Bug #78270 for IO-Pager: Support C<say> => Unicode & non-paged robustness

Show quoted text
It does return an object, the object is of the subclass type instantiated. For instance, the tests are replete with isa_ok IO::Pager::Buffered, etc. Oh, you mean when it's piped... I don't think it should return/implement a sub-class actually, since it wouldn't be one. I could see it maybe returning an IO::String object for non-interactive, but you'd still have to or your test.

I've not yet figured out the best way to implement a test for it, nor documented it, but the latest commit falls back to IO::Handle IFF C<new> was used to invoke IO::Pager, thus preserving transparent procedural programming if using C<open> and adding transparent OO programming. Feel free to kick the tires

On Jul 12, 2012, at 2:53 AM, belg4mit@pthbb.org via RT wrote: Show quoted text
There are two problems with this. The first is that the passed file handle is ignored. You can fix that with this patch: diff --git a/lib/IO/Pager/Buffered.pm b/lib/IO/Pager/Buffered.pm index 9cc2360..29347e5 100644 --- a/lib/IO/Pager/Buffered.pm +++ b/lib/IO/Pager/Buffered.pm @@ -16,7 +16,8 @@ sub new(;$) { # [FH] return $_[1] if defined($_[2]) && $_[2] eq 'procedural'; #...fall back to IO::Handle for transparent OO programming - eval "require IO::Handle" && return new IO::Handle; + eval "require IO::Handle" or die $@; + return IO::Handle->new_from_fd(fileno($_[1]), 'w'); } $!=$@, return 0 if $@ =~ 'pipe'; diff --git a/lib/IO/Pager/Unbuffered.pm b/lib/IO/Pager/Unbuffered.pm index 67e06a7..7d043fa 100644 --- a/lib/IO/Pager/Unbuffered.pm +++ b/lib/IO/Pager/Unbuffered.pm @@ -16,7 +16,8 @@ sub new(;$) { # [FH], procedural return $_[1] if defined($_[2]) && $_[2] eq 'procedural'; #...fall back to IO::Handle for transparent OO programming - eval "require IO::Handle" && return new IO::Handle; + eval "require IO::Handle" or die $@; + return IO::Handle->new_from_fd(fileno($_[1]), 'w'); } $!=$@, return 0 if $@ =~ 'pipe'; The second problem is that IO::Handle and IO::Pager do not share a common base class, as far as I can tell. While returning an object is nice, if it does not derive from some class that IO::Pager also derives from, the promise of new() is still somewhat broken, IMHO. I am not sure what one can do about this, though, since Tie::Handle doesn't derive from IO::Handle, either. :-( David

On Jul 12, 2012, at 4:46 PM, belg4mit@pthbb.org via RT wrote: Show quoted text
It means I have to keep some sort of `where` statement in the Moose type definition, something like this: isa => type('IO::Pager' => where { eval { $_->isa('IO::Pager') || $_->isa('IO::Handle') } }), Though I could probably use Moose::Meta::TypeConstraint::DuckType, instead: isa => duck_type([qw(say print printf)]), So I can live with that. Show quoted text
It just seems weird to me. I wouldn't expect `DateTime->new` to return a Date::Object object, you know? Show quoted text
Yeah. The only "correct" thing to do would probably be to add another subclass, IO::Pager::Unpaged or something, that inherits from IO::Pager and either provides the TIED implementation to print to the real file handle, or else also inherits from IO::Handle. HTH, David

On 2012-07-12 11:33:10, belg4mit@pthbb.org wrote: Show quoted text
*BUMP*. Had a chance to give it some thought? —Theory

26 letters should be enough for anyone X-P I'll add this to the list...

Updated subject. Should be able to round up some tuits this weekend. I don't expect -CAS to be too hard, perlrun(1) documents ${^UNICODE} The bigger pain is an identical feature-set for non-paged output, without reimplimenting everything a second time. IO::Handle actually turns out to be largely useless for this purpose. I may just have to bite the bullet though, and do that. If I'm lucky I can just replace the init method in the non-paged class (IO::Pager::less = pagerless). On the other hand, as a result of tackling this I've refactored some of the code to not be *quite* so hairy, and the various new()s and open()s in the current code (not yet push'd) are more rational.

Perhaps, but I'd argue that the order of operations in your example is "wrong" since there's a working binmode method that can be applied to the object. I can try to use PerlIO::get_layers($fh), but that would seem to preclude any sort of (incidental) support for stdio perls.

On Aug 2, 2012, at 7:23 PM, Jerrad Pierce via RT wrote: Show quoted text
Oh, maybe. Seems weird, though, that if I passed a file handle that happened to already have been `binmode`d (not unusual for STDOUT; see utf8::all), and then have to do it again after getting the object back. David

On Aug 22, 2012, at 8:32 PM, Jerrad Pierce via RT wrote: Show quoted text
This is the main thing I want to see fixed in a new release. Once it works, Sqitch will be built as a PPM and available to all ActivePerl users, which would be very cool. Show quoted text
Sounds like a release of what you have so far would be very useful in the short term, eh? Thanks, David

On Aug 23, 2012, at 4:07 PM, belg4mit@pthbb.org via RT wrote: Show quoted text
I updated my clone. Test 12 fails on OS X: PERL_DL_NONLAZY=1 /usr/local/bin/perl "-MExtUtils::Command::MM" "-e" "test_harness(0, 'blib/lib', 'blib/arch')" t/*.t t/01-load.t ........................ ok t/02-which.t ....................... ok t/02-which_interactive.t ........... ok t/03-bald_interactive.t ............ ok t/04-buffered_interactive.t ........ ok t/05-binmode_interactive.t ......... ok t/06-scalar_interactive.t .......... ok t/07-oo_interactive.t .............. ok t/08-redirect.t .................... ok t/09-open.t ........................ ok t/10-close_interactive.t ........... ok t/11-redirect-oo.t ................. ok t/12-preservelayers_interactive.t .. print() on unopened filehandle GEN0 at /usr/local/lib/perl5/5.16.0/darwin-thread-multi-2level/IO/Handle.pm line 430. print() on unopened filehandle GEN0 at /usr/local/lib/perl5/5.16.0/darwin-thread-multi-2level/IO/Handle.pm line 430. close() on unopened filehandle GEN0 at /usr/local/lib/perl5/5.16.0/darwin-thread-multi-2level/IO/Handle.pm line 384. t/12-preservelayers_interactive.t .. Failed 1/1 subtests Test Summary Report ------------------- t/12-preservelayers_interactive.t (Wstat: 0 Tests: 0 Failed: 0) Parse errors: Bad plan. You planned 1 tests but ran 0. Files=13, Tests=21, 7 wallclock secs ( 0.07 usr 0.04 sys + 1.71 cusr 0.13 csys = 1.95 CPU) Result: FAIL Failed 1/13 test programs. 0/21 subtests failed. make: *** [test_dynamic] Error 255 BTW, you might want to ask Christian Walde (MITHALDU) to take it for a spin on Win32 via RT#75181. Best, David

On Aug 23, 2012, at 5:35 PM, belg4mit@pthbb.org via RT wrote: Show quoted text
Yes, I am prompted. If I hit Y it passes. If I hit n it fails. It should not prompt at all, of course. David

Great, seems to work for me, thanks! I will watch for your release. Best, David

Wtf? I just wrote last niht that they'd been uploaded to CPAN Did you remove yourself from the Window bug requestors? Here it is http://search.cpan.org/~jpgierce/IO-Pager-0.30/

You mean to roll a new PPM?