Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
Overview
| Comment: | Edits to the VFS documentation. Better cross-referencing to VFS. |
|---|---|
| Downloads: | Tarball | ZIP archive | SQL archive |
| Timelines: | family | ancestors | descendants | both | trunk |
| Files: | files | file ages | folders |
| SHA1: |
e63c0ce4fb1d824ffcac9360ecc6f24c |
| User & Date: | drh 2011-05-23 19:04:56 |
Context
|
2011-05-24
| ||
| 00:44 | Restrict website search text to omit non-alphanumeric characters. (check-in: 2528553d5a user: drh tags: trunk) | |
|
2011-05-23
| ||
| 19:04 | Edits to the VFS documentation. Better cross-referencing to VFS. (check-in: e63c0ce4fb user: drh tags: trunk) | |
|
2011-05-20
| ||
| 23:21 | Further tightening of the validity rules for valid HTTP requests in althttpd.c. (check-in: 73bc374298 user: drh tags: trunk) | |
Changes
Changes to pages/atomiccommit.in.
83 83 never start writing a sector in the middle and work towards the 84 84 ends. We do not know if this assumption is always true but it 85 85 seems reasonable.</p> 86 86 87 87 <p>The previous paragraph states that SQLite does not assume that 88 88 sector writes are atomic. This is true by default. But as of 89 89 SQLite version 3.5.0, there is a new interface called the 90 -Virtual File System (VFS) interface. The VFS is the only means 90 +Virtual File System ([VFS]) interface. The [VFS] is the only means 91 91 by which SQLite communicates to the underlying filesystem. The 92 92 code comes with default VFS implementations for Unix and Windows 93 93 and there is a mechanism for creating new custom VFS implementations 94 94 at runtime. In this new VFS interface there is a method called 95 95 xDeviceCharacteristics. This method interrogates the underlying 96 96 filesystem to discover various properties and behaviors that the 97 97 filesystem may or may not exhibit. The xDeviceCharacteristics ................................................................................ 122 122 file space originally contains garbage and then later is filled in 123 123 with the data actually written. In other words, SQLite assumes that 124 124 the file size is updated before the file content. This is a 125 125 pessimistic assumption and SQLite has to do some extra work to make 126 126 sure that it does not cause database corruption if power is lost 127 127 between the time when the file size is increased and when the 128 128 new content is written. The xDeviceCharacteristics method of 129 -the VFS might indicate that the filesystem will always write the 129 +the [VFS] might indicate that the filesystem will always write the 130 130 data before updating the file size. (This is the 131 131 SQLITE_IOCAP_SAFE_APPEND property for those readers who are looking 132 132 at the code.) When the xDeviceCharacteristics method indicates 133 133 that files content is written before the file size is increased, 134 134 SQLite can forego some of its pedantic database protection steps 135 135 and thereby decrease the amount of disk I/O needed to perform a 136 136 commit. The current implementation, however, makes no such assumptions
Changes to pages/compile.in.
371 371 372 372 <li> No locking style. If none of the above can be supported, this 373 373 locking style is used. No database locking mechanism is used. When 374 374 this system is used it is not safe for a single database to be 375 375 accessed by multiple clients. 376 376 </ul> 377 377 378 - Additionally, five extra VFS implementations are provided as well as the 378 + Additionally, five extra [VFS] implementations are provided as well as the 379 379 default. By specifying one of the extra VFS implementations 380 380 when calling [sqlite3_open_v2()], an application may bypass the file-system 381 381 detection logic and explicitly select one of the above locking styles. The 382 - five extra VFS implementations are called "unix-posix", "unix-afp", 382 + five extra [VFS] implementations are called "unix-posix", "unix-afp", 383 383 "unix-flock", "unix-dotfile" and "unix-none". 384 384 } 385 385 386 386 COMPILE_OPTION {SQLITE_ENABLE_MEMORY_MANAGEMENT} { 387 387 This option adds extra logic to SQLite that allows it to release unused 388 388 memory upon request. This option must be enabled in order for the 389 389 [sqlite3_release_memory()] interface to work. If this compile-time
Changes to pages/fileformat2.in.
