 set user_payload
[mem one
{SELECT int
(sum
(payload
)) FROM space_used
602 WHERE NOT is_index AND name NOT LIKE 'sqlite_schema'
}]
603 set user_percent
[percent
$user_payload $file_bytes]
605 # Output the summary statistics calculated above.
607 puts "/** Disk-Space Utilization Report For $root_filename"
609 statline
{Page size in bytes
} $pageSize
610 statline
{Pages in the whole
file (measured
)} $file_pgcnt
611 statline
{Pages in the whole
file (calculated
)} $file_pgcnt2
612 statline
{Pages that store data
} $inuse_pgcnt $inuse_percent
613 statline
{Pages on the freelist
(per header
)} $free_pgcnt2 $free_percent2
614 statline
{Pages on the freelist
(calculated
)} $free_pgcnt $free_percent
615 statline
{Pages of auto-vacuum overhead
} $av_pgcnt $av_percent
616 statline
{Number of tables in the database
} $ntable
617 statline
{Number of indices
} $nindex
618 statline
{Number of defined indices
} $nmanindex
619 statline
{Number of implied indices
} $nautoindex
621 statline
{Size of uncompressed content in bytes
} $file_bytes
622 set efficiency
[percent
$true_file_size $file_bytes]
623 statline
{Size of compressed
file on disk
} $true_file_size $efficiency
625 statline
{Size of the
file in bytes
} $file_bytes
627 statline
{Bytes of user payload stored
} $user_payload $user_percent
629 # Output table rankings
632 titleline
"Page counts for all tables with their indices"
634 mem
eval {SELECT tblname
, count
(*) AS cnt
,
635 int
(sum
(int_pages
+leaf_pages
+ovfl_pages
)) AS size
636 FROM space_used GROUP BY tblname ORDER BY size
+0 DESC
, tblname
} {} {
637 statline
[string toupper
$tblname] $size [percent
$size $file_pgcnt]
640 titleline
"Page counts for all tables and indices separately"
645 int
(int_pages
+leaf_pages
+ovfl_pages
) AS size
647 ORDER BY size
+0 DESC
, name
} {} {
648 statline
$nm $size [percent
$size $file_pgcnt]
652 titleline
"Bytes of disk space used after compression"
655 mem
eval {SELECT tblname
,
656 int
(sum
(compressed_size
)) +
657 $compressOverhead*sum
(int_pages
+leaf_pages
+ovfl_pages
)
659 FROM space_used GROUP BY tblname ORDER BY csize
+0 DESC
, tblname
} {} {
661 statline
[string toupper
$tblname] $csize [percent
$csize $true_file_size]
663 set overhead
[expr {$true_file_size - $csum}]
665 statline
{Header and free space
} $overhead [percent
$overhead $true_file_size]
672 subreport
{All tables and indices
} 1 0
674 subreport
{All tables
} {NOT is_index
} 0
676 subreport
{All indices
} {is_index
} 0
678 foreach tbl
[mem
eval {SELECT DISTINCT tblname name FROM space_used
681 set name
[string toupper
$tbl]
682 set n
[mem
eval {SELECT count
(*) FROM space_used WHERE tblname
=$tbl}]
684 set idxlist
[mem
eval "SELECT name FROM space_used
685 WHERE tblname='$qn' AND is_index
687 subreport
"Table $name and all its indices" "tblname='$qn'" 0
688 subreport
"Table $name w/o any indices" "name='$qn'" 1
689 if {[llength $idxlist]>1} {
690 subreport
"Indices of table $name" "tblname='$qn' AND is_index" 0
692 foreach idx
$idxlist {
693 set qidx
[quote
$idx]
694 subreport
"Index [string toupper $idx] of table $name" "name='$qidx'" 1
697 subreport
"Table $name" "name='$qn'" 1
701 # Output instructions on what the numbers above mean.
704 titleline Definitions
708 The number of bytes in a single page of the database
file.
711 Number of pages in the whole
file
713 puts " The number of $pageSize-byte pages that go into forming the complete
716 Pages that store data
718 The number of pages that store data
, either as primary B
*Tree pages or
719 as overflow pages. The number at the right is the data pages divided by
720 the total number of pages in the
file.
