|
NAMETest::ManyParams - module to test many params as one testSYNOPSISuse Test::ManyParams; all_ok {foo(@_)} [ [$arg1a, $arg2a], [$arg2b, $arg2b, $arg3b, $arg4b] ], "Testing that foo returns true for every combination of the arguments"; all_ok {bar(shift())} [qw/arg1 arg2 arg3 arg4 arg5 arg6/], "Testing every argument with bar"; all_are CODE VALUE, PARAMETERS, [ TEST_NAME ] all_arent CODE VALUE, PARAMETERS, [ TEST_NAME ] any_ok {drunken_person() eq shift()} ["Jim Beam", "Jonny Walker", "Jack Daniels"]; any_is {ask_for_sense_of_life(shift())} 42, ["Jack London", "Douglas Adams"]; most_ok {$img->colorAt($_[0],$_[1]) == BLACK} [ [0 .. 10_000], [0 .. 10_000] ] => 100, "100 random pixels of a black image should be black"; [NOT YET IMPLEMENTED] all_are_deeply CODE SCALAR, PARAMETERS, [ TEST_NAME ] all_like CODE REGEXP, PARAMETERS, [ TEST_NAME ] all_unlike CODE REGEXP, PARAMETERS, [ TEST_NAME ] all_can CODE METHODS, PARAMETERS, [ TEST_NAME ] all_dies_ok CODE PARAMETERS, [TEST_NAME] all_lives_ok CODE PARAMETERS, [TEST_NAME] all_throws_ok CODE REGEXP, PARAMETERS, [TEST_NAME] DESCRIPTIONGENERAL PRINCIPLESThis module helps to tests many parameters at once. In general, it calls the given subroutine with every combination of the given parameter values. The combinations are created with building a cross product.Especially it avoids writing ugly, boring code like: my $ok = 1; foreach my $x ($arg1a, $arg2a) { foreach my $y ($arg2a, $arg2b, $arg3b, $arg4b) { $ok &&= foo($x,$y); } } ok $ok, $testname; Instead you simpler write all_ok {foo(@_)} [ [$arg1a, $arg2a], [$arg2b, $arg2b, $arg3b, $arg4b] ] $testname; Additionally the output contains also some useful information about the parameters that should be tested and the first parameters the test failed. E.g. all_ok {$_[0] != 13 and $_[1] != 13} [ [1 .. 100], [1 .. 100] ], "No double bad luck"; would print: not ok 1 - No double bad luck # Failed test (x.pl at line 5) # Tests with the parameters: $VAR1=[[10,11,12,13,14,15],[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]]; # Failed first using these parameters: $VAR1=[10,13]; The parameters passed to "all_ok" can be passed in two ways. If you need to test a crossproduct of more than one parameterlist, you have to write it as all_ok CODE [ \@arglist1, \@arglist2, ..., \@arglistn ], TEST_NAME; The CODE-routine will be called with every combination of the arglists, passed as simple arguments. E.g. all_ok {foo(@_)} [ ["red", "green", "blue"], ["big", "medium", "little"] ]; would call foo("red","big"); foo("red","medium"); foo("red","little"); foo("green","big"); foo("green","medium"); foo("green","little"); foo("blue","big"); foo("blue","medium"); foo("blue","little"); Note, that the order of calling shouldn't play any role, as it could be changed in future versions without any notice about. Please always remember, that a crossproduct of the lists can be very, very big. So don't write something like "all_ok {&foo} [ [1 .. 32000], [1 .. 32000], [1 .. 32000] ]", as it would test 32_768_000_000_000 parameter combinations. If you only want to test one parameter with different values, you can write in general all_ok CODE \@values, TEST_NAME So "all_ok {&foo} [1,2,3]" would call "foo(1); foo(2); foo(3)". (In this way and only in this, $_ will be set to the passed argument. Thus it is "$_ = $_[0]" in this special case. Reason for this behaviour is just convenience.) Please take care, that the first element of the values list isn't an array ref, as Test::ManyParams would assume that you want to test combinations of the above. If it is important to pass values that are array refs, you have to write it this way: my @values = ( [1 .. 10], [100 .. 110], [990 .. 1000] ); all_ok {&foo} [ [@values] ]; # calls foo(1), ... foo(10), foo(100), ..., foo(110), foo(990), ..., foo(1000) what is very different to all_ok {&foo} [ @values ]; # what would call foo(1,100,900), foo(1,100,901), ... Of course, the test name is always optional, but recommended. FUNCTIONS
IMPORTINGIn the most cases, you will simply import this module ("use Test::ManyParams").But when you want to set the seed for the randomization of the "most_ok" method, the way for importing looks like "use Test::ManyParams seed =" 42>. At the time, it only will call "srand(42)", but later on, it will hold foreach test script and file it's own random numbers, so that several modules all using this Test::ManyParams module won't collide. If you don't seet a seed value, at default the "time ^ $$" value will be taken as seed (the default value could be changed in future versions without any notice). That's not a very good seed, but I hope it will be good and quick enough for the most cases. You can access the setted seed with the variable $Test::ManyParams::seed. That's a non-exported, readonly variable. Please take care only to import Test::ManyParams one times in one package, as the use command is executed while compile time and so different seedings will be confusing. EXPORT"all_ok" "all_are" "all_arent" "any_is" "any_isnt" "most_ok"BUGSThe representation of the parameters uses Data::Dumper. As this module neither set $Data::Dumper::Indent, nor reads it out, setting $Data::Dumper::Indent to some strange values can destroy a useful parameter outprint. I don't plan to fix this behaviour in the next time, as I there a more important things to do. (Who changes global variables harvest what he/she/it has seed.)The "most_ok" method is hard to test. I wrote some tests that helped to remove the obviously bugs, but there could be some subtle bugs. Especially I didn't test what happens, if you try to test more parameters than there could be. There are perhaps many mistakes in this documentation. Please tell me everything you can find. TODOThere are a lot of methods I'd like to implement still. The most of them are simple Here's a list of them:
Similar methods are planned with the prefix any_. The "most_ok" method should accept also a percentage rate and a time slot or ... for specification of how many tests should be run. Typical examples for that could be '1000', '1%', '5s', '100 + bounds'. The pro and contra of had been discussed a bit on perl.qa. One of the results is that random parameter tests are very sensful, but the reproducibility is very important. So the module has to seed (or recognise the seed) of the random generator and to give the possibility to set them (e.g. with "use Test::ManyParams seed =" 42>). Recognising a failed test, this seed has to be printed. It always seems to be sensful to set an own random numbering for each package using this module. The last part still has to be done. That's only a short synopsis of this discussion, it will be better explained when these features are built in. Of course, there will be also some methods like "most_are","most_arent",... . SEE ALSOThis module had been and will be discussed on perl.qa.Test::More Test::Exception THANKSThanks to Nicholas Clark and to Tels (http://www.bloodgate.com) for giving a lot of constructive ideas.AUTHORJanek Schleicher, <bigj@kamelfreund.de>COPYRIGHT AND LICENSECopyright 2002 by Janek SchleicherThis library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.
Visit the GSP FreeBSD Man Page Interface. |