360 360 that contains the bytes at offsets between 1073741824 and 1073742335, 361 361 inclusive. A database file that is less than or equal to 1073741824 bytes 362 362 in size contains no lock-byte page. A database file larger than 363 363 1073741824 contains exactly one lock-byte page. 364 364 </p> 365 365 366 366 <p>The lock-byte page is set aside for use by the operating-system specific 367 -VFS implementation in implementing the database file locking primitives. 367 +[VFS] implementation in implementing the database file locking primitives. 368 368 ^SQLite does not use the lock-byte page. ^The SQLite core 369 -will never read or write, though operating-system specific VFS 369 +will never read or write, though operating-system specific [VFS] 370 370 implementations may choose to read or write bytes on the lock-byte 371 371 page according to the 372 372 needs and proclivities of the underlying system. The unix and win32 373 -VFS implementations that come built into SQLite do not write to the 373 +[VFS] implementations that come built into SQLite do not write to the 374 374 lock-byte page, but third-party VFS implementations for 375 375 other operating systems might.</p> 376 376 377 377 <tcl>hd_fragment {freelist} {freelist} {free-page list}</tcl> 378 378 <h3>1.4 The Freelist</h3> 379 379 380 380 <p>A database file might contain one or more pages that are not in
Changes to pages/pragma.in.
743 743 <p>^(<b>PRAGMA synchronous; 744 744 <br>PRAGMA synchronous = </b> 745 745 <i>0 | OFF | 1 | NORMAL | 2 | FULL</i><b>;</b></p> 746 746 747 747 <p>Query or change the setting of the "synchronous" flag.)^ 748 748 ^The first (query) form will return the synchronous setting as an 749 749 integer. ^When synchronous is FULL (2), the SQLite database engine will 750 - use the xSync method of the VFS to ensure that all content is safely 750 + use the xSync method of the [VFS] to ensure that all content is safely 751 751 written to the disk surface prior to continuing. 752 752 This ensures that an operating system crash or power failure will 753 753 not corrupt the database. 754 754 FULL synchronous is very safe, but it is also slower. 755 755 ^When synchronous is NORMAL (1), the SQLite database 756 756 engine will still sync at the most critical moments, but less often 757 757 than in FULL mode. There is a very small (though non-zero) chance that
Changes to pages/selfcontained.in.
33 33 The date and time SQL functions provided by SQLite require 34 34 some additional C library support, but those functions can 35 35 be also be omitted from the build using compile-time options. 36 36 </p> 37 37 38 38 <p> 39 39 Communications between SQLite and the operating system and disk are 40 -mediated through an interchangeable 41 -[sqlite3_vfs | Virtual File System (VFS)] layer. 40 +mediated through an interchangeable [VFS] layer. 42 41 VFS modules for Unix (Linux and Mac OS X), OS/2, and Windows (Win32 and WinCE) 43 42 are provided in the source tree. It is a simple matter to devise an 44 43 alternative VFS for embedded devices. 45 44 </p> 46 45 47 46 <p> 48 47 For safe operation in multi-threaded environments, SQLite requires
Changes to pages/testing.in.
