ModelicaSpecification
ModelicaSpecification copied to clipboard
modelica
Add argmin and argmax array functions returning indices
Reported by beutlich on 3 May 2017 12:54 UTC Section 10.3.4 declares min and max array functions. However, the indices satisfying the min or max are not returned. In MATLAB the min and max reduction functions have a second output argument returning these indices.
See also #1980 for reshape.
Migrated-From: https://trac.modelica.org/Modelica/ticket/2187
Comment by hansolsson on 10 May 2017 14:30 UTC I agree that it seems like useful functionality, but exactly how?
- Is it the first, last, or any element being equal to max/min?
- For a matrix (or higher dimensional entity) do we return one argument per array subscript, or one vector of subscripts?
- Only for the array functions, right?
Comment by jmattsson on 29 May 2017 08:06 UTC Replying to [comment:1 Hans Olsson]:
Only for the array functions, right?
Preferably for the reduction expression variants, as well. Otherwise you get:
(x, i) = min(y[j] * j for j in 1:n); // This is an error
(x, i) = min({y[j] * j for j in 1:n}); // this is allowed
It's bad enough that sum() behaves different with an array argument and as a reduction expression without adding more.
Comment by hansolsson on 16 Jun 2017 14:45 UTC As noted in #2190 if they can be used in continuous-time equations the additional output need to generate events. (Another option is to restrict the use of argmin/argmax.)
An argument for restricting the use is to avoid the numerical issues if multiple array elements are at the maximum/minimum (which does not seem unlikely).