item(...) vs .item and $(...) issue

[2colours]

Raku/doc#4539

So this discussion happened on IRC a long time ago: https://irclogs.raku.org/perl6/2017-02-24.html#19:32 (Note the item(...) solution.)

Which was taken care as Raku/doc@b33402c, under the assumption that .item works like item(...) which is something you might assume based on the docs available.

In fact, $(...) calls .item and they return the container as it is with https://github.com/rakudo/rakudo/blob/c7004109c8535dcc926f5ed8c709155c5a88d2b3/src/core.c/Mu.rakumod#L816 (when there is one - overridden for container-ish types) . On the other hand, the subroutine item makes a copy a la https://github.com/rakudo/rakudo/blob/c7004109c8535dcc926f5ed8c709155c5a88d2b3/src/core.c/Any.rakumod#L555 (this is the default behavior of the multi at least; moreover it mangles Captures but that's another story).

Seems like the two have never been interchangable which really makes you wonder this is not a documentation issue at all, or it goes way beyond.

[2colours]

More on this topic in the upcoming release(?)

rakudo/rakudo@e34b767

rakudo/rakudo@a6608c2121

From what I can tell, the second commit is a partial revert of the changes introduced in the first commit. I haven't checked what it might change but I think in the bigger picture, it probably doesn't change the discrepancy between the duo of $(...) and .item and the item(...) subroutine call. In any case, I do think Liz is raising some good general points about the distinction of itemization and containerization.