On Wed, Aug 14, 2013 at 11:26 AM, James Laver <james.la...@gmail.com> wrote: > perlglossary is wrong. It doesn't only occur as an lvalue but can > occur by access as well.
Yes and no. autovivification in general applies in writish kinds of context, but that is a somewhat imprecise concept. Note that keys() can be in lvalue context (allowing you to set the hash bucket size); it just so happens that keys always puts a hash dereference argument in lvalue context whether or not keys itself is; perlglossary is quite correct in terms of the context applied to the dereference itself.