Fix Problem's timed effects in Protobuf Writer

[Framba-Luca]

Fix #536

[arbimo]

🤔 I am surprised this bug was not encountered earlier as protobuf testing is done by a round-trip conversion of all test cases in the UP, checking that the result is the same as the original.
So it may be worth checking that we have an example with timed effects in the test suite and if not adding one.

[arbimo]

I just expanded a bit the example to better test the compliance of validators.

It seems however that we now hit a limitation of the PDDL writer that does not support TILs...

[Framba-Luca]

Thanks for expanding. I did a quick fix to mypy, so now we see the PDDL test failing. I think I just have to add a skip if there are TILs to the problem in the PDDL testing file

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (830f340) 84.91% compared to head (6ad31db) 84.95%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #538      +/-   ##
==========================================
+ Coverage   84.91%   84.95%   +0.04%     
==========================================
  Files         200      200              
  Lines       26403    26428      +25     
==========================================
+ Hits        22420    22453      +33     
+ Misses       3983     3975       -8     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[alvalentini]

@arbimo I'm going to add TILs support in the PDDLWriter in a different PR.

[arbimo]

One last thing. The test cases trigger a corner case of the temporal planning semantics: when should the goals hold in the presence of timed initial literals ? The PDDL 2.2 semantics require the goals to hold after all TILs are computed (the real final state).
Another possible interpretation would be "immediately after the end of the last action".

Looking at report it seems that:

• tamer and aries-val adheres to the PDDL 2.2 semantics
• aries and tamer-val adheres to the second one

I believe we should enforce the PDDL2.2 semantics as they are the least surprising ones (and most established). This is the one enforced in the test-cases added in this PR.

Note that this is truly a corner case as it is very unusual to have goal conditions on exogeneous state variables in real problems.

@mikand Feel free to jump in on this, otherwise I will merge it as it is (which means that down the line, aries and tamer-val will need to align themselves on the PDDL semantics).

[alvalentini]

@arbimo I agree with you and the PDDL 2.2 semantics. I'm going to fix the tamer validator.

[arbimo]

@alvalentini Perfect, we agree. I'll merge this one then (and add this aries' fix to my todos)

Thanks a lot to both of you!