Clean up, no moose here.

[catagits/Catalyst-Devel.git] / gsoc / gsoc-devel-plan-5-23.txt

15:36 <t0m> right. so the first thing to do is probably write a test to check 
            generating an application doesn't change.
15:37 <dhoss> per app?
15:38 <t0m> no, I mean whilst you're ripping ::Helper apart.
15:38 <dhoss> OH
15:38 <dhoss> gotcha 
15:38 <t0m> if you install your branch.. Then catalyst.pl t/TestAppForComparison
15:39 <dhoss> so we have the same results now as before, yes?
15:39 <t0m> then you write a test which says catalyst.pl /tmp/TempTestApp; my 
            $diff = `diff -urN t/TestAppForComparison /tmp/TempTestApp`; ok 
            !length($diff) or warn $diff;
15:40 <t0m> exactly. and that'll test you rip all of the files out of Helper.pm 
            correctly..
15:40 <dhoss> okay, TestAppForComparison created. i'll commit that initial 
              version
15:40 <t0m> you'll probably want to nuke it after that, as testing the 
            generated strings of an entire app dir is gonna be a PITA
15:41 <t0m> to maintain
15:41 <dhoss> i could see that
15:41 <t0m> But it's totally the right solution whilst you're trying to rip 
            stuff out without changing it
15:41 <dhoss> temp testing :-D
15:44 <t0m> I've got some ideas on doing better testing that the TestApp we 
            generate is good.
15:44 <t0m> but they're a bit more involved, and not needed for this at all ;)
15:45 <dhoss> this being this refactor, or this bit of the refactor?
15:45 <t0m> the latter.
15:45 <dhoss> okay - got it
15:46 <t0m> and once this is done - you can branch again to work on the code, 
            and again to work on the generated app templates (as we'll want to 
            moosify those, so your app is moosetastic by default
15:46 <t0m> and as it's all in different files - you can do both in parallel as 
            you feel like and safely merge back
15:47 <t0m> did you get to grips with MX::GetOpt and do you have a plan for 
            that yet?
15:48 <dhoss> so one branch for helper_refactor, one for code (which code?), 
              and one for app templates
15:48 <t0m> As I guess catalyst.pl should be taught to use it, and so should 
            the generated scripts :)
15:48 <dhoss> i've been playing with MX::GetOpt, i need a few more details on 
              what exactly needs to be beat up
15:48 <t0m> yy, but you branch from completed helper_refactor :)
15:49 <t0m> and 'code' is the actual refactor / api changes for helpers
15:49 <t0m> bletch
15:50 <dhoss> okay, so the 'code' branch will be simply the C::H code branched 
              from helper_refactor
15:50 <t0m> get_file has to keep working as everything using ::Helper needs it
15:50 <t0m> yes, exactly.
15:51 <t0m> so you need to make a get_file_from_sharedir function or something
15:51 <dhoss> okay, making sense now :-)
15:51 <t0m> and change things over to use that.
15:51 <t0m> that's fine though - means you can extract 1 file at a time
15:52 <dhoss> which would be better, no? better atomicity and for testing
15:52 <t0m> yy
15:52 <t0m> bletch was at the fact that 'get_file' is a good method name, and 
            it's now deprecated
15:53 <dhoss> is that in C::H?
15:53 <t0m> yy
15:53  * dhoss is a little behind in my api knowledge
15:53 <dhoss> off the top of my head at least
15:53 <t0m> it's the 'get me a file out the <DATA> section' method
15:53 <dhoss> aha right
15:54 <t0m> which you're replacing in favour of 'get me a file out of the share 
            dir'
15:54 <t0m> and 'copy file out of sharedir' methods.
15:54 <t0m> The former for templates
15:54 <t0m> the latter for binaries, which gets rid of the hateful decoding and 
            writing out business
15:54 <dhoss> get -> we're going to interpolate shit, copy -> we're giong to 
              use directly?
15:55 <t0m> just so
15:55 <t0m> eliminating the 'please turn this giant hex string back into 
            binary' thing
15:55 <dhoss> definitely :-P
16:00 <t0m> sub _mk_favicon {
16:00 <t0m>     my ( $self ) = @_;
16:00 <t0m>     $self->copy_share_file(
16:00 <t0m>         File::Spec->catfile( ( caller(0) )[0], 'favicon.ico' ),
16:00 <t0m>         File::Spec->catfile( $self->{root}, "favicon.ico" )
16:00 <t0m>     );
16:00 <t0m> }
16:00 <t0m> sumfin like dat
16:01 <dhoss> easy enough
16:01 <t0m> the actually, all the caller crap dies too
16:02 <dhoss> hrm, i suppose we have to keep the atrocious 
              Class::Accessor::Fast method name artifacts?
16:02 <t0m> that's to clue get_file what file to borg crap out off
16:03 <t0m> I vote that most of them can burn, as we'll be working out what 
            shit to copy a lot more dynamically
16:03 <t0m> and the ones which can't get deprecated, but preserved.
16:04 <dhoss> so that's just for the current get_file, right? 
16:04 <t0m> yy
########################################################################
## done
16:04 <t0m>     $self->copy_share_file(                                 
16:04 <t0m>         'favicon.ico',
16:04 <t0m>         File::Spec->catfile( $self->{root}, "favicon.ico" )
16:04 <t0m>     );
#########################################################################
16:05 <t0m> more like it
16:05 <t0m> so we have sub mk_foo { shift->generate_foo(@_); }

