Database Name Is A File Name¶

SQLite creates a file per a database. You should pass the "path" of the database file (with or without a parent directory) in the DBI connection string (as a database "name"):

  my $dbh = DBI->connect("dbi:SQLite:dbname=$dbfile","","");

The file is opened in read/write mode, and will be created if it does not exist yet.

Although the database is stored in a single file, the directory containing the database file must be writable by SQLite because the library will create several temporary files there.

If the filename $dbfile is ":memory:", then a private, temporary in-memory database is created for the connection. This in-memory database will vanish when the database connection is closed. It is handy for your library tests.

Note that future versions of SQLite might make use of additional special filenames that begin with the ":" character. It is recommended that when a database filename actually does begin with a ":" character you should prefix the filename with a pathname such as "./" to avoid ambiguity.

If the filename $dbfile is an empty string, then a private, temporary on-disk database will be created. This private database will be automatically deleted as soon as the database connection is closed.

Accessing A Database With Other Tools¶

To access the database from the command line, try using "dbish" which comes with the DBI::Shell module. Just type:

  dbish dbi:SQLite:foo.db

On the command line to access the file foo.db.

Alternatively you can install SQLite from the link above without conflicting with DBD::SQLite and use the supplied "sqlite3" command line tool.

Blobs¶

As of version 1.11, blobs should "just work" in SQLite as text columns. However this will cause the data to be treated as a string, so SQL statements such as length(x) will return the length of the column as a NUL terminated string, rather than the size of the blob in bytes. In order to store natively as a BLOB use the following code:

  use DBI qw(:sql_types);
  my $dbh = DBI->connect("dbi:SQLite:dbfile","","");
  
  my $blob = `cat foo.jpg`;
  my $sth = $dbh->prepare("INSERT INTO mytable VALUES (1, ?)");
  $sth->bind_param(1, $blob, SQL_BLOB);
  $sth->execute();

And then retrieval just works:

  $sth = $dbh->prepare("SELECT * FROM mytable WHERE id = 1");
  $sth->execute();
  my $row = $sth->fetch;
  my $blobo = $row->[1];
  
  # now $blobo == $blob

Functions And Bind Parameters¶

As of this writing, a SQL that compares a return value of a function with a numeric bind value like this doesn't work as you might expect.

  my $sth = $dbh->prepare(q{
    SELECT bar FROM foo GROUP BY bar HAVING count(*) > ?;
  });
  $sth->execute(5);

This is because DBD::SQLite assumes that all the bind values are text (and should be quoted) by default. Thus the above statement becomes like this while executing:

  SELECT bar FROM foo GROUP BY bar HAVING count(*) > "5";

There are two workarounds for this.

Use bind_param() explicitly

As shown above in the "BLOB" section, you can always use "bind_param()" to tell the type of a bind value.

  use DBI qw(:sql_types);  # Don't forget this
  
  my $sth = $dbh->prepare(q{
    SELECT bar FROM foo GROUP BY bar HAVING count(*) > ?;
  });
  $sth->bind_param(1, 5, SQL_INTEGER);
  $sth->execute();
    

Add zero to make it a number

This is somewhat weird, but works anyway.

  my $sth = $dbh->prepare(q{
    SELECT bar FROM foo GROUP BY bar HAVING count(*) > (? + 0);
  });
  $sth->execute(5);
    

Foreign Keys¶

BE PREPARED! WOLVES APPROACH!!

SQLite has started supporting foreign key constraints since 3.6.19 (released on Oct 14, 2009; bundled with DBD::SQLite 1.26_05). To be exact, SQLite has long been able to parse a schema with foreign keys, but the constraints has not been enforced. Now you can issue a pragma actually to enable this feature and enforce the constraints.

