Re^4: Main routines, unit tests, and sugar

[BrowserUk]

Any important code resides under one App::MyApp namespace, and not split across App::MyApp and myapp.pl

It was mostly humour, but ... you still need myapp.pl in order to invoke your application.

Which means you added an extra file (your App::MyAPP.pm) -- and an extra level of callback trace to every error message, etc. etc. -- and made the (still) required file almost empty and what did you gain?

IMO nothing. You added the need to chase off and track through where this module file lives -- not always obvious with lib, site/lib, site/vendor/lib, use lib et al. -- just in order to see what it does; what its dependencies are ...

That almost empty file, the .pl, is still used. What is the point in adding an extra file in order to make this one almost empty? There is no logical justification for it.

[vsespb]

Well, if the application is small enought to fit to one file, your arguments are valid

If application already consists of several files (i.e. it's not a small 500 lines script), the benefits I described outweigh that fact that you need another file

You added the need to chase off and track through where this module file lives -- not always obvious with lib, site/lib, site/vendor/lib

It must be irrelevant if we talk about CPAN deploy. More that that, if user has a mess with @INC dir, it's actually better to have all application files in one @INC dir, than having half of application is @INC and half in bin (/usr/bin, /usr/local/bin/ ~/bin /opt/myapp/bin etc) (if we assume that there is a mess with @INC, we can assume there is a mess with $PATH too)

Having site/lib, site/vendor/lib is actually a feature, so people can deploy with CPAN or with OS package manager at same time, it's an advantage

extra file

Minor issue, if you already have several files

an extra level of callback trace

There won't be extra level of stacktrace if you just split code to files. Extra level added if you split it to subroutines

So it's valid argument if we talk about having some startup code in one "main" subroutine vs having some code unwrapped to subroutine in main package

callback trace to every error message

Imho stacktrace should appear if unexpected/programmer error happened (assertion, like "confess"), otherwise there should be clear error message and no stacktrace is needed. But it's probably matter of script quality

[BrowserUk]

the benefits I described

You had not -- and still have not -- described any "benefits".

At best you've outlined your personal preference for adding an extra layer to your applications for no good reason.

But they are your applications, so do as you will. (Just don't try to feign some CompSci justifiction for it. )