296 296 course, and so crash testing is done in simulation. An alternative 297 297 [sqlite3_vfs | Virtual File System] is inserted that allows the test 298 298 harness to simulate the state of the database file following a crash.</p> 299 299 300 300 <p>In the TCL test harness, the crash simulation is done in a separate 301 301 process. The main testing process spawns a child process which runs 302 302 some SQLite operation and randomly crashes somewhere in the middle of 303 -a write operation. A special VFS randomly reorders and corrupts 303 +a write operation. A special [VFS] randomly reorders and corrupts 304 304 the unsynchronized 305 305 write operations to simulate the effect of buffered filesystems. After 306 306 the child dies, the original test process opens and reads the test 307 307 database and verifies that the changes attempted by the child either 308 308 completed successfully or else were completely rolled back. The 309 309 [integrity_check] [PRAGMA] is used to make sure no database corruption 310 310 occurs.</p> 311 311 312 312 <p>The TH3 test harness needs to run on embedded systems that do not 313 313 necessarily have the ability to spawn child processes, so it uses 314 -an in-memory VFS to simulate crashes. The in-memory VFS can be rigged 314 +an in-memory [VFS] to simulate crashes. The in-memory [VFS] can be rigged 315 315 to make a snapshot of the entire filesystem after a set number of I/O 316 316 operations. Crash tests run in a loop. On each iteration of the loop, 317 317 the point at which a snapshot is made is advanced until the SQLite 318 318 operations being tested run to completion without ever hitting a 319 319 snapshot. Within the loop, after the SQLite operation under test has 320 320 completed, the filesystem is reverted to the snapshot and random file 321 321 damage is introduced that is characteristic of the kinds of damage ................................................................................ 420 420 421 421 <p>The SQLite core has 100% branch test coverage under [TH3] as of 422 422 2009-07-25, in its default configuration as measured by 423 423 [http://gcc.gnu.org/onlinedocs/gcc/Gcov.html | gcov] 424 424 utility on SuSE Linux 10.1 on x86 hardware with the GCC 4.0.1 compiler.</p> 425 425 426 426 <p>The "SQLite core" in the previous paragraph excludes the 427 -operating-system dependent [sqlite3_vfs | VFS] backends, since it is 427 +operating-system dependent [VFS] backends, since it is 428 428 not possible to write cross-platform tests for those modules. Extensions 429 429 such as FTS3 and RTree are also excluded from the analysis.</p> 430 430 431 431 <tcl>hd_fragment stmtvbr</tcl> 432 432 <h3>7.1 Statement versus branch coverage</h3> 433 433 434 434 <p>There are many ways to measure test coverage. The most popular ................................................................................ 708 708 <tcl>hd_fragment journaltest</tcl> 709 709 <h3>8.5 Journal Tests</h3> 710 710 711 711 <p>One of the things that SQLite does to ensure that transactions 712 712 are atomic across system crashes and power failures is to write 713 713 all changes into the rollback journal file prior to changing the 714 714 database. The TCL test harness contains an alternative 715 -[sqlite3_vfs | Virtual File System ] implementation that helps to 715 +[OS backend] implementation that helps to 716 716 verify this is occurring correctly. The "journal-test VFS" monitors 717 717 all disk I/O traffic between the database file and rollback journal, 718 718 checking to make sure that nothing is written into the database 719 719 file which has not first been written and synced to the rollback journal. 720 720 If any discrepancies are found, an assertion fault is raised.</p> 721 721 722 722 <p>The journal tests are an additional double-check over and above
Changes to pages/th3.in.
23 23 errors, and power loss during transaction commit. </p></li> 24 24 25 25 <li><p> TH3 exercises SQLite in a variety of run-time configurations 26 26 (UTF8 vs UTF16, different pages sizes, varying journal modes, etc.) 27 27 </p></li> 28 28 29 29 <li><p> TH3 achieves 100% branch test coverage over SQLite core. 30 - (Test coverage of the operating-system specific VFSes and extensions 30 + (Test coverage of the operating-system specific [VFSes] and extensions 31 31 such as FTS and RTREE is less than 100%). </p></li> 32 32 </ul> 33 33 34 34 <p>TH3 was originally written for validation testing only, but has 35 35 subsequently been used for development testing and debugging 36 36 as well, and has proven very helpful in those roles. A full-coverage 37 37 test run for TH3 takes less than 10 minutes on a workstation and hence
Changes to pages/vfs.in.
