Repository navigation
Conversation
|
I think that part of this (getting x - y to work for structs) was addressed here: SciML/ModelingToolkit.jl#5372 |
|
Looks like it - I'll check against the new main and see. |
|
Hmm, no, this is still partly needed. The existing fixes cover the |
|
what do you mean by residualification? |
|
Whoops, sorry, what I mean by residualification is that right now it wants to rewrite x ~ y into 0 ~ x - y really badly, which is not defined for most structures. This avoids the formation of that residual form (which is why I call it that). |
eefd3f1 to
d1c9a7d
Compare
|
The problem with the rewrite being that |
|
I don't know what |
Struct-valued equations (e.g.
struct Foo x y end) beget equations of the formx ~ ywherexandyareFooInstances, which currently neither clock infer properly and error weirdly when MTK tries to put them into0 ~ x - yform. This instead keeps them intact and makes sure that clock inference recurses through them.