To do this, issue the following pragma (see below), preferably as soon as you connect to a database and you're not in a transaction:

  $dbh->do("PRAGMA foreign_keys = ON");

And you can explicitly disable the feature whenever you like by turning the pragma off:

  $dbh->do("PRAGMA foreign_keys = OFF");

As of this writing, this feature is disabled by default by the sqlite team, and by us, to secure backward compatibility, as this feature may break your applications, and actually broke some for us. If you have used a schema with foreign key constraints but haven't cared them much and supposed they're always ignored for SQLite, be prepared, and please do extensive testing to ensure that your applications will continue to work when the foreign keys support is enabled by default. It is very likely that the sqlite team will turn it default-on in the future, and we plan to do it NO LATER THAN they do so.

See <http://www.sqlite.org/foreignkeys.html> for details.

Pragma¶

SQLite has a set of "Pragma"s to modifiy its operation or to query for its internal data. These are specific to SQLite and are not likely to work with other DBD libraries, but you may find some of these are quite useful. DBD::SQLite actually sets some (like "foreign_keys" above) for you when you connect to a database. See <http://www.sqlite.org/pragma.html> for details.

Performance¶

SQLite is fast, very fast. Matt processed my 72MB log file with it, inserting the data (400,000+ rows) by using transactions and only committing every 1000 rows (otherwise the insertion is quite slow), and then performing queries on the data.

Queries like count(*) and avg(bytes) took fractions of a second to return, but what surprised him most of all was:

  SELECT url, count(*) as count
  FROM access_log
  GROUP BY url
  ORDER BY count desc
  LIMIT 20

To discover the top 20 hit URLs on the site (<http://axkit.org>), and it returned within 2 seconds. He was seriously considering switching his log analysis code to use this little speed demon!

Oh yeah, and that was with no indexes on the table, on a 400MHz PIII.

For best performance be sure to tune your hdparm settings if you are using linux. Also you might want to set:

  PRAGMA default_synchronous = OFF

Which will prevent sqlite from doing fsync's when writing (which slows down non-transactional writes significantly) at the expense of some peace of mind. Also try playing with the cache_size pragma.

The memory usage of SQLite can also be tuned using the cache_size pragma.

  $dbh->do("PRAGMA cache_size = 800000");

The above will allocate 800M for DB cache; the default is 2M. Your sweet spot probably lies somewhere in between.

Database Handle Attributes¶

sqlite_version

Returns the version of the SQLite library which DBD::SQLite is using, e.g., "2.8.0". Can only be read.

sqlite_unicode

If set to a true value, DBD::SQLite will turn the UTF-8 flag on for all text strings coming out of the database (this feature is currently disabled for perl < 5.8.5). For more details on the UTF-8 flag see perlunicode. The default is for the UTF-8 flag to be turned off.

Also note that due to some bizarreness in SQLite's type system (see <http://www.sqlite.org/datatype3.html>), if you want to retain blob-style behavior for some columns under "$dbh->{sqlite_unicode} = 1" (say, to store images in the database), you have to state so explicitly using the 3-argument form of "bind_param" in DBI when doing updates:

  use DBI qw(:sql_types);
  $dbh->{sqlite_unicode} = 1;
  my $sth = $dbh->prepare("INSERT INTO mytable (blobcolumn) VALUES (?)");
  
  # Binary_data will be stored as is.
  $sth->bind_param(1, $binary_data, SQL_BLOB);
    

Defining the column type as "BLOB" in the DDL is not sufficient.

This attribute was originally named as "unicode", and renamed to "sqlite_unicode" for integrity since version 1.26_06. Old "unicode" attribute is still accessible but will be deprecated in the near future.

table_info¶

  $sth = $dbh->table_info(undef, $schema, $table, $type, \%attr);

Returns all tables and schemas (databases) as specified in "table_info" in DBI. The schema and table arguments will do a "LIKE" search. You can specify an ESCAPE character by including an 'Escape' attribute in \%attr. The $type argument accepts a comma seperated list of the following types 'TABLE', 'VIEW', 'LOCAL TEMPORARY' and 'SYSTEM TABLE' (by default all are returned). Note that a statement handle is returned, and not a direct list of tables.