3 3 4 4 <h1 align="center"> 5 5 The SQLite OS Interface or "VFS" 6 6 </h1> 7 7 8 8 <p> 9 9 This article describes the SQLite OS portability layer or "VFS" - the 10 -thin module that is at the bottom of the SQLite implementation stack 11 -and which provides portability across operating systems. 10 +module at the bottom of the SQLite implementation stack 11 +that provides portability across operating systems. 12 12 </p> 13 13 14 14 <img src="images/vfs1.gif" align="right" hspace="10"> 15 15 <h2>1.0 The VFS In Relation To The Rest Of SQLite</h2> 16 16 17 17 <p> 18 -The internal organization of the SQLite library can be viewed as a 19 -stack of modules such as shown to the right. 18 +The internal organization of the SQLite library can be viewed as the 19 +stack of modules shown to the right. 20 20 The Tokenizer, Parser, and Code Generator components are used to 21 21 process SQL statements and convert them into executable programs 22 22 in a virtual machine language or byte code. 23 23 Roughly speaking, these top three layers implement 24 24 [sqlite3_prepare_v2()]. The byte code generated by the top three 25 25 layers is a [prepared statement]. 26 26 The Virtual Machine module is responsible for running the SQL statement ................................................................................ 38 38 39 39 <p> 40 40 This article is about the bottom layer. 41 41 </p> 42 42 43 43 <p>The OS Interface - also called the "VFS" - is what makes SQLite 44 44 portable across operating systems. Whenever any of the other modules 45 -in SQLite needs to communicate with the filesystem, or the operating 46 -system, they invoke methods in the VFS to communicate with the 47 -operating system on their behalf. Hence, porting SQLite to a new 45 +in SQLite needs to communicate with the operating 46 +system, they invoke methods in the VFS. The VFS then invokes the 47 +operating-specific code needed to satisfy the request. 48 +Hence, porting SQLite to a new 48 49 operating system is simply a matter of writing a new OS interface layer 49 50 or "VFS".</p> 50 51 51 52 <h2>2.0 Multiple VFSes</h2> 52 53 53 54 <p> 54 55 The standard SQLite source tree contains built-in VFSes for os/2, unix, ................................................................................ 59 60 60 61 <p> 61 62 Multiple VFSes can be registered at the same time. 62 63 Each VFS has a unique names. 63 64 Separate [database connections] within the same process can be using 64 65 different VFSes at the same time. For that matter, if a single 65 66 database connection has multiple database files open using 66 -the [ATTACH] command, then each database might be using a 67 +the [ATTACH] command, then each attached database might be using a 67 68 different VFS. 68 69 </p> 69 70 70 71 <p> 71 72 Unix builds come with multiple VFSes built-in. The default VFS 72 -for unix is called "unix" and is what is used in an overwhelming 73 +for unix is called "unix" and is the VFS used in an overwhelming 73 74 majority of applications. Other VFSes that can be found in unix 74 75 may include: 75 76 </p> 76 77 77 78 <ol> 78 79 <li><p><b>unix-dotfile</b> - uses dot-file locking rather than 79 80 POSIX advisory locks. ................................................................................ 84 85 <li><p><b>unix-none</b> - all file locking operations are no-ops. 85 86 <li><p><b>unix-namedsem</b> - uses named semaphores for file locking. 86 87 VXWorks only. 87 88 </ol> 88 89 89 90 <p> 90 91 The various unix VFSes differ only in the way they handle file locking - 91 -they share most of the implementation in common with one another and 92 -are all located in the same SQLite source file: "<tt>os_unix.c</tt>". 