====
5-28
03:04 < dhoss> SO. official next step: 1) #done trim crap out 2) #done write test for how 
               helpers are invoking Catalyst::Helper 3)??? 4) profit!
03:05 < t0m> 3)#done Re arrange files in sharedir so we can make the comversion guts 
             totally generic
03:05 < t0m> 4) merge&profit
03:05 < t0m> *conversion
03:05 < dhoss> rearrange meaning mv foo MyApp?
03:06 < t0m> erm, yes, as described above, so that the contents of share/ is 
             laid out the same as a newly generated MyApp
03:06 < dhoss> OH
03:06 < dhoss> okay
03:06 < dhoss> instead of my current fuckery :-)
03:06 < dhoss> autarch: btw if you have an idea for that i'd love an email to 
               the one in topic
03:06 < t0m> cause then the entire generate application becomes 'copy this 
             directory from X => Y, applying some file mangling rules)
             
=============================================================================
04:26 < t0m> dhoss: erm, yeah - so the two things to do is: unit tests for the 
             methods like get_file which we're not actually using any more..
04:27 < t0m> (do that by making t/lib/ExampleHelper.pm - copy the TT one from 
             CPAN or something, and then write some tests that get_file can 
             successfully pull chunks out the __DATA__ segment
04:28 < t0m> and rename everything in the sharedir to be named as-per where it 
             ends up in your app, but suffixed by .tt or .bin
04:29 < t0m> e.g. share/Makefile.PL.tt share/root/favicon.ico.bin 
             share/lib/MyApp.pm.tt
04:30 < t0m> the exact reasons for this nameing scheme become aparent post 
             first merge, when we remove most of the code :)
04:30 < t0m> but all of share/ needs to be in it's final resting place pre-merge
04:31 < marcus> http://lumberjaph.net/blog/index.php/2009/05/30/catalystxdispatcherasgraph/ # <3 lazyweb
04:31 < t0m> so tests the stuff we're no longer using in Helper.pm (e.g. 
             get_file) still works. Rename share/ bits
04:31 < t0m> marcus: Indeed :)
04:31 < t0m> ^^ those two, then first merge.
23:36 <@kd> dhoss: just found a bug in ::Devel ...
23:36 <@kd> we're using #!/usr/bin/env perl now
23:36 <@kd> that means we have to add the use warnings; pragma beneath it
23:37 <@kd> that doesn't appear to be happeining in the scripts
23:37 <@kd> please fix
23:38 <@kd> dhoss: i.e. http://gist.github.com/122081
==============================================================================

