search

ipc2023-classical / ipc2023-dataset

2 stars 1 forks source link

Conversion script adds incorrect PDDL requirements #1

Open jendrikseipp opened 4 months ago

jendrikseipp commented 4 months ago

Hi @danfis ! For all four domains that had normalized versions in the IPC, I get the following requirements, when I run the conversion Apptainer script:

(:requirements
    :strips
    :typing
    :negative-preconditions
    :disjunctive-preconditions
    :equality
    :existential-preconditions
    :universal-preconditions
    :conditional-effects
)

But isn't the whole point that the script removes some of these?

danfis commented 4 months ago

Sun, Apr 14, 2024 at 11:29:21PM CEST, @.***:

Hi @danfis ! For all four domains that had normalized versions in the IPC, I get the following requirements, when I run the conversion Apptainer script:

(:requirements
    :strips
    :typing
    :negative-preconditions
    :disjunctive-preconditions
    :equality
    :existential-preconditions
    :universal-preconditions
    :conditional-effects
)

But isn't the whole point that the script removes some of these? It seems the translator keeps the :requirements from the input pddl untouched. The actual features should be compiled away. Thanks for pointing this out, I'll put this on my todo list to fix it. I think I didn't notice because most translators just ignore requirements (for example Fast Downward), or are ok with having superset of features appearing in the pddl file (which is a correct behaviour).

For now, if you need accurate :requirements section, I would suggest using sed as postprocessing step (sed -i 's/:unwanted-feature//' domain.pddl).

Cheers, Dan

-- Reply to this email directly or view it on GitHub: https://github.com/ipc2023-classical/ipc2023-dataset/issues/1 You are receiving this because you were mentioned.

Message ID: @.***>

jendrikseipp commented 4 months ago

Thanks! Yes, that's the approach I used to fix this. Just wanted to mention it here, so you have it on your radar.