Issue 753: ugly hacks - Fast Downward issue tracker

Title	ugly hacks
Priority	meta	Status	chatting
Superseder		Nosy List	florian, guillem, jendrik, malte, silvan
Assigned To		Keywords
Optional summary	1. The "pref=..." option of the lmcount heuristic (see http://www.fast-downward.org/OptionCaveats). ~~2. The "cost_type" and "lm_cost_type" options in the landmark code (issue697 and see http://www.fast-downward.org/OptionCaveats).~~ ~~3. The multi-path dependency code (issue198).~~ 4. ~~The global copy of~~ the initial state data and the complex way of initializing it (see msg5101 and issue509).

Created on 2017-11-30.17:37:10 by malte, last changed by florian.

1. The "pref=..." option of the lmcount heuristic (see
http://www.fast-downward.org/OptionCaveats).
2. The "cost_type" and "lm_cost_type" options in the landmark code (issue697 
and see
http://www.fast-downward.org/OptionCaveats).
3. The multi-path dependency code (issue198).
4. The global copy of the initial state data and the complex way of 
initializing
it (see msg5101 and issue509).

Messages
msg7770 (view)	Author: florian	Date: 2018-09-22.13:53:04
> I don't know if no. 4 is addressed with Florian's recent changes. Only parts of it is done. The data is no longer global and its less complicated than it was but it is still complicated and it still has a "HACK" comment.
msg7704 (view)	Author: malte	Date: 2018-09-21.11:37:17
Numbers 2 and 3 have been addressed. Despite our usual policy to let the struck-through issue numbers speak in the summaries to show that something was done, let me explicitly strike through 2. and 3. here, because in this case the related issues mentioned (at least for 2.) were not so much about fixing the particular hack and more to provide general context to what the hack is. (In particular, issue697 was closed long before the hack was actually resolved, so using the issue-strikethrough as a sign that the task was done would not have worked in this case.) I don't know if no. 4 is addressed with Florian's recent changes.
msg6669 (view)	Author: florian	Date: 2017-12-01.08:27:21
The initial state data is stored globally without axioms evaluated and is accessed when constructing state registries in a very intricate way where the task constructs a StateRegistry that accesses the task itself but gets the state data on a separate channel to avoid an endless loop (see msg5101 and issue509).
msg6664 (view)	Author: jendrik	Date: 2017-11-30.18:55:29
Reference issue697.
msg6661 (view)	Author: malte	Date: 2017-11-30.17:37:24
Fix typo in summary.
msg6660 (view)	Author: malte	Date: 2017-11-30.17:37:10
This meta issue collects ugly hacks that we would like to get rid of. It is meant mainly for things where the usual concepts in the code are violated in some way and hence work-arounds are required, especially when these are visible on the user level. For example, usually a heuristic object doesn't care whether it's going to be used for preferred operators, but if it is the landmark heuristic, it does care, and the user needs to specify a special option. Also, the way that action cost modifications are specified usually makes some sense, but it doesn't in the case of the landmark heuristic and/or landmark synergy.

