Merged.

This was actually not implemented in the normalization. Now it is. :-)

The translator paper indeed says that such existentially quantified variables in
effect conditions should be compiled to universally quantified variables of the
effect (last sentence of Section 4.2).

Can you confirm that you see this operator dump after normalization?
That would indicate that there's a bug in normalize.py.

I'm not sure I've seen existential conditions in universal effects before, so
maybe that is not properly implemented. (If so, that's a bug though, not an
intentionally missing feature.)

Existential conditions that happen in the precondition should be treated like
operator parameters; by analogy, I think it should be correct to treat
existential conditions within universal effects as additional quantified
variables for the universal effect? I'll check what the paper says about this
and what the normalization code does.

I get invalid plans on one of my compiled tasks (simplified version attached).

The problem seems to be somewhere in the instantiation, because for action
have-party it throws conditional effects that make someone_angry_in_world true away.

More precisely, in the effect

  forall ?w: object
    if
      ExistentialCondition ?p: object
        Conjunction
          Atom inworld(?p, ?w)
          NegatedAtom invited_in_world(?p, ?w)
    then
      Atom someone_angry_in_world(?w)

the translator thinks that the condition is for ?w=world1 impossible: When it
instantiates Atom inworld(?p, ?w), it checks whether "inworld(?p, world1)" is
either in fluent_facts or in the initial state. So it seems to be a problem with
the variable mapping for ?p which should somehow be handled in the instantiation
of the existential condition...? 

Shouldn't existential conditions not already have been compiled away at this stage?