# DONE
1. Rename everything into final layout:
  . tt files all named .tt
  . none tt files all named .bin
  . directory structure as generated, e.g. share/lib/MyApp.pm.tt, share/root/favicon.ico.bin

2. get new layout working


# DONE
3. get tests for back compat - i.e. the methods no longer used in Helper.pm like 'get_file'
  . copy Catalyst::Helper::View::TT into t/lib/TestBackCompat.pm
  . write test t/deprecated_methods_backcompat.t which says:
     use FindBin qw/$Bin/;
     use lib "$Bin/lib";
     use Test::More tests => 1;

     my $helper = TestBackCompat->new( %maybe_some_params_here );
     my $a_section = $helper->get_file('FileName');
     is $a_section, 'FILECONTENTSFROM__DATA__';

3. RFC - DONE

Looking ahead

MooseX::Getopt

################################################################################
17:46 < t0m> dhoss: erm, oh, so the _other_ thing I didn't think of
17:46 < t0m> Is Cat works on windows
17:46 < t0m> I have stuffed loads of "t/foo.t"
17:46 < t0m> into strings.
17:46 < t0m> which will break win
17:47 < t0m> so, all those places where I just "string/append/a/file.name"
17:47 < t0m> you need to use Path::Class qw/file/; file(qw/ string append a 
             file.name /);
17:47 < t0m> like ^^^
17:48 < t0m> then it will still work on win32
17:48 < dhoss> noted
17:48 < t0m> There is already File::Spec->catfile
17:48 < t0m> which does the same thing
17:48 < t0m> but more ugly
17:48 < t0m> either/or
17:48 < purl> either/or is "public load_first_existing_class" or "load_class 
              passes multiple args to load_first_existing_class"
17:48 < t0m> don't care...
17:49 < t0m> Just "file/name" is BAD
17:49 < t0m> and will hate windows users

