Change magnet modeling workflow to use CadQuery #139
Conversation
parastell/magnet_coils.py
Outdated
| parallel_parts = [] | ||
| for dir, tangent in zip(normal_dirs, self.tangents): | ||
| parallel_parts.append(np.dot(dir, tangent)) | ||
| parallel_parts = np.array(parallel_parts) |
There was a problem hiding this comment.
If these are numpy arrays then this block becomes
| parallel_parts = [] | |
| for dir, tangent in zip(normal_dirs, self.tangents): | |
| parallel_parts.append(np.dot(dir, tangent)) | |
| parallel_parts = np.array(parallel_parts) | |
| parallel_parts = np.dot(normal_dirs, self.tangents) |
Or close (maybe need to specify an axis?)
There was a problem hiding this comment.
As mentioned above, these are all numpy arrays. Unfortunately, the numpy functions that perform the dot product, dot and inner, do not have a way of specifying an axis along which to operate. Without a specified axis, they will just perform a full matrix operation.
There was a problem hiding this comment.
Can we structure the matrices such that a full matrix operation is the right thing to do?
connoramoreno
force-pushed
the
magnets-cq
branch
from
f1e019b to
13bf0db
Compare
parastell/magnet_coils.py
Outdated
| parallel_parts = [] | ||
| for dir, tangent in zip(normal_dirs, self.tangents): | ||
| parallel_parts.append(np.dot(dir