Bug #84020 for Method-Signatures: Make Moose optional

> As far as I know Moose is only needed for type checking. Would it be > possible to factor this out, so that Method::Signatures can be used

> it (and thus with fewer dependencies and a smaller foot-print)?

> <URL: https://rt.cpan.org/Ticket/Display.html?id=84020 > > > > Moose is never loaded by Method::Signatures. It will load _Mouse_, but > only if you use types in your signatures. So Method::Signatures can > easily be used without Moose, and even without Mouse (if you don't do any > type-checking at all).

> According to my perlbloat script, MS adds about 18.2Mb (on my machine with > my version of Perl). Of that, I'm guessing the bulk of it is from PPI > (which perlbloat tells me weighs in at 13.8Mb). Using PPI is how we avoid > many whitespace-dependent bugs, make sure that comments in the signatures > don't cause weird bugs, and properly parse things like code blocks. I > think it would be prohibitively expensive to try to do without it, but I'm > open to ideas.

> In fact, if you have any specific ideas for reducing the memory > footprint of MS, please let us know. We're always open to doing that. :-)

> The only idea I have (and that would probably work in my context) is to have > an option to disable those checks. This would allow to have the checks > during development and testing, but have a smaller foot print on the > production system. But this might be difficult to implement and probably be > a bit against the spirit of the module ...

> Integrate it into the Perl core ... ;-)

> <URL: https://rt.cpan.org/Ticket/Display.html?id=84020 > > > On Thu, Mar 21, 2013 at 1:16 AM, Fritz Zaucker via RT > <bug-Method-Signatures@rt.cpan.org> wrote: > > I know Buddy and I were talking about this, but I can't find it in a > Github issue. We talked about having an environment variable to turn > off types, but had to be careful of coercion.

> Buddy, do you remember? > > > I wish. And it has been floated with some core folks.

> The only idea I have (and that would probably work in my context) is to have > an option to disable those checks. This would allow to have the checks > during development and testing, but have a smaller foot print on the > production system. But this might be difficult to implement and probably be > a bit against the spirit of the module ...

> I was actually talking about the checks using PPI, as this seems to be > the > biggest module being used in MS.

> <URL: https://rt.cpan.org/Ticket/Display.html?id=84020 > > > Fritz, > > > As schwern says, I think I wasn't clear enough in my first email. We > don't use PPI for any validation checks. We use it to parse the signature > in the first place. Without PPI, we can't generate the code to unpack @_, > much less generate the validation checks. Now, schwern mentions that it > might conceivably be possible to parse the sig without PPI, and that's > true, but for just one simple example of how valuable PPI is: > > One of the very first bugs I ever discovered in MS was that comments > inside signatures (e.g. if your signature spanned multiple lines) would > throw off the parsing. I contacted schwern about it, and he said "well, > we're using PPI; just use that to fix it." One quick head-slap later, I > spent ten minutes reading the PPI docs, added a *single* line of code, and > the bug was fixed. > > That's the kind of flexibility I would really hate to give up by > abandoning PPI.