SYNOPSIS

use DBI;
use SQL::Translator;
my $dbh = DBI->connect('dsn', 'user', 'pass',
{
RaiseError => 1,
FetchHashKeyName => 'NAME_lc',
}
);
my $translator = SQL::Translator->new(
parser => 'DBI',
parser_args => {
dbh => $dbh,
},
);

Or:

  use SQL::Translator;
  my $translator      =  SQL::Translator->new(
      parser          => 'DBI',
      parser_args     => {
          dsn         => 'dbi:mysql:FOO',
          db_user     => 'guest',
          db_password => 'password',
    }
  );

DESCRIPTION

This parser accepts an open database handle (or the arguments to create one) and queries the database directly for the information.

The following are acceptable arguments:

• dbh

  An open DBI database handle. NB: Be sure to create the database with the ``FetchHashKeyName => 'NAME_lc''' option as all the DBI parsers expect lowercased column names.

• dsn

  The DSN to use for connecting to a database.

• db_user

  The user name to use for connecting to a database.

• db_password

  The password to use for connecting to a database.

There is no need to specify which type of database you are querying as this is determined automatically by inspecting $dbh->{'Driver'}{'Name'}. If a parser exists for your database, it will be used automatically; if not, the code will fail automatically (and you can write the parser and contribute it to the project!).

Currently parsers exist for the following databases:

• MySQL
• SQLite
• Sybase
• PostgreSQL (still experimental)

Most of these parsers are able to query the database directly for the structure rather than parsing a text file. For large schemas, this is probably orders of magnitude faster than traditional parsing (which uses Parse::RecDescent, an amazing module but really quite slow).

Though no Oracle parser currently exists, it would be fairly easy to query an Oracle database directly by using DDL::Oracle to generate a DDL for the schema and then using the normal Oracle parser on this. Perhaps future versions of SQL::Translator will include the ability to query Oracle directly and skip the parsing of a text file, too.

AUTHOR

Ken Y. Clark <[email protected]>.