93 -Note that because they do use different locking strategies, they are 94 -subtle imcompatible with one another. If two processes are accessing 95 -the same SQLite database using different unix VFSes, they will probably 92 +they share most of their implementation in common with one another and 93 +are all located in the same SQLite source file: 94 +[http://www.sqlite.org/src/doc/trunk/src/os_unix.c | os_unix.c]. 95 +Note that except for "unix" and "unix-excl", the various unix VFSes all 96 +use incompatible locking implementations. If two processes are accessing 97 +the same SQLite database using different unix VFSes, they may 96 98 not see each others locks and may end up interfering with one another, 97 99 resulting in database corruption. The "unix-none" VFS in particular 98 100 does no locking at all and will easily result in database corruption if 99 -used to access a database by two or more database connections at a time. 101 +used by two or more database connections at the same time. 102 +Programmers are encouraged to use only "unix" or "unix-excl" unless 103 +there is a compelling reason to do otherwise. 100 104 </p> 101 105 102 106 <h2>2.1 Specifying Which VFS To Use</h2> 103 107 104 108 <p> 105 109 There is always one VFS which is the default VFS. On unix systems, 106 110 the "unix" VFS comes up as the default and on windows it is "win32". ................................................................................ 107 111 If no other actions are taken, new database connections will make use 108 112 of the default VFS. 109 113 </p> 110 114 111 115 <p> 112 116 The default VFS can be changed by registering or re-registering the 113 117 VFS using the [sqlite3_vfs_register()] interface with a second parameter 114 -of 1. Hence, of a (unix) process to make the "unix-nolock" VFS the 115 -default in place of "unix", the following code would work: 118 +of 1. Hence, if a (unix) process to always use the "unix-nolock" VFS 119 +in place of "unix", the following code would work: 116 120 </p> 117 121 118 122 <blockquote><pre> 119 123 sqlite3_vfs_register(sqlite3_vfs_find("unix-nolock"), 1); 120 124 </pre></blockquote> 121 125 122 126 <p> ................................................................................ 145 149 a VFS specified as the fourth argument to [sqlite3_open_v2()]. The 146 150 default VFS is used if no VFS is specified otherwise. 147 151 </p> 148 152 149 153 <h2>2.2 VFS Shims</h2> 150 154 151 155 <p> 152 -From the point of view of the uppers layers of the SQLite stack, there 153 -is only one VFS in use at a time. But in practice, a particular VFS might 154 -just be a thin wrapper around another VFS does the real work. 156 +From the point of view of the uppers layers of the SQLite stack, each 157 +open database file uses exactly one VFS. 158 +But in practice, a particular VFS might 159 +just be a thin wrapper around another VFS that does the real work. 155 160 We call a wrapper VFS a "shim". 156 161 </p> 157 162 158 163 <p> 159 164 A simple example of a shim is the "vfstrace" VFS. This is a VFS 160 -(imnplemented in the 161 -[http://www.sqlite.org/src/finfo?name=src/test_vfstrace.c | test_vfstrace.c] 165 +(implemented in the 166 +[http://www.sqlite.org/src/doc/trunk/src/test_vfstrace.c | test_vfstrace.c] 162 167 source file) that writes a message associated with each VFS method call 163 168 into a log file, then passes control off to another VFS to do the actual 164 169 work. 165 170 </p> 166 171 167 172 <h2>2.3 Other Example VFSes</h2> 168 173 ................................................................................ 169 174 <p> 170 175 The following are other VFS implementations available in the public 171 176 SQLite source tree: 172 177 </p> 173 178 174 179 <ul> 175 180 <li><p> 176 -<b>test_demovfs.c</b> - 181 +[http://www.sqlite.org/src/doc/trunk/src/test_demovfs.c | test_demovfs.c] - 177 182 This file implements a very simple VFS named "demo" that uss POSIX 178 183 functions such as 179 184 open(), read(), write(), fsync(), close(), fsync(), sleep(), time(), 180 185 and so forth. This VFS only works on unix systems. But it is not 181 186 intended as a replacement for the standard "unix" VFS used by default 182 187 on unix platforms. The "demo" VFS is deliberately kept very simple 183 188 so that it can be used as a learning aid or as template for building 184 189 other VFSes or for porting SQLite to new operating systems. 