722 Pages on the freelist
724 The number of pages that are not currently in use but are reserved
for
725 future use. The percentage at the right is the number of freelist pages
726 divided by the total number of pages in the
file.
728 Pages of auto-vacuum overhead
730 The number of pages that store data used by the database to facilitate
731 auto-vacuum. This is zero
for databases that do not support auto-vacuum.
733 Number of tables in the database
735 The number of tables in the database
, including the SQLITE_SCHEMA table
736 used to store schema information.
740 The total number of indices in the database.
742 Number of defined indices
744 The number of indices created using an explicit CREATE INDEX statement.
746 Number of implied indices
748 The number of indices used to implement PRIMARY KEY or UNIQUE constraints
751 Size of the
file in bytes
753 The total amount of disk space used by the entire database files.
755 Bytes of user payload stored
757 The total number of bytes of user payload stored in the database. The
758 schema information in the SQLITE_SCHEMA table is not counted when
759 computing this number. The percentage at the right shows the payload
760 divided by the total
file size.
762 Percentage of total database
764 The amount of the complete database
file that is devoted to storing
765 information described by this category.
769 The total number of B-Tree key
/value pairs stored under this category.
771 Bytes of storage consumed
773 The total amount of disk space required to store all B-Tree entries
774 under this category. The is the total number of pages used times
779 The amount of payload stored under this category. Payload is the data
780 part of table entries and the key part of index entries. The percentage
781 at the right is the bytes of payload divided by the bytes of storage
786 The amount of formatting and structural information stored in the
787 table or index. Metadata includes the btree page header
, the cell pointer
788 array, the size field
for each cell
, the left child pointer or non-leaf
789 cells
, the overflow pointers
for overflow cells
, and the rowid value
for
790 rowid table cells. In other words
, metadata is everything that is neither
791 unused space nor content. The record header in the payload is counted as
792 content
, not metadata.
794 Average payload per
entry
796 The average amount of payload on each
entry. This is just the bytes of
797 payload divided by the number of entries.
799 Average unused bytes per
entry
801 The average amount of free space remaining on all pages under this
802 category on a per-entry basis. This is the number of unused bytes on
803 all pages divided by the number of entries.
807 The number of pages in the table or index that are out of sequence.
808 Many filesystems are optimized
for sequential
file access so a small
809 number of non-sequential pages might result in faster queries
,
810 especially
for larger database files that do not fit in the disk cache.
811 Note that
after running VACUUM
, the root page of each table or index is
812 at the beginning of the database
file and all other pages are in a
813 separate part of the database
file, resulting in a single non-
816 Maximum payload per
entry
818 The largest payload size of any
entry.
820 Entries that use overflow
822 The number of entries that user one or more overflow pages.
826 This is the number of pages used to hold all information in the current
827 category. This is the sum of index
, primary
, and overflow pages.
831 This is the number of pages in a table B-tree that hold only key
(rowid
)
832 information and no data.
836 This is the number of B-tree pages that hold both key and data.
840 The total number of overflow pages used
for this category.
842 Unused bytes on index pages
844 The total number of bytes of unused space on all index pages. The
845 percentage at the right is the number of unused bytes divided by the
846 total number of bytes on index pages.
848 Unused bytes on primary pages
850 The total number of bytes of unused space on all primary pages. The
851 percentage at the right is the number of unused bytes divided by the
852 total number of bytes on primary pages.
854 Unused bytes on overflow pages
856 The total number of bytes of unused space on all overflow pages. The
857 percentage at the right is the number of unused bytes divided by the
858 total number of bytes on overflow pages.
860 Unused bytes on all pages
862 The total number of bytes of unused space on all primary and overflow
863 pages. The percentage at the right is the number of unused bytes
864 divided by the total number of bytes.
867 # Output a dump of the in-memory database. This can be used for more
868 # complex offline analysis.
871 puts "The entire text of this report can be sourced into any SQL database"
872 puts "engine for further analysis. All of the text above is an SQL comment."
873 puts "The data used to generate this report follows:"
878 mem
eval {SELECT
* FROM space_used
} x
{
879 puts -nonewline "INSERT INTO space_used VALUES"
883 if {$v=="" ||
![string is double
$v]} {set v '
[quote
$v]'
}
884 puts -nonewline $sep$v