DBD::File - Base class for writing file based DBI drivers


This module is a base class for writing other DBDs. It is not intended to function as a DBD itself (though it is possible). If you want to access flat files, use DBD::AnyData, or DBD::CSV (both of which are subclasses of DBD::File).


The DBD::File module is not a true DBI driver, but an abstract base class for deriving concrete DBI drivers from it. The implication is, that these drivers work with plain files, for example CSV files or INI files. The module is based on the SQL::Statement module, a simple SQL engine.

See DBI for details on DBI, SQL::Statement for details on SQL::Statement and DBD::CSV, DBD::DBM or DBD::AnyData for example drivers.


The following attributes are handled by DBI itself and not by DBD::File, thus they all work as expected:

    CompatMode             (Not used)
    Warn                   (Not used)

The following DBI attributes are handled by DBD::File:


Always on.




Valid after $sth->execute.


Valid after $sth->prepare.


Valid after $sth->execute; undef for Non-Select statements.


Not really working, always returns an array ref of ones, except the affected table has been created in this session. Valid after $sth->execute; undef for non-select statements.

Unsupported DBI attributes and methods





DBD::File specific attributes

In addition to the DBI attributes, you can use the following dbh attributes:


This attribute is used for setting the directory where the files are opened and it defaults to the current directory (.). Usually you set it on the dbh but it may be overridden per table (see f_meta).

When the value for f_dir is a relative path, it is converted into the appropriate absolute path name (based on the current working directory) when the dbh attribute is set.

  f_dir => "/data/foo/csv",


This optional attribute can be set to pass a list of folders to also find existing tables. It will not be used to create new files.

  f_dir_search => [ "/data/bar/csv", "/dump/blargh/data" ],


This attribute is used for setting the file extension. The format is:


where the /flag is optional and the extension is case-insensitive. f_ext allows you to specify an extension which:

  f_ext => ".csv/r",
  • makes DBD::File prefer table.extension over table.

  • makes the table name the filename minus the extension.


In the above example and when f_dir contains both table.csv and table, DBD::File will open table.csv and the table will be named "table". If table.csv does not exist but table does that file is opened and the table is also called "table".

If f_ext is not specified and table.csv exists it will be opened and the table will be called "table.csv" which is probably not what you want.

NOTE: even though extensions are case-insensitive, table names are not.


The r flag means the file extension is required and any filename that does not match the extension is ignored.

Usually you set it on the dbh but it may be overridden per table (see f_meta).


This will set the schema name and defaults to the owner of the directory in which the table file resides. You can set f_schema to undef.

    my $dbh = DBI->connect ("dbi:CSV:", "", "", {
        f_schema => undef,
        f_dir    => "data",
        f_ext    => ".csv/r",
        }) or die $DBI::errstr;

By setting the schema you affect the results from the tables call:

    my @tables = $dbh->tables ();

    # no f_schema

    # f_schema => "dbi"

    # f_schema => undef

Defining f_schema to the empty string is equal to setting it to undef so the DSN can be "dbi:CSV:f_schema=;f_dir=.".


The f_lock attribute is used to set the locking mode on the opened table files. Note that not all platforms support locking. By default, tables are opened with a shared lock for reading, and with an exclusive lock for writing. The supported modes are:

  0: No locking at all.

  1: Shared locks will be used.

  2: Exclusive locks will be used.

But see KNOWN BUGS below.


If you wish to use a lockfile extension other than .lck, simply specify the f_lockfile attribute:

  $dbh = DBI->connect ("");
  $dbh->{f_lockfile} = ".foo";
  $dbh->{dbm_tables}{qux}{f_lockfile} = ".foo";

If you wish to disable locking, set the f_lockfile to 0.

  $dbh = DBI->connect ("dbi:DBM:f_lockfile=0");
  $dbh->{f_lockfile} = 0;
  $dbh->{dbm_tables}{qux}{f_lockfile} = 0;


With this attribute, you can set the encoding in which the file is opened. This is implemented using binmode $fh, ":encoding(<f_encoding>)".


Private data area aliasing "sql_meta" in DBI::DBD::SqlEngine which contains information about the tables this module handles. Table meta data might not be available until the table has been accessed for the first time e.g., by issuing a select on it however it is possible to pre-initialize attributes for each table you use.

DBD::File recognizes the (public) attributes f_ext, f_dir, f_file, f_encoding, f_lock, f_lockfile, f_schema, in addition to the attributes "sql_meta" in DBI::DBD::SqlEngine already supports. Be very careful when modifying attributes you do not know, the consequence might be a destroyed or corrupted table.

f_file is an attribute applicable to table meta data only and you will not find a corresponding attribute in the dbh. Whilst it may be reasonable to have several tables with the same column names, it is not for the same file name. If you need access to the same file using different table names, use SQL::Statement as the SQL engine and the AS keyword:

    SELECT * FROM tbl AS t1, tbl AS t2 WHERE =

f_file can be an absolute path name or a relative path name but if it is relative, it is interpreted as being relative to the f_dir attribute of the table meta data. When f_file is set DBD::File will use f_file as specified and will not attempt to work out an alternative for f_file using the table name and f_ext attribute.

