Execute a callback before attempting DBI->connect

[rwstauner]

Separating this idea from #19, accepting a coderef that DBIx::Connector would call before calling DBI->connect could be a handy feature.

After reading DBI docs and code I found that this effect can be achieved with something like:

DBIx::Connector->new('dbi:SQLite:dbname=:memory:', '', '',
    { dbi_connect_method => sub { do_something; shift->connect(@_) } }) 

but that seems a bit hackish and slightly cumbersome: Each time you have to remember to also do the shift->connect(@_) part. Also passing a coderef as dbi_connect_method is undocumented, and making a separate sub or class each time would be even more cumbersome.

What do you think? Or perhaps do you have any other pearls of wisdom I've overlooked again? :-)

Thanks!

[theory]

Why wouldn't you just execute said code before calling DBIx::Connector->new?

Otherwise, I recommend using connect_cached and the accompanying connect_cached.new and connect_cached.reused callbacks, which are called before connecting.

If you can't use connect_cached, maybe ask @timbunce if dbi_connect_method could be officially blessed for this purpose in the docs?

[rwstauner]

I was thinking about doing something for preparing the connection.

Some example situations I've run into:

• need to fork and exec java server before attempting to use DBD::JDBC
• need to (re)open an ssh tunnel before connecting to the database

I've since changed the java thing to be run by a daemon that tries to handle keeping java running so that the dbi-using code isn't concerned with it. It recently occurred to me that perhaps one solution for the ssh tunnel could be setting up a dbi proxy server that would manage that (aside from the other possibility of merely running a separate process to manage it).

I also didn't realize that connect_cached.new actually fired before the connect rather than after, so that would work. Since I'd definitely like to minimize the goofy cases like these, I'll go that route.

I have one other old project that sets up subclasses and uses connect_cached. I've been scheming about how to replace some of that code with DBIx::Connector. Combining connect_cached.new with RootClass looks like everything should be doable through DBIx::Connector, so thank you very much!

{ package MYDBI; package MYDBI::db; our @ISA = qw(DBI::db);
  sub connected { print STDERR "hi mom!"; } package MYDBI::st; our @ISA = qw(DBI::st); };
$INC{"MYDBI.pm"} = 1;
my $dbc = DBIx::Connector->new('dbi:SQLite:dbname=:memory:', '', '', {
    dbi_connect_method => "connect_cached",
    Callbacks => { "connect_cached.new" => sub { print STDERR "CONNECTING"; } },
    RootClass => "MYDBI",
});

I apologize again for bothering you with these questions, but I really appreciate your showing me how to achieve the same effect through existing interfaces. Thanks for the help, and for DBIx::Connector which I'm looking forward to using more... but don't worry, I think I'm done asking you for things about it :-)

[theory]

Bah! I just realized that you can't use the connect_cached.* callbacks with DBIx::Connector, because it calls connect. @timbunce is there any other way to get a callback in before the actual connection code? As Randy noted, dbi_connect_method is not documented. Maybe it could be? Or some other callback name could be created to support it?

[rwstauner]

Setting dbi_connect_method to connect_cached (or any other function name) as in the example in my last comment is documented:

The C<dbi_connect_method> attribute can be used to specify which driver
method should be called to establish the connection. The only useful
values are 'connect', 'connect_cached', or some specialized case like
'Apache::DBI::connect' (which is automatically the default when running
within Apache).

However the docs say nothing about using an anonymous coderef as in my first comment.

[theory]

Oh, so not complete undocumented, and you can coerce into using connect_cached. That's kind of cool. Would be nice to document it as available for callbacks, too. If, of course, it is.

[timbunce]

I'll take a doc patch to officially bless using a code ref for dbi_connect_method (asuming it work, I've not tried/looked).

For this use case perhaps DBIx::Connector->new could take an extra argument, or notice a dbix_connect_foo => ... in the DBI connect attributes. Either way, DBIx::Connector->new could take that argument and turn it into a dbi_connectmethod code ref that calls the supplied code ref and then does shift->connect(@);

[theory]

Meh. I have steered clear of adding parameters to DBIx::Connector->new. Only those supported by DBI->connect are allowed. There are a couple of other things one can change in the DBIx::Connector object, but only via accessors. And although it does muck with a few DBI params, it does not change their meaning at all.