|
Created on 2025-05-12.15:52:37 by haz, last changed by haz.
This is akin to --keep-unimportant-variables, but for the operaters/actions.
We'd like the command line to have a translator option to --keep-noop-operators
so that they
aren't discarded. This touches on several places where an operator or action
(depending on where
in the process you are) is removed if the effects are empty. Default should be
false, but this
would help those of us using the FD translator for other paradigms (FOND, PPOS,
PPDDL, etc).
|
File name |
Uploaded |
Type |
Edit |
Remove |
domain.pddl
|
haz,
2025-06-24.18:04:46
|
application/octet-stream |
|
|
msg11850 (view) |
Author: haz |
Date: 2025-06-25.15:21:46 |
|
Thanks for having a look into it! I'd feel more comfortable having the full test
to check for regressions with the code & corner cases, so no worries.
|
msg11849 (view) |
Author: malte |
Date: 2025-06-25.12:02:42 |
|
Thanks, looks good! I left one last polishing comment.
For me it looks ready to merge, but I would still like to run an experiment first, which is taking me a bit of time to set up because I haven't done it in a while and there have been some significant changes in our infrastructure. I hope I can finish the setup and queue the experiment later today.
|
msg11848 (view) |
Author: haz |
Date: 2025-06-24.18:04:46 |
|
Thanks! Changes make sense -- revisions are now all set. I'm attaching a slightly
changed gripper domain (you can pair it with
misc/tests/benchmarks/gripper/prob01.pddl) that has a noop operator for testing.
|
msg11847 (view) |
Author: malte |
Date: 2025-06-24.14:59:42 |
|
Hi Christian, thanks for the pull request and sorry for the slow reaction.
I'm now off the busy stretch and expect my next reaction should come much more quickly.
I'd like to request a change in the style of the code -- rather than passing the new option around a lot, access it from the options module where needed. That, plus one or two smaller things. See my review of the pull request for more details and some rationale. This should also make the pull request much smaller and touch fewer files.
Let me know if you are OK to make these changes. Otherwise I can also try to make them myself.
Can you attach or link one or two benchmarks that trigger the new changes, so that we can test this?
I don't think it's very important for this specific change, but as a matter of best practice I'd also like to run a larger before-after translator experiment before we merge. I'll see if I can set something up.
|
msg11846 (view) |
Author: malte |
Date: 2025-06-24.14:27:51 |
|
pull request: https://github.com/aibasel/downward/pull/263
|
|
Date |
User |
Action |
Args |
2025-06-25 15:21:46 | haz | set | messages:
+ msg11850 |
2025-06-25 12:02:42 | malte | set | messages:
+ msg11849 |
2025-06-24 18:04:46 | haz | set | files:
+ domain.pddl messages:
+ msg11848 summary: This is akin to --keep-unimportant-variables, but for the operaters/actions.
We'd like the command line to have a translator option to --keep-noop-operators so that they
aren't discarded. This touches on several places where an operator or action (depending on where
in the process you are) is removed if the effects are empty. Default should be false, but this
would help those of us using the FD translator for other paradigms (FOND, PPOS, PPDDL, etc). -> This is akin to --keep-unimportant-variables, but for the operaters/actions.
We'd like the command line to have a translator option to --keep-noop-operators
so that they
aren't discarded. This touches on several places where an operator or action
(depending on where
in the process you are) is removed if the effects are empty. Default should be
false, but this
would help those of us using the FD translator for other paradigms (FOND, PPOS,
PPDDL, etc). |
2025-06-24 14:59:42 | malte | set | messages:
+ msg11847 |
2025-06-24 14:27:51 | malte | set | status: unread -> chatting assignedto: malte messages:
+ msg11846 |
2025-05-12 18:39:14 | malte | set | nosy:
+ malte, haz, jendrik |
2025-05-12 15:52:37 | haz | create | |
|