Need docs demonstrating how to replace 'goto LABEL'

[jkeenan]

As part of my research on Fatalization of currently deprecated goto
construct I have been
re-reading our current documentation of goto in blead and have been trying
to locate in our test suite examples of each claim made in that documentation.
What we find -- or fail to find -- there will have an impact on the
documentation updates provided as part of a pull request like the current
#23782. I've posted the preliminary results
of that documentation review
here.
This is the first in a series of GH issues addressing the items which I
believe are lacking in that documentation.

perldoc -f goto states that goto:

can be used to go almost anywhere else within the dynamic scope,
including out of subroutines, but it's usually better to use
some other construct such as "last" or "die".

Can this value judgment be illustrated via tests in the Perl 5 test suite?
(So far I have not been able to locate any.)

Such examples would be very helpful in repairing any damage to CPAN
libraries that occur as part of a fatalization of certain aspects of goto LABEL.

[mauke]

I'm not sure what exactly you're looking for here. Technically t/op/loopctl.t is full of them because every last/next/redo could be replaced by a goto. The reason they're "better" than goto is that goto is completely unrestrained, so it doesn't tell a reader much about what the code is trying to do. On the other hand, last tells a reader specifically that we want to exit from a surrounding loop, die indicates that we want to throw an exception and unwind the call stack, etc. So while you could replace every last/next/redo and die (and while/until and subroutine calls and return ...) with goto, using a more specific construct (in cases where one exists) better conveys your intent and tends to make the code more readable.

But how do you write a test for "tends to make code more readable"?

[jkeenan]

I'm not sure what exactly you're looking for here. Technically t/op/loopctl.t is full of them because every last/next/redo could be replaced by a goto. The reason they're "better" than goto is that goto is completely unrestrained, so it doesn't tell a reader much about what the code is trying to do. On the other hand, last tells a reader specifically that we want to exit from a surrounding loop, die indicates that we want to throw an exception and unwind the call stack, etc. So while you could replace every last/next/redo and die (and while/until and subroutine calls and return ...) with goto, using a more specific construct (in cases where one exists) better conveys your intent and tends to make the code more readable.

But how do you write a test for "tends to make code more readable"?

Thanks for your comments; they helped clarify my thinking. What we need more than tests is instructive documentation, i.e., examples in the docs that demonstrate why goto LABEL is inferior to those other control-flow operators. Subject line revised to reflect that.

In particular, we already know that once the fatalization planned in #23618 is implemented, there will be some breakage on CPAN (though not as much as I earlier feared). If we have examples of how to replace goto LABEL in place, people will be able to write patches and pull requests to fix that breakage even before perl-5.44 is released.