History
Date	User	Action	Args
2018-09-22 13:53:04	florian	set	messages: + msg7770 summary: 1. The "pref=..." option of the lmcount heuristic (see http://www.fast-downward.org/OptionCaveats). ~~2. The "cost_type" and "lm_cost_type" options in the landmark code (issue697 and see http://www.fast-downward.org/OptionCaveats).~~ ~~3. The multi-path dependency code (issue198).~~ 4. The global copy of the initial state data and the complex way of initializing it (see msg5101 and issue509). -> 1. The "pref=..." option of the lmcount heuristic (see http://www.fast-downward.org/OptionCaveats). ~~2. The "cost_type" and "lm_cost_type" options in the landmark code (issue697 and see http://www.fast-downward.org/OptionCaveats).~~ ~~3. The multi-path dependency code (issue198).~~ 4. ~~The global copy of~~ the initial state data and the complex way of initializing it (see msg5101 and issue509).
2018-09-21 11:37:17	malte	set	messages: + msg7704 summary: 1. The "pref=..." option of the lmcount heuristic (see http://www.fast-downward.org/OptionCaveats). 2. The "cost_type" and "lm_cost_type" options in the landmark code (issue697 and see http://www.fast-downward.org/OptionCaveats). 3. The multi-path dependency code (issue198). 4. The global copy of the initial state data and the complex way of initializing it (see msg5101 and issue509). -> 1. The "pref=..." option of the lmcount heuristic (see http://www.fast-downward.org/OptionCaveats). ~~2. The "cost_type" and "lm_cost_type" options in the landmark code (issue697 and see http://www.fast-downward.org/OptionCaveats).~~ ~~3. The multi-path dependency code (issue198).~~ 4. The global copy of the initial state data and the complex way of initializing it (see msg5101 and issue509).
2017-12-01 15:50:58	silvan	set	nosy: + silvan
2017-12-01 11:47:08	guillem	set	nosy: + guillem summary: 1. The "pref=..." option of the lmcount heuristic (see http://www.fast-downward.org/OptionCaveats). 2. The "cost_type" and "lm_cost_type" options in the landmark code (issue697 and see http://www.fast-downward.org/OptionCaveats). 3. The multi-path dependency code (issue198). 4. The global copy of the initial state data and the complex way of initializing it (see msg5101 and issue509). -> 1. The "pref=..." option of the lmcount heuristic (see http://www.fast-downward.org/OptionCaveats). 2. The "cost_type" and "lm_cost_type" options in the landmark code (issue697 and see http://www.fast-downward.org/OptionCaveats). 3. The multi-path dependency code (issue198). 4. The global copy of the initial state data and the complex way of initializing it (see msg5101 and issue509).
2017-12-01 08:27:21	florian	set	nosy: + florian messages: + msg6669 summary: 1. The "pref=..." option of the lmcount heuristic (see http://www.fast-downward.org/OptionCaveats). 2. The "cost_type" and "lm_cost_type" options in the landmark code (issue697 and see http://www.fast-downward.org/OptionCaveats). 3. The multi-path dependency code (issue198). -> 1. The "pref=..." option of the lmcount heuristic (see http://www.fast-downward.org/OptionCaveats). 2. The "cost_type" and "lm_cost_type" options in the landmark code (issue697 and see http://www.fast-downward.org/OptionCaveats). 3. The multi-path dependency code (issue198). 4. The global copy of the initial state data and the complex way of initializing it (see msg5101 and issue509).
2017-11-30 18:55:29	jendrik	set	messages: + msg6664 summary: 1. The "pref=..." option of the lmcount heuristic (see http://www.fast-downward.org/OptionCaveats). 2. The "cost_type" and "lm_cost_type" options in the landmark code (see http://www.fast-downward.org/OptionCaveats). 3. The multi-path dependency code (issue198). -> 1. The "pref=..." option of the lmcount heuristic (see http://www.fast-downward.org/OptionCaveats). 2. The "cost_type" and "lm_cost_type" options in the landmark code (issue697 and see http://www.fast-downward.org/OptionCaveats). 3. The multi-path dependency code (issue198).
2017-11-30 17:41:14	jendrik	set	nosy: + jendrik
2017-11-30 17:37:24	malte	set	messages: + msg6661 summary: 1. The "pref=..." option of the lmcount heuristic (see http://www.fast-downward.org/OptionCaveats). 2. The "cost_type" and "lm_cost_type" options in the landmark code (eee http://www.fast-downward.org/OptionCaveats). 3. The multi-path dependency code (issue198). -> 1. The "pref=..." option of the lmcount heuristic (see http://www.fast-downward.org/OptionCaveats). 2. The "cost_type" and "lm_cost_type" options in the landmark code (see http://www.fast-downward.org/OptionCaveats). 3. The multi-path dependency code (issue198).
2017-11-30 17:37:10	malte	create

Issue753