man DBIx::DBSchema (3): Database-independent schema objects

SYNOPSIS

use DBIx::DBSchema;
$schema = new DBIx::DBSchema @dbix_dbschema_table_objects;
$schema = new_odbc DBIx::DBSchema $dbh;
$schema = new_odbc DBIx::DBSchema $dsn, $user, $pass;
$schema = new_native DBIx::DBSchema $dbh;
$schema = new_native DBIx::DBSchema $dsn, $user, $pass;
$schema->save("filename");
$schema = load DBIx::DBSchema "filename" or die $DBIx::DBSchema::errstr;
$schema->addtable($dbix_dbschema_table_object);
@table_names = $schema->tables;
$DBIx_DBSchema_table_object = $schema->table("table_name");
@sql = $schema->sql($dbh);
@sql = $schema->sql($dsn, $username, $password);
@sql = $schema->sql($dsn); #doesn't connect to database - less reliable
$perl_code = $schema->pretty_print;
%hash = eval $perl_code;
use DBI qw(:sql_types); $schema = pretty_read DBIx::DBSchema \%hash;

DESCRIPTION

DBIx::DBSchema objects are collections of DBIx::DBSchema::Table objects and represent a database schema.

This module implements an OO-interface to database schemas. Using this module, you can create a database schema with an OO Perl interface. You can read the schema from an existing database. You can save the schema to disk and restore it in a different process. You can write SQL CREATE statements statements for different databases from a single source. You can transform one schema to another, adding any necessary new columns, tables, indices and foreign keys.

Currently supported databases are MySQL, PostgreSQL and SQLite. Sybase and Oracle drivers are partially implemented. DBIx::DBSchema will attempt to use generic SQL syntax for other databases. Assistance adding support for other databases is welcomed. See DBIx::DBSchema::DBD, ``Driver Writer's Guide and Base Class''.

METHODS

new TABLE_OBJECT, TABLE_OBJECT, ...

Creates a new DBIx::DBSchema object.

new_odbc DATABASE_HANDLE | DATA_SOURCE USERNAME PASSWORD [ ATTR ]

Creates a new DBIx::DBSchema object from an existing data source, which can be specified by passing an open DBI database handle, or by passing the DBI data source name, username, and password. This uses the experimental DBI type_info method to create a schema with standard (ODBC) SQL column types that most closely correspond to any non-portable column types. Use this to import a schema that you wish to use with many different database engines. Although primary key and (unique) index information will only be read from databases with DBIx::DBSchema::DBD drivers (currently MySQL and PostgreSQL), import of column names and attributes *should* work for any database. Note that this method only uses ``ODBC'' column types; it does not require or use an ODBC driver.

new_native DATABASE_HANDLE | DATA_SOURCE USERNAME PASSWORD [ ATTR ]

Creates a new DBIx::DBSchema object from an existing data source, which can be specified by passing an open DBI database handle, or by passing the DBI data source name, username and password. This uses database-native methods to read the schema, and will preserve any non-portable column types. The method is only available if there is a DBIx::DBSchema::DBD for the corresponding database engine (currently, MySQL and PostgreSQL).

load FILENAME

Loads a DBIx::DBSchema object from a file. If there is an error, returns false and puts an error message in $DBIx::DBSchema::errstr;

save FILENAME

Saves a DBIx::DBSchema object to a file.

addtable TABLE_OBJECT

Adds the given DBIx::DBSchema::Table object to this DBIx::DBSchema.

tables

Returns a list of the names of all tables.

table TABLENAME

Returns the specified DBIx::DBSchema::Table object.

sql [ DATABASE_HANDLE | DATA_SOURCE [ USERNAME PASSWORD [ ATTR ] ] ]

Returns a list of SQL `CREATE' statements for this schema.

The data source can be specified by passing an open DBI database handle, or by passing the DBI data source name, username and password.

Although the username and password are optional, it is best to call this method with a database handle or data source including a valid username and password - a DBI connection will be opened and used to check the database version as well as for more reliable quoting and type mapping. Note that the database connection will be used passively, not to actually run the CREATE statements.