The following fields are returned:

TABLE_CAT: Always NULL, as SQLite does not have the concept of catalogs.

TABLE_SCHEM: The name of the schema (database) that the table or view is in. The default schema is 'main', temporary tables are in 'temp' and other databases will be in the name given when the database was attached.

TABLE_NAME: The name of the table or view.

TABLE_TYPE: The type of object returned. Will be one of 'TABLE', 'VIEW', 'LOCAL TEMPORARY' or 'SYSTEM TABLE'.

$dbh->sqlite_last_insert_rowid()¶

This method returns the last inserted rowid. If you specify an INTEGER PRIMARY KEY as the first column in your table, that is the column that is returned. Otherwise, it is the hidden ROWID column. See the sqlite docs for details.

Generally you should not be using this method. Use the DBI last_insert_id method instead. The usage of this is:

  $h->last_insert_id($catalog, $schema, $table_name, $field_name [, \%attr ])

Running "$h->last_insert_id("","","","")" is the equivalent of running "$dbh->sqlite_last_insert_rowid()" directly.

$dbh->sqlite_busy_timeout()¶

Retrieve the current busy timeout.

$dbh->sqlite_busy_timeout( $ms )¶

Set the current busy timeout. The timeout is in milliseconds.

$dbh->sqlite_create_function( $name, $argc, $code_ref )¶

This method will register a new function which will be useable in an SQL query. The method's parameters are:

$name

The name of the function. This is the name of the function as it will be used from SQL.

$argc

The number of arguments taken by the function. If this number is -1, the function can take any number of arguments.

$code_ref

This should be a reference to the function's implementation.

For example, here is how to define a now() function which returns the current number of seconds since the epoch:

  $dbh->sqlite_create_function( 'now', 0, sub { return time } );

After this, it could be use from SQL as:

  INSERT INTO mytable ( now() );

REGEXP function

SQLite includes syntactic support for an infix operator 'REGEXP', but without any implementation. The "DBD::SQLite" driver automatically registers an implementation that performs standard perl regular expression matching, using current locale. So for example you can search for words starting with an 'A' with a query like

  SELECT * from table WHERE column REGEXP '\bA\w+'

If you want case-insensitive searching, use perl regex flags, like this :

  SELECT * from table WHERE column REGEXP '(?i:\bA\w+)'

The default REGEXP implementation can be overriden through the "create_function" API described above.

Note that regexp matching will not use SQLite indices, but will iterate over all rows, so it could be quite costly in terms of performance.

$dbh->sqlite_create_collation( $name, $code_ref )¶

This method manually registers a new function which will be useable in an SQL query as a COLLATE option for sorting. Such functions can also be registered automatically on demand: see section "COLLATION FUNCTIONS" below.

The method's parameters are:

$name

The name of the function exposed to SQL.

$code_ref

Reference to the function's implementation. The driver will check that this is a proper sorting function.

$dbh->sqlite_collation_needed( $code_ref )¶

This method manually registers a callback function that will be invoked whenever an undefined collation sequence is required from an SQL statement. The callback is invoked as

  $code_ref->($dbh, $collation_name)

and should register the desired collation using "sqlite_create_collation".

An initial callback is already registered by "DBD::SQLite", so for most common cases it will be simpler to just add your collation sequences in the %DBD::SQLite::COLLATION hash (see section "COLLATION FUNCTIONS" below).

$dbh->sqlite_create_aggregate( $name, $argc, $pkg )¶

This method will register a new aggregate function which can then be used from SQL. The method's parameters are:

$name

The name of the aggregate function, this is the name under which the function will be available from SQL.

$argc

This is an integer which tells the SQL parser how many arguments the function takes. If that number is -1, the function can take any number of arguments.

$pkg

This is the package which implements the aggregator interface.

