Channel operations inside a janet_call could wreak more
havoc later, when scheduled fibers are shooting off when they shouldn't
be. The easiest way to prevent this is simply check that we are not
inside janet_call before doing channel operations. We also get nicer
error messages this way.
optionals.
This fixes some undesirable interplay between lookaheads and looping
(repeating) rules where lookaheads caused the loop to immediately
terminate _and_ discard any captures. This change adds an exception if
there is a 0-width match at the first iteration of the loop, in which
case captures will be kept, although the loop will still be terminated
(any further iteration of the loop would possibly cause an infinite
loop).
This allows optionals to work as expected when combined with lookahead.
There is also a question of whether or not optional should be
implemented as separate rule rather than as `(between 0 1 subrule)`.
Mutable keys are a minefield for comparisons, as resolving
equality require re-implementing a lot of the internal structures, as
well as dealing with multiple mutable keys that are in the same
equivalency class by deep=.
Simplifying the implementation to not resole mutable keys is much
simpler, faster, and has the benefit that deep= and deep-not= do not
need to allocate.
This was partially implemented before, but not in the case where the
await or other signal itself was created by a C function. We have to
separate code paths for generating signals - one via normal returns in
janet_vm_continue, and the other via longjump. This adds handling for
the longjump case, as well as improved messaging.
These rules allow selecting from a number of sub-captures
while dropping the rest. `nth` is more succinct in many cases, but `only-tags` is
more general and corresponds to an internal mechanism already present.
Weak containers did not preserve their weakness when marshalled. This
fixes that for tables and arrays, as well as adds some tests for this.
Also exposes functions for creating weak tables in janet.h