While f_meta is a private and readonly attribute (which means, you cannot modify it's values), derived drivers might provide restricted write access through another attribute. Well known accessors are csv_tables for DBD::CSV, ad_tables for DBD::AnyData and dbm_tables for DBD::DBM.

New opportunities for attributes from DBI::DBD::SqlEngine


$dbh->{sql_table_source} can be set to DBD::File::TableSource::FileSystem (and is the default setting of DBD::File). This provides usual behaviour of previous DBD::File releases on

  @ary = DBI->data_sources ($driver);
  @ary = DBI->data_sources ($driver, \%attr);
  @ary = $dbh->data_sources ();
  @ary = $dbh->data_sources (\%attr);

  @names = $dbh->tables ($catalog, $schema, $table, $type);
  $sth = $dbh->table_info ($catalog, $schema, $table, $type);
  $sth = $dbh->table_info ($catalog, $schema, $table, $type, \%attr);

  $dbh->func ("list_tables");


$dbh->{sql_data_source} can be set to either DBD::File::DataSource::File, which is default and provides the well known behavior of DBD::File releases prior to 0.41, or DBD::File::DataSource::Stream, which reuses already opened file-handle for operations.

Internally private attributes to deal with SQL backends

Do not modify any of these private attributes unless you understand the implications of doing so. The behavior of DBD::File and derived DBDs might be unpredictable when one or more of those attributes are modified.


Contains the version of loaded DBI::SQL::Nano.


Contains the version of loaded SQL::Statement.


Contains either the text 'SQL::Statement' or 'DBI::SQL::Nano'.


Contains optionally temporary tables.


Contains optional flags to instantiate the SQL::Parser parsing engine when SQL::Statement is used as SQL engine. See SQL::Parser for valid flags.

Driver private methods

Default DBI methods


The data_sources method returns a list of subdirectories of the current directory in the form "dbi:CSV:f_dir=$dirname".

If you want to read the subdirectories of another directory, use

    my ($drh)  = DBI->install_driver ("CSV");
    my (@list) = $drh->data_sources (f_dir => "/usr/local/csv_data");

Additional methods

The following methods are only available via their documented name when DBD::File is used directly. Because this is only reasonable for testing purposes, the real names must be used instead. Those names can be computed by replacing the f_ in the method name with the driver prefix.



  sub f_versions (;$)
    my ($table_name) = @_;
    $table_name ||= ".";

Returns the versions of the driver, including the DBI version, the Perl version, DBI::PurePerl version (if DBI::PurePerl is active) and the version of the SQL engine in use.

    my $dbh = DBI->connect ("dbi:File:");
    my $f_versions = $dbh->func ("f_versions");
    print "$f_versions\n";
    # DBD::File              0.41 using IO::File (1.16)
    #   DBI::DBD::SqlEngine  0.05 using SQL::Statement 1.406
    # DBI                    1.623
    # OS                     darwin (12.2.1)
    # Perl                   5.017006 (darwin-thread-multi-ld-2level)

Called in list context, f_versions will return an array containing each line as single entry.

Some drivers might use the optional (table name) argument and modify version information related to the table (e.g. DBD::DBM provides storage backend information for the requested table, when it has a table name).


  • This module uses flock () internally but flock is not available on all platforms. On MacOS and Windows 95 there is no locking at all (perhaps not so important on MacOS and Windows 95, as there is only a single user).

  • The module stores details about the handled tables in a private area of the driver handle ($drh). This data area is not shared between different driver instances, so several DBI->connect () calls will cause different table instances and private data areas.

    This data area is filled for the first time when a table is accessed, either via an SQL statement or via table_info and is not destroyed until the table is dropped or the driver handle is released. Manual destruction is possible via f_clear_meta.

    The following attributes are preserved in the data area and will evaluated instead of driver globals:


    The following attributes are preserved in the data area only and cannot be set globally.


    The following attributes are preserved in the data area only and are computed when initializing the data area:


    For DBD::CSV tables this means, once opened "foo.csv" as table named "foo", another table named "foo" accessing the file "foo.txt" cannot be opened. Accessing "foo" will always access the file "foo.csv" in memorized f_dir, locking f_lockfile via memorized f_lock.

    You can use f_clear_meta or the f_file attribute for a specific table to work around this.

  • When used with SQL::Statement and temporary tables e.g.,


    the table data processing bypasses DBD::File::Table. No file system calls will be made and there are no clashes with existing (file based) tables with the same name. Temporary tables are chosen over file tables, but they will not covered by table_info.


This module is currently maintained by

H.Merijn Brand < h.m.brand at > and Jens Rehsack < rehsack at >

The original author is Jochen Wiedmann.


 Copyright (C) 2009-2013 by H.Merijn Brand & Jens Rehsack
 Copyright (C) 2004-2009 by Jeff Zucker
 Copyright (C) 1998-2004 by Jochen Wiedmann

All rights reserved.

You may freely distribute and/or modify this module under the terms of either the GNU General Public License (GPL) or the Artistic License, as specified in the Perl README file.


DBI, DBD::DBM, DBD::CSV, Text::CSV, Text::CSV_XS, SQL::Statement, and DBI::SQL::Nano