The aggregator interface consists of defining three methods:

new()

This method will be called once to create an object which should be used to aggregate the rows in a particular group. The step() and finalize() methods will be called upon the reference return by the method.

step(@_)

This method will be called once for each row in the aggregate.

finalize()

This method will be called once all rows in the aggregate were processed and it should return the aggregate function's result. When there is no rows in the aggregate, finalize() will be called right after new().

Here is a simple aggregate function which returns the variance (example adapted from pysqlite):

  package variance;
  
  sub new { bless [], shift; }
  
  sub step {
      my ( $self, $value ) = @_;
  
      push @$self, $value;
  }
  
  sub finalize {
      my $self = $_[0];
  
      my $n = @$self;
  
      # Variance is NULL unless there is more than one row
      return undef unless $n || $n == 1;
  
      my $mu = 0;
      foreach my $v ( @$self ) {
          $mu += $v;
      }
      $mu /= $n;
  
      my $sigma = 0;
      foreach my $v ( @$self ) {
          $sigma += ($x - $mu)**2;
      }
      $sigma = $sigma / ($n - 1);
  
      return $sigma;
  }
  
  $dbh->sqlite_create_aggregate( "variance", 1, 'variance' );

The aggregate function can then be used as:

  SELECT group_name, variance(score)
  FROM results
  GROUP BY group_name;

For more examples, see the DBD::SQLite::Cookbook.

$dbh->sqlite_progress_handler( $n_opcodes, $code_ref )¶

This method registers a handler to be invoked periodically during long running calls to SQLite.

An example use for this interface is to keep a GUI updated during a large query. The parameters are:

$n_opcodes

The progress handler is invoked once for every $n_opcodes virtual machine opcodes in SQLite.

$code_ref

Reference to the handler subroutine. If the progress handler returns non-zero, the SQLite operation is interrupted. This feature can be used to implement a "Cancel" button on a GUI dialog box.

Set this argument to "undef" if you want to unregister a previous progress handler.

$dbh->sqlite_commit_hook( $code_ref )¶

This method registers a callback function to be invoked whenever a transaction is committed. Any callback set by a previous call to "sqlite_commit_hook" is overridden. A reference to the previous callback (if any) is returned. Registering an "undef" disables the callback.

When the commit hook callback returns zero, the commit operation is allowed to continue normally. If the callback returns non-zero, then the commit is converted into a rollback (in that case, any attempt to explicitly call "$dbh->rollback()" afterwards would yield an error).

$dbh->sqlite_rollback_hook( $code_ref )¶

This method registers a callback function to be invoked whenever a transaction is rolled back. Any callback set by a previous call to "sqlite_rollback_hook" is overridden. A reference to the previous callback (if any) is returned. Registering an "undef" disables the callback.

$dbh->sqlite_update_hook( $code_ref )¶

This method registers a callback function to be invoked whenever a row is updated, inserted or deleted. Any callback set by a previous call to "sqlite_update_hook" is overridden. A reference to the previous callback (if any) is returned. Registering an "undef" disables the callback.

The callback will be called as

  $code_ref->($action_code, $database, $table, $rowid)

where

$action_code

is an integer equal to either "DBD::SQLite::INSERT", "DBD::SQLite::DELETE" or "DBD::SQLite::UPDATE" (see "Action Codes");

$database

is the name of the database containing the affected row;

$table

is the name of the table containing the affected row;

$rowid

is the unique 64-bit signed integer key of the affected row within that table.

$dbh->sqlite_set_authorizer( $code_ref )¶

This method registers an authorizer callback to be invoked whenever SQL statements are being compiled by the "prepare" in DBI method. The authorizer callback should return "DBD::SQLite::OK" to allow the action, "DBD::SQLite::IGNORE" to disallow the specific action but allow the SQL statement to continue to be compiled, or "DBD::SQLite::DENY" to cause the entire SQL statement to be rejected with an error. If the authorizer callback returns any other value, then then "prepare" call that triggered the authorizer will fail with an error message.