185 190 186 191 <li><p> 187 -<b>test_quota.c</b> - 192 +[http://www.sqlite.org/src/doc/trunk/src/test_quota.c | test_quota.c] - 188 193 This file implements a shim called "quota" that enforces cumulative 189 194 file size limits on a collection of database files. An auxiliary 190 195 interface is used to define "quote groups". A quota group is a 191 196 set of files (database files, journals, and temporary files) whose 192 197 names all match a [GLOB] pattern. The sum of the sizes of all files 193 198 in each quota group is tracked, and if that sum exceeds a threshold 194 199 defined for the quota group, a callback function is invoked. That 195 200 callback can either increase the threshold or cause the operation 196 201 that would have exceeded the quota to fail with an 197 202 [SQLITE_FULL] error. One of the uses of this shim is used to enforce 198 203 resource limits on application databases in Firefox. 199 204 200 205 <li><p> 201 -<b>test_multiplex.c</b> - 206 +[http://www.sqlite.org/src/doc/trunk/src/test_multiplex.c | test_multiplex.c] - 202 207 This file implements a shim that allows database files to exceed the 203 208 maximum file size of the underlying filesystem. This shim presents 204 209 an interface to the upper six layers of SQLite that makes it look like 205 210 very large files are being used, when in reality each such large file 206 211 is split up into many smaller files on the underlying system. 207 212 This shim has been used, for example, to allow databases to grow 208 213 larger than 2 gibibytes on FAT16 filesystems. 209 214 210 215 <li><p> 211 -<b>test_onefile.c</b> - 216 +[http://www.sqlite.org/src/doc/trunk/src/test_onefile.c | test_onefile.c] - 212 217 This file implements a demonstration VFS named "fs" that shows how SQLite 213 218 can be used on an embedded device that lacks a filesystem. Content is 214 219 written directly to the underlying media. A VFS derived from this 215 220 demonstration code could be used by a gadget with a limited amount of 216 221 flash memory to make SQLite behave as the filesystem for the flash memory 217 222 on the device. 218 223 219 224 <li><p> 220 -<b>test_journal.c</b> - 225 +[http://www.sqlite.org/src/doc/trunk/src/test_journal.c | test_journal.c] - 221 226 This file implements a shim used during SQLite testing that verifies that 222 227 the database and rollback journal are written in the correct order and 223 228 are "synced" at appropriate times in order to guarantee that the database 224 229 can recover from a power lose are hard reset at any time. The shim 225 230 checks several invariants on the operation of databases and rollback 226 231 journals and raises exceptions if any of those invariants are violated. 227 232 These invariants, in turn, assure that the database is always recoverable. 228 233 Running a large suite of test cases using this shim provides added 229 234 assurance that SQLite databases will not be damaged by unexpected 230 235 power failures or device resets. 231 236 232 237 <li><p> 233 -<b>test_vfs.c</b> - 238 +[http://www.sqlite.org/src/doc/trunk/src/test_vfs.c | test_vfs.c] - 234 239 This file implements a shim that can be used to simulate filesystem faults. 235 240 This shim is used during testing to verify that SQLite responses sanely 236 241 to hardware malfunctions or to other error conditions such as running out 237 242 of filesystem space that are difficult to test on a real system. 238 243 </ul> 239 244 240 245 <p>
Changes to pages/wal.in.
