[META] Surprising Design Decisions

[leonerd]

This meta-issue maintains a list of comments on what could be considered "surprising" design decisions. I.e. parts of the current implementation that intentionally differ from existing structure or what might be expected to be their design, and attempts to discuss why.

[leonerd]

ADJUST phasers are not like BEGIN/INIT/CHECK/...

I found the way that the existing perl parser handles the existing phaser blocks (BEGIN, INIT, etc...) to be very strange. It works by pretending that there was a prefixed sub keyword in front, then parses the whole thing as a regularly named sub. Once finished, late in the code flow these special symbol names are detected, and the newly-constructed sub is installed in one of the phaser AVs instead of the regular symbol table. This has a bunch of weird properties to it; e.g. that the sub keyword can still be optionally written, or that you cannot actually have a regular named sub of these names.

Instead, the ADJUST phasers are implemented by different parser syntax that parses them as a different kind of true block; similar to the way e.g. defer blocks work.

[leonerd]

Class attributes are not like other perl attributes

Attributes applied to class declarations are implemented in a fundamentally different way to existing attributes on subroutines or variables. #24 has more details.