Support heterogeneous parameters for linearize and remake #2283
Conversation
Co-authored-by: Fredrik Bagge Carlson <[email protected]>
| @@ -1274,14 +1275,34 @@ See also [`linearize`](@ref) which provides a higher-level interface. | |||
| function linearization_function(sys::AbstractSystem, inputs, | |||
| outputs; simplify = false, | |||
| initialize = true, | |||
| op = Dict(), | |||
There was a problem hiding this comment.
This is a very bad interface, now the user has to provide both p and op even if op already contains p. The new changes are also lacking documentation, how is the user supposed to know that they now have to provide also p for the results to be correct?
I would prefer to think a bit harder about this design, the current state feels like a quick hack that is going to be very confusing to use
There was a problem hiding this comment.
The alternative is being wrong though.
There was a problem hiding this comment.
The op here doesn't require the correct numerical value. It only cares about types.
There was a problem hiding this comment.
The op here doesn't require the correct numerical value. It only cares about types.
What does this mean? op is the user-provided operating point around which to linearize.
There was a problem hiding this comment.
Since we use heterogenous parameters by default now, we need to know the exact type of the parameters to create the linearized function correctly.
YingboMa
changed the title
YingboMa
force-pushed
the
bgc/split_params_bug
branch
from
360ddb3 to
2a553a7
Compare
Requires SciML/ModelingToolkitStandardLibrary.jl#225