An authorizer is used when preparing SQL statements from an untrusted source, to ensure that the SQL statements do not try to access data they are not allowed to see, or that they do not try to execute malicious statements that damage the database. For example, an application may allow a user to enter arbitrary SQL queries for evaluation by a database. But the application does not want the user to be able to make arbitrary changes to the database. An authorizer could then be put in place while the user-entered SQL is being prepared that disallows everything except SELECT statements.

The callback will be called as

  $code_ref->($action_code, $string1, $string2, $database, $trigger_or_view)

where

$action_code

is an integer that specifies what action is being authorized (see "Action Codes").

$string1, $string2

are strings that depend on the action code (see "Action Codes").

$database

is the name of the database ("main", "temp", etc.) if applicable.

$trigger_or_view

is the name of the inner-most trigger or view that is responsible for the access attempt, or "undef" if this access attempt is directly from top-level SQL code.

$dbh->sqlite_backup_from_file( $filename )¶

This method accesses the SQLite Online Backup API, and will take a backup of the named database file, copying it to, and overwriting, your current database connection. This can be particularly handy if your current connection is to the special :memory: database, and you wish to populate it from an existing DB.

$dbh->sqlite_backup_to_file( $filename )¶

This method accesses the SQLite Online Backup API, and will take a backup of the currently connected database, and write it out to the named file.

$dbh->sqlite_enable_load_extension( $bool )¶

Calling this method with a true value enables loading (external) sqlite3 extensions. After the call, you can load extensions like this:

  $dbh->sqlite_enable_load_extension(1);
  $sth = $dbh->prepare("select load_extension('libsqlitefunctions.so')")
  or die "Cannot prepare: " . $dbh->errstr();

Authorizer Return Codes¶

  OK
  DENY
  IGNORE

Action Codes¶

The "set_authorizer" method registers a callback function that is invoked to authorize certain SQL statement actions. The first parameter to the callback is an integer code that specifies what action is being authorized. The second and third parameters to the callback are strings, the meaning of which varies according to the action code. Below is the list of action codes, together with their associated strings.

  # constant              string1         string2
  # ========              =======         =======
  CREATE_INDEX            Index Name      Table Name
  CREATE_TABLE            Table Name      undef
  CREATE_TEMP_INDEX       Index Name      Table Name
  CREATE_TEMP_TABLE       Table Name      undef
  CREATE_TEMP_TRIGGER     Trigger Name    Table Name
  CREATE_TEMP_VIEW        View Name       undef
  CREATE_TRIGGER          Trigger Name    Table Name
  CREATE_VIEW             View Name       undef
  DELETE                  Table Name      undef
  DROP_INDEX              Index Name      Table Name
  DROP_TABLE              Table Name      undef
  DROP_TEMP_INDEX         Index Name      Table Name
  DROP_TEMP_TABLE         Table Name      undef
  DROP_TEMP_TRIGGER       Trigger Name    Table Name
  DROP_TEMP_VIEW          View Name       undef
  DROP_TRIGGER            Trigger Name    Table Name
  DROP_VIEW               View Name       undef
  INSERT                  Table Name      undef
  PRAGMA                  Pragma Name     1st arg or undef
  READ                    Table Name      Column Name
  SELECT                  undef           undef
  TRANSACTION             Operation       undef
  UPDATE                  Table Name      Column Name
  ATTACH                  Filename        undef
  DETACH                  Database Name   undef
  ALTER_TABLE             Database Name   Table Name
  REINDEX                 Index Name      undef
  ANALYZE                 Table Name      undef
  CREATE_VTABLE           Table Name      Module Name
  DROP_VTABLE             Table Name      Module Name
  FUNCTION                undef           Function Name
  SAVEPOINT               Operation       Savepoint Name

Definition¶

SQLite v3 provides the ability for users to supply arbitrary comparison functions, known as user-defined "collation sequences" or "collating functions", to be used for comparing two text values. <http://www.sqlite.org/datatype3.html#collation> explains how collations are used in various SQL expressions.