If passed a DBI data source (or handle) such as `DBI:mysql:database' or `DBI:Pg:dbname=database', will use syntax specific to that database engine. Currently supported databases are MySQL and PostgreSQL.

If not passed a data source (or handle), or if there is no driver for the specified database, will attempt to use generic SQL syntax.

sql_update_schema [ OPTIONS_HASHREF, ] PROTOTYPE_SCHEMA [ DATABASE_HANDLE | DATA_SOURCE [ USERNAME PASSWORD [ ATTR ] ] ]

Returns a list of SQL statements to update this schema so that it is idential to the provided prototype schema, also a DBIx::DBSchema object.

Right now this method knows how to add new tables and alter existing tables, including indices. If specifically requested by passing an options hashref with drop_tables set true before all other arguments, it will also drop tables.

See ``sql_alter_table'' in DBIx::DBSchema::Table, ``sql_add_column'' in DBIx::DBSchema::Column and ``sql_alter_column'' in DBIx::DBSchema::Column for additional specifics and limitations.

The data source can be specified by passing an open DBI database handle, or by passing the DBI data source name, username and password.

If not passed a data source (or handle), or if there is no driver for the specified database, will attempt to use generic SQL syntax.

update_schema [ OPTIONS_HASHREF, ] PROTOTYPE_SCHEMA, DATABASE_HANDLE | DATA_SOURCE [ USERNAME PASSWORD [ ATTR ] ]

Same as sql_update_schema, except actually runs the SQL commands to update the schema. Throws a fatal error if any statement fails.

pretty_print

Returns the data in this schema as Perl source, suitable for assigning to a hash.

pretty_read HASHREF

This method is not recommended. If you need to load and save your schema to a file, see the ``load'' and ``save'' methods.

Creates a schema as specified by a data structure such as that created by pretty_print method.

AUTHORS

Ivan Kohler <[email protected]>

Charles Shapiro <[email protected]> and Mitchell Friedman <[email protected]> contributed the start of a Sybase driver.

Daniel Hanks <[email protected]> contributed the Oracle driver.

Jesse Vincent contributed the SQLite driver and fixes to quiet down internal usage of the old API.

Slaven Rezic <[email protected]> contributed column and table dropping, Pg bugfixes and more.

CONTRIBUTIONS

Contributions are welcome! I'm especially keen on any interest in the top items/projects below under BUGS.

REPOSITORY

The code is available from our public git repository:

  git clone git://git.freeside.biz/DBIx-DBSchema.git

Or on the web:

  http://freeside.biz/gitweb/?p=DBIx-DBSchema.git
  Or:
  http://freeside.biz/gitlist/DBIx-DBSchema.git

COPYRIGHT

Copyright (c) 2000-2007 Ivan Kohler Copyright (c) 2000 Mail Abuse Prevention System LLC Copyright (c) 2007-2015 Freeside Internet Services, Inc. All rights reserved. This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.

BUGS AND TODO

Multiple primary keys are not yet supported.

Foreign keys: need to support dropping, NOT VALID, reverse engineering w/mysql

Need to port and test with additional databases

Each DBIx::DBSchema object should have a name which corresponds to its name within the SQL database engine (DBI data source).

Need to support ``using'' index attribute in pretty_read and in reverse engineering

sql CREATE TABLE output should convert integers (i.e. use DBI qw(:sql_types);) to local types using DBI->type_info plus a hash to fudge things

PRETTY_ BUGS

pretty_print is actually pretty ugly.

pretty_print isn't so good about quoting values... save/load is a much better alternative to using pretty_print/pretty_read

pretty_read is pretty ugly too.

pretty_read should *not* create and pass in old-style unique/index indices when nothing is given in the read.

Perhaps pretty_read should eval column types so that we can use DBI qw(:sql_types) here instead of externally.

perhaps we should just get rid of pretty_read entirely. pretty_print is useful for debugging, but pretty_read is pretty bunk.