20 20 <li>WAL uses many fewer fsync() operations and is thus less vulnerable to 21 21 problems on systems where the fsync() system call is broken. 22 22 </ol> 23 23 24 24 <p>But there are also disadvantages:</p> 25 25 26 26 <ol> 27 -<li>WAL normally requires that the [sqlite3_vfs | VFS] 28 - supports shared-memory primitives. 27 +<li>WAL normally requires that the [VFS] 28 + support shared-memory primitives. 29 29 (Exception: [WAL without shared memory]) 30 30 The built-in unix and windows VFSes 31 31 support this but third-party extension VFSes for custom operating 32 32 systems might not. 33 33 <li>All processes using a database must be on the same host computer; 34 34 WAL does not work over a network filesystem. 35 35 <li>Transactions that involve changes against multiple [ATTACH | ATTACHed] ................................................................................ 243 243 244 244 <blockquote><pre> 245 245 PRAGMA journal_mode=WAL; 246 246 </pre></blockquote> 247 247 248 248 <p>The journal_mode pragma returns a string which is the new journal mode. 249 249 On success, the pragma will return the string "<tt>wal</tt>". If 250 -the conversion to WAL could not be completed (for example, if the VFS 250 +the conversion to WAL could not be completed (for example, if the [VFS] 251 251 does not support the necessary shared-memory primitives) then the 252 252 journaling mode will be unchanged and the string returned from the 253 253 primitive will be the prior journaling mode (for example "<tt>delete</tt>"). 254 254 255 255 <a name="how_to_checkpoint"></a> 256 256 <h3>Automatic Checkpoint</h3> 257 257 ................................................................................ 309 309 operations.</p> 310 310 311 311 <p>An efficient implementation of the [WAL read algorithm] requires that 312 312 there exist a hash table in shared memory over the content of the WAL file. 313 313 This hash table is called the [wal-index]. 314 314 The wal-index is in shared memory, and so technically it does not have 315 315 to have a name in the host computer filesystem. Custom 316 -[sqlite3_vfs | VFS] implementations are free to implement shared 316 +[VFS] implementations are free to implement shared 317 317 memory in any way they see fit, but the default unix and windows 318 318 drivers that come built-in with SQLite implement shared memory 319 319 using <a href="http://en.wikipedia.org/wiki/Mmap">mmapped files</a> 320 320 named using the suffix "<tt>-shm</tt>" and 321 321 located in the same directory as the database file. The wal-index must 322 322 be rebuilt upon first access, even by readers, and so in order to open 323 323 the WAL database, write access is required on the "<tt>-shm</tt>" shared ................................................................................ 361 361 think this is a major concern since the wal-index rarely exceeds 362 362 32 KiB in size and is never synced. Furthermore, the wal-index 363 363 backing file is deleted when the last database connection disconnects, 364 364 which often prevents any real disk I/O from ever happening.</p> 365 365 366 366 <p>Specialized applications for which the default implementation of 367 367 shared memory is unacceptable can devise alternative methods via a 368 -custom [sqlite3_vfs | VFS]. 368 +custom [VFS]. 369 369 For example, if it is known that a particular database 370 370 will only be accessed by threads within a single process, the wal-index 371 371 can be implemented using heap memory instead of true shared memory.</p> 372 372 373 373 <tcl>hd_fragment noshm {WAL without shared memory}</tcl> 374 374 <h2>Use of WAL Without Shared-Memory</h2> 375 375 ................................................................................ 376 376 <p>Beginning in SQLite version 3.7.4, ^WAL databases can be created, read, and 377 377 written even if shared memory is unavailable as long as the 378 378 [locking_mode] is set to EXCLUSIVE before the first attempted access. 379 379 In other words, a process can interact with 380 380 a WAL database without using shared memory if that 381 381 process is guaranteed to be the only process accessing the database. 382 382 ^This feature allows WAL databases to be created, read, and written 383 -by legacy [sqlite3_vfs | VFSes] that lack the "version 2" shared-memory 383 +by legacy [VFSes] that lack the "version 2" shared-memory 384 384 methods xShmMap, xShmLock, xShmBarrier, and xShmUnmap on the 385 385 [sqlite3_io_methods] object.</p> 386 386 387 387 <p>^(If EXCLUSIVE locking mode is set prior to the first WAL-mode 388 388 database access, then SQLite never attempts to call any of the 389 389 shared-memory methods and hence no shared-memory 390 390 wal-index is ever created.)^