Builtin collation sequences¶

The following collation sequences are builtin within SQLite :

BINARY

Compares string data using memcmp(), regardless of text encoding.

NOCASE

The same as binary, except the 26 upper case characters of ASCII are folded to their lower case equivalents before the comparison is performed. Note that only ASCII characters are case folded. SQLite does not attempt to do full UTF case folding due to the size of the tables required.

RTRIM

The same as binary, except that trailing space characters are ignored.

In addition, "DBD::SQLite" automatically installs the following collation sequences :

perl

corresponds to the Perl "cmp" operator

perllocale

Perl "cmp" operator, in a context where "use locale" is activated.

Usage¶

You can write for example

  CREATE TABLE foo(
      txt1 COLLATE perl,
      txt2 COLLATE perllocale,
      txt3 COLLATE nocase
  )

or

  SELECT * FROM foo ORDER BY name COLLATE perllocale

Unicode handling¶

If the attribute "$dbh->{sqlite_unicode}" is set, strings coming from the database and passed to the collation function will be properly tagged with the utf8 flag; but this only works if the "sqlite_unicode" attribute is set before the first call to a perl collation sequence . The recommended way to activate unicode is to set the parameter at connection time :

  my $dbh = DBI->connect(
      "dbi:SQLite:dbname=foo", "", "",
      {
          RaiseError     => 1,
          sqlite_unicode => 1,
      }
  );

Adding user-defined collations¶

The native SQLite API for adding user-defined collations is exposed through methods "sqlite_create_collation" and "sqlite_collation_needed".

To avoid calling these functions every time a $dbh handle is created, "DBD::SQLite" offers a simpler interface through the %DBD::SQLite::COLLATION hash : just insert your own collation functions in that hash, and whenever an unknown collation name is encountered in SQL, the appropriate collation function will be loaded on demand from the hash. For example, here is a way to sort text values regardless of their accented characters :

  use DBD::SQLite;
  $DBD::SQLite::COLLATION{no_accents} = sub {
    my ( $a, $b ) = map lc, @_;
    tr[aaaaeaaa~cd`eeeeeiiiien~ooooeo~ouuuuey]
      [aaaaaacdeeeeiiiinoooooouuuuy] for $a, $b;
    $a cmp $b;
  };
  my $dbh  = DBI->connect("dbi:SQLite:dbname=dbfile");
  my $sql  = "SELECT ... FROM ... ORDER BY ... COLLATE no_accents");
  my $rows = $dbh->selectall_arrayref($sql);

The builtin "perl" or "perllocale" collations are predefined in that same hash.

The COLLATION hash is a global registry within the current process; hence there is a risk of undesired side-effects. Therefore, to prevent action at distance, the hash is implemented as a "write-only" hash, that will happily accept new entries, but will raise an exception if any attempt is made to override or delete a existing entry (including the builtin "perl" and "perllocale").

If you really, really need to change or delete an entry, you can always grab the tied object underneath %DBD::SQLite::COLLATION --- but don't do that unless you really know what you are doing. Also observe that changes in the global hash will not modify existing collations in existing database handles: it will only affect new requests for collations. In other words, if you want to change the behaviour of a collation within an existing $dbh, you need to call the "create_collation" method directly.

Warnings Upgrade¶

We currently use a horridly hacky method to issue and suppress warnings. It suffices for now, but just barely.

Migrate all of the warning code to use the recommended DBI warnings.

Leak Detection¶

Implement one or more leak detection tests that only run during AUTOMATED_TESTING and RELEASE_TESTING and validate that none of the C code we work with leaks.

Stream API for Blobs¶

Reading/writing into blobs using "sqlite2_blob_open" / "sqlite2_blob_close".

Flags for sqlite3_open_v2¶

Support the full API of sqlite3_open_v2 (flags for opening the file).