Rcc4#187
Conversation
|
@pet1330 Test this plz when you have time, so I can merge it and rebase rcc5 and solve its conflicts in one go. |
|
PS: checks will fail, but it should be the multiple only failing. |
|
Why have you left it out of the docs? |
|
RCC3 is actually wrong in its current implementation as it does not account for inverse relations. When I used it, it was fine for the purposes I wanted it as inverse did not caused me troubles. RCC4 is the same as RCC3 with inverse included and unified vocabulary across the RCC relations. My suggestion would be to migrate to RCC4 from RCC3 when you have time. In no rush to deprecate it. |
|
I don't think it's wrong, I think what you're describing is a different relation. RCC3 should have [O, PO, DC] That makes it a different relation. All it will mean is that I end up using RCC4 to simply combine the two PP states into the same relation. You may as well leave RCC3 as as it is and just add RCC4 with this additional functionality. |
|
Regarding the unconventional strings of RCC3 (my fault I know) it should be [PP, PO, DC] I think. Mapping from RCC8 to RCC3 would be:
Is that what you mean and happy with it? |
|
Due to the way quantisation works, it needs to remain what it would have mapped to, otherwise we will get different results for each RCC when using the quantisation factor |
|
Not sure what you mean with respect to the q-factor, it is simply a mapping from one to the other (can be done via a dict), so q-factor is kind of irrelevant to this mapping. The current implementation maps dc to dc, ec to po, and all the rest are as o. In the o case it does not distinguish between RCC8's [po, tpp, ntpp and the inverses]. Hmmm, I think I see what you mean now, you don't want a distinction between partially overlaps and proper part of (tangential or not), but have them as occludes or better overlaps. That is fine for me but I think that in that case EC should remain as EC as transforming it to PO (partial overlap) is confusing. This would result in the following RCC3 relations [DC, EC, O] with mapping everything except DC and EC to O. Right? If that's the case fine by me and it makes it different from RCC4 also. |
|
Check out #195, it shows you at the bottom all of the relations. When you adjust the quantisation factor, check that they are mapped correctly 😄 |
|
See my other comment, which basically I think that debugging RCC PRs that map from RCC8 is as simple as checking the mapping dictionary. So for RCC5 you can do this. @pet1330 HOWEVER, for this PR since I had to put back RCC3, and since you are using it in your current work, you might want to take it for a spin to make sure that things still work for you. We can talk and do about minor changes at a later stage. |
|
Tested. This is okay to merge. 👍 |
|
@yianni resolve conflicts |
Conflicts: qsr_lib/CMakeLists.txt qsr_lib/src/qsrlib_qsrs/__init__.py
2 participants
No description provided.