# from Jonathan Swartz
# on Friday 26 October 2007 06:53:
>I'd like to avoid actually running a single script per class, for
>efficiency reasons - i.e. I agree with Ovid and Adrian here:
> http://use.perl.org/~Ovid/journal/31172
>
>It seems like it should be possible (easy?) to create a new
>TAP::Harness that treated each Test::Class class as its own "script"
>in the output.
Building it as a TAP::Harness plugin sort-of implies that the tap stream
contains diagnostics which tie the results to a given class.
Perhaps a solution which involves a pre-fork model of Test::Class would
be a little neater. The "test scripts" would then be sub-sub-processes
and the harness just needs filehandles from each of those.
# $self is (roughly) a TAP::Harness
my $driver = Test::ClassDriver->new(%some_params);
foreach my $test ($driver->classes) {
my $name = $test->name;
my $out = $test->stdout; # initiates the sub-sub-process
$self->process_results($name, $out);
}
This is just a quick scribble to illustrate what I mean, but I think
$driver could fairly easily be a process which had pre-loaded all of
the classes.
If each $test were a fork, it would conceivably parallelize just by
partitioning $driver->classes.
--Eric
--
"Left to themselves, things tend to go from bad to worse."
--Murphy's Corollary
---------------------------------------------------
http://scratchcomputing.com
---------------------------------------------------
