http://qs1969.pair.com?node_id=417951


in reply to Mutator chaining considered harmful

Chained mutators can Here's a snippet from my own code. This probes a real-time audio application with a five channel OpenGL oscilloscope:
my $s= Scope->new('probe' => $netlist) ->anatrace('in', 'System input', 'active', 1) ->anatrace('out_1', 'Lpf bank1', 'active', 1) ->anatrace('out_2', 'Lpf bank2', 'active', 1024) ->anatrace('out', 'System output', 'active', 1) ->anatrace('bank2', 'Switch pos', 'active', 1) ->generate();
Want to see another channel? Call anatrace again.

I don't think that features with potential downsides should be 'considered harmful'.

It is possible to create a 'little language' out of chained method calls. I avoid creating little languages because I have seen a few of them grow into voracious blood-sucking monsters. I can enjoy chaining method calls without fear of triggering a science-related memetic disorder.

It should work perfectly the first time! - toma

Replies are listed 'Best First'.
Re^2: Mutator chaining considered harmful
by kappa (Chaplain) on Dec 29, 2004 at 14:11 UTC
    This gives you almost nothing in comparison with:
    ... $s->anatrace('bank2', 'Switch pos', 'active', 1); $s->generate();
    And when the objects have long names or are expressions you can always:
    for (Scope->new('probe' => $netlist)) { $_->anatrace('in', 'System input', 'active', 1); ... $_->generate(); }
    --kap
Re^2: Mutator chaining considered harmful
by Aristotle (Chancellor) on Dec 29, 2004 at 19:39 UTC

    Apart from kappa's points: read my root node update. Here's what that would look like with a good interface:

    my $s = Scope->new( 'probe' => $netlist ); $s->anatrace( in => [ 'System input', 'active', 1 ], out_1 => [ 'Lpf bank1', 'active', 1 ], out_2 => [ 'Lpf bank2', 'active', 1024 ], out => [ 'System output', 'active', 1 ], bank2 => [ 'Switch pos', 'active', 1 ], ); $s->generate();

    Makeshifts last the longest.

      I suppose I could also use this :-)
      my $s = Scope->new( probe => $netlist ) ->anatrace( in => [ 'System input', 'active', 1 ], out_1 => [ 'Lpf bank1', 'active', 1 ], out_2 => [ 'Lpf bank2', 'active', 1024 ], out => [ 'System output', 'active', 1 ], bank2 => [ 'Switch pos', 'active', 1 ], ) ->generate();
      The => after the first arguments look nice, especially since the first arguments are instance handles, and I use them as a hash key.

      In this example the arguments to anatrace really aren't mutators, they are constructors of sub-objects. They are waveforms being added to a scope display.

      update
      Is the problem that you perceive just with mutators, or does it also apply to sub-object constructors? Can you provide examples of acceptable method chaining, and contrast it with problematic method chaining?

      It should work perfectly the first time! - toma