Summary:
17:43 < dhoss> 1. fix pod coverage, 2. remove TestAppForInvocation 3. ??? # DONE
17:43 < t0m> 3. merge # DONE
################################################################################
--- Log opened Wed Jun 17 18:23:06 2009
18:23 -!- Irssi: Starting query in perl with t0m
18:23 <t0m> yo. wanna pick my brain affore I crash?
18:23 <t0m> I wanna branch, and start hacking local::lib stuff
18:23 <t0m> but that'll do a 4am on me again
18:24 <dhoss> okay, yes
18:25 <dhoss> i stopped after the last thing we talked about, the Moose-ification of Helper.pm. started that but nothing significant
18:26 <t0m> no worries. There's at least 2 things you can work on right now obviously
18:26 <t0m> one being that
18:26 <t0m> one being hacking the generated
18:26 <t0m> +code
18:26 <t0m> one being tests for the generation stuff
18:27 <t0m> at least 2 :)
18:27 <dhoss> :-D
18:27 <dhoss> lemme check hiveminder
18:28 <t0m> but yeah, first 2 can be independent branches. +tests can just go in trunk
18:29 <dhoss> so branch again
18:29 <dhoss> and work on scripts
18:29 <dhoss> which are tasks...hold on
18:30 <t0m> and branch again for moosify helper. if you want. I mean work on what you wanna work on - I'm just saying after stage 1, there is a whole spread of stuff to do now
18:30 <dhoss> #PYMI,#PYM9?
18:30 <dhoss> yea, i'd love to moosify the helper
18:30 <t0m> Also, probably not enough todos to cover it - there is removing all the hardcoded filenames...
18:31 <t0m> so many stuff.
18:31 <t0m> cool, do that then :)
18:31 <t0m> then we'll actually refactor/rewrite it
18:31 <t0m> so all those _mk_foo things go away
18:31 <dhoss> okay, sweet, yea i forgot the about the hardcoded shits
18:31 <dhoss> okay, so my famous list:
18:31 <t0m> that's going to involve some fun Module::Install hacking ;)
18:32 <dhoss> making _mk_blargh go away is?
18:32 <t0m> just so
18:32 <t0m> oh, sorry - yeah. _mk_favicon
18:32 <dhoss> just so == yes? (sorry, my crosspondia lang isn't up to snuff)
18:32 <t0m> etc etc etc
18:33 <t0m> just so = yes = exactly like you descibed but I misread what you said.
18:33 <dhoss> ahaha :-)
18:33 <t0m> basically, the code has sub _mk_foo { render('lib/foo..pm.tt', 'lib/foo.pm') } (pseudocode)
18:34 <dhoss> okay, so: 1. #PMYI 2. #PYM9 3. Moosification of halper
18:35 <t0m> if you could do foreach my $from_fn_-_to_fn ( $self->get_list_of_things_in_sharedir ) {
18:35 <t0m> ^^ That would suck so much less
18:36 <dhoss> yea i've been looking at something like that
18:38 <t0m> that's fairly easy, with the way we did the rename
18:38 <t0m> initially, if you just found everything not /\.tt$/
18:38 <dhoss> exactly, phrew 
18:39 <t0m> and transplanted them
18:39 <t0m> that would be a good start
18:39 <t0m> in fact, you could add that
18:39 <t0m> not take any code away
18:39 <t0m> just copy the files twice
18:39 <t0m> not fail tests
18:39 <t0m> commit
18:39 <dhoss> copy the files twice?
18:39 <t0m> take away the _mk_favicon etc methods one by one
18:40 <dhoss> aha, and add that loop
18:40 <t0m> well, if you run foreach my $binfile ($self->find_all_bin_files) { $self->copybinfile_to_app($binfile) }
18:40 <t0m> (pseudocode)
18:41 <t0m> ^^ Adding that to the current generation process won't break anything
18:41 <t0m> You just generate the favicon.ico (etc) twice
18:41 <t0m> once in that loop
18:41 <t0m> once in pre-existing $self->_mk_favicon
18:41 <t0m> in the same place..
18:41 <t0m> so you overwrite it the 2nd time
18:41 <dhoss> okay, so *leave* _mk_blargh and copy it into the for loop, i get it
18:42 <t0m> well, do the generic thing in the for loop
18:42 <t0m> in fact - here is a smart move. You know the 'check generated files' test?
18:42 <dhoss> yes
18:42 <t0m> unlink each file
18:43 <t0m> after you pass a test as it exists
18:43 <t0m> then add an extra test doing a File::Find
18:43 <t0m> and fail if it finds anything
18:43 <dhoss> so check file, unlink, then another test to look for leftovers
18:44 <t0m> exactly
18:44 <t0m> then you add something generic to copy bin files
18:44 <dhoss> and, forgive me, but what does that help us make sure we've done?
18:45 <t0m> Makes sure you're not suddenly generating t/01app.t and t/01/app.t.tt # Oops, you fucked up the 'what file name do I generate test'
18:45 <t0m> s/test'$/code'/
18:46 <t0m> so you don't ever accidentally generate duplicate files with diff names
18:46 <t0m> or blank files with a weird name
18:46 <dhoss> okay, i think i understand
18:47 <t0m> or genrally any shit you're not specificially saying 'we test it generates a file named X'
