Alien-Selenium
view release on metacpan or search on metacpan
inc/My/Tests/Below.pm view on Meta::CPAN
no pre-processing step (e.g. C<inline2test>) and no temporary file
creation is required with My::Tests::Below. This goes a long ways
towards shortening the debugging cycle (no need to re-run "./Build
code" nor "make" each time)
=item *
L<Test::Inline> has a lot of dependencies, and using it would cause
the installation of small modules to become unduly burdensome.
=back
=cut "
package My::Tests::Below;
use strict;
use File::Temp ();
our $VERSION = 2.0;
## This is done at the top level, not in a sub, as "require
## My::Tests::Below" is what gets the ball rolling:
our $singleton = __PACKAGE__->_parse(\*main::DATA, caller(0));
unless (defined $singleton) {
die "My::Tests::Below invoked, but no tests were found below!";
}
close(main::DATA);
$singleton->run();
## Creates an instance of My::Tests::Below from a source file
## that has tests at the bottom.
sub _parse {
my ($class, $fd, $package, $packfilename, $packline) = @_;
if (!defined $package) { # Handle self-testing case
$package="My::Tests::Below";
$0 =~ m/^(.*)$/; $packfilename = $1;
}
my $self = bless {
package => $package,
packfilename => $packfilename
}, $class;
$self->{testsuite} = do { no warnings; scalar <$fd> };
return undef if ! defined $self->{testsuite};
$self->{testsuite} .= join('',<$fd>); # The rest of it
=head2 Comfort features
Unlike the C<< eval >> form recommended in L<perlmodlib>,
My::Tests::Below provides a couple of comfort features that help
making the system smooth to use for tests.
=over
=item I<Support for code and data snippets in the POD>
A mechanism similar to the now-deprecated L<Pod::Tests> is proposed to
test documented examples such as code fragments in the SYNOPSIS. See
L</CLASS METHODS> below.
=cut
## Parse the whole source file again in order to provide said
## features. Yes, seeking to start also works on main::DATA!
## Gotta love Perl :-)
seek($fd, 0, 0) or die $!; $. = 0;
my $insnippet;
SOURCELINE: while(<$fd>) {
if (m/^=for\s+My::Tests::Below\s+"([^"]*)"(.*)$/) {
my $snipkey = $1; my @args = split m/\s+/, $2;
if (grep { lc($_) eq "end" } @args) {
die qq'Not in an "=for My::Tests::Below" directive'.
qq' at line $.\n' unless (defined $insnippet);
die qq'Badly nested "=for My::Tests::Below" directives'.
qq' at line $.\n' unless ($insnippet eq $snipkey);
undef $insnippet;
} else { # Assume "begin" for compatibility with old tests
die qq'Duplicate "=for My::Tests::Below" directive'.
qq' for label $snipkey at line $.\n' if
(exists $self->{podsnippets}{$snipkey});
die qq'Badly nested "=for My::Tests::Below" directives'.
qq' at line $.\n' if (defined $insnippet);
$self->{podsnippets}{$snipkey}->{lineno} = $. + 1;
$insnippet = $snipkey;
}
} elsif (m/^=for\s+My::Tests::Below/) {
die qq'Parse error in "=for My::Tests::Below"'.
qq' directive at line $.\n';
} else {
$self->{podsnippets}{$insnippet}->{text} .= $_ if (defined $insnippet);
};
next if (defined($packline) && $. <= $packline); # Be sure to
# catch the first marker *after* the require directive, and
# mind the self-test case too.
next SOURCELINE unless (m/^__(END|DATA)__\s+$/);
=item I<Line counting for the debugger>
You can step through the test using a GUI debugger (e.g. perldb in
Emacs) because the line numbers are appropriately translated.
=cut
$self->{testsuite}="#line ".($.+1)." \"$self->{packfilename}\"\n".
$self->{testsuite};
last SOURCELINE;
}
=item I<Tests always start in package main>
The perlmodlib idiomatics puts you either in C<main> or in the package
where the eval was called from, depending on the version of Perl.
=cut
$self->{testsuite}="package main;\n" . $self->{testsuite};
( run in 1.294 second using v1.01-cache-2.11-cpan-b16cb0d3907 )