18:47 <t0m> you test it generates X
18:47 <t0m> you test it does not generate anything other than X
18:47 <dhoss> right
18:47 <dhoss> oaky
18:47 <dhoss> that's what i figured
18:47 <t0m> sorry => tangent :)
18:48 <dhoss> it's okay, i'm a one track mind with this stuff :-)
18:48 <t0m> I can't find pmyi, but pym9 cool
18:49 <dhoss> Try rewriting script_create.pl and see how far you can get - the big issue is going to be help display
18:49 <dhoss> that's pmyi
18:49 <t0m> yes
18:49 <t0m> exactly so :)
18:49 <t0m> Oh, no, hang on
18:49 <t0m> Rewrite script_server.pl first
18:50 <t0m> in fact, rewrite everything that isn't _create first
18:50 <t0m> _create is the hardest
18:50 <t0m> and needs the most thinking
18:50 <t0m> by a fucking long shot
18:50 <t0m> I think _create probably involves MooseX::App::Cmd
18:51 <dhoss> so, rewrite with Moose sex getops
18:51 <t0m> do the simple ones
18:51 <t0m> first
18:51 <t0m> like myapp_server.pl
18:51 <t0m> like you just said :)
18:51 <t0m> which will get you up to scratch on the Moo sex
18:51 <t0m> before you have to hack it
18:52 <t0m> Also, we can then start the bikeshedding about the right thing to do with moving the real code back into a CPAN dist
18:52 <t0m> rather than it being in generated scripts
18:52 <dhoss> right, that would be awesome :-)
18:53 <t0m> myapp_XXXXX.pl should say: use Catalyst::Script::XXXXX; Catalyst::Script::XXXXX->new_with_options->run;
18:53 <t0m> + shebang line
18:53 <dhoss> right, i saw that in the list
18:53 <t0m> kk. So make the simple scripts MX::Getopt packages
18:54 <t0m> with the ->new_with_options->run incantation at the end of the source
18:54 <t0m> hit until myapp_server.pl works like it used to.
18:54 <t0m> rip Catalyst::Script::DevServer out
18:55 <t0m> job done
18:55 <dhoss> okay. i'll give it a crash
--- Log closed Wed Jun 17 19:00:38 2009
--- Log opened Wed Jun 17 19:00:44 2009
19:00 <t0m> http://scsys.co.uk:8001/29932. 
19:00 <t0m> I guess something like that but with less syntax errors works
19:01 <t0m> expand out till original functionality replicated :)
19:01 <dhoss> Moose<3
19:01 <dhoss> thanks man
19:04 <dhoss> i'm gonna get some grub. i'll ping you if i have any questions, i appreciate the help :-)
19:06 <t0m> no worries. I'm at 2am
19:06 <t0m> so => bed
19:08 <dhoss> hehe sleep up man
--- Log closed Wed Jun 17 19:13:38 2009
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
12:39 < t0m> to check we generate an application which can pass it's own tests 
             :)
12:40 < dhoss> so write a test to do taht?
12:40 < t0m> yy, that would be cool. Can just be on the bottom of the generated 
             files one
12:41 < t0m> run Makefile.PL, check exist status 0, check 'Makefile' exists, 
             run make test, check exit status zero
12:41 < dhoss> i trust something like `$perl Makefile.pl; make test` would do?
12:41 < t0m> *exit status
12:41 < dhoss> or is that windows unfriendly?
12:41 < dhoss> OH SHIT
12:41 < dhoss> thats right
12:41 < t0m> yeah, pretty much
12:41 < dhoss> i have to fix those /'s
12:41 < dhoss> make em windows happy
12:41 < t0m> yy, good one, File::Spec->catfile it up :)
12:42 < dhoss> i'm working on hacking buubot to read my mind
12:42 < dhoss> so i can just tell it to grab shit i say and todo it
12:42 < t0m> other than that, lets leave it till we can actually see a need for 
             more tests..
12:42 < dhoss> that's *fine* by me :-)
12:43 < dhoss> so i can branch and moosify, and hack on the create scripts to 
               make them moosey
12:43 < dhoss> also, what does MX::Getopt need to play nice?
12:43 < t0m> yy. I guess if you hack on the scripts, you want to write tests 
             that they still perl -c in the generated app
12:43 < t0m> yy. I guess if you hack on the scripts, you want to write tests 
             that they still perl -c in the generated app
12:43 < t0m> and attack the simple scripts like myapp_server.pl first
12:44 < t0m> And you'll work out what GetOpt struggles with :)
12:44 < dhoss> which one was the pain in the ass one? was that create?
12:44 < t0m> yy
12:44 < dhoss> ok, i got them backwards
12:44 < t0m> create has much more elaborate option parsin
12:44 < t0m> *parsing
12:44 < dhoss> okay. that will be a later beast
12:45 < t0m> so yeah, just branch and jfdi, add tests for things as you touch 
             them - as you touch scripts add a 'scripts still compile' test
12:46 < t0m> when each script is nicely getopt'd and refactored, it'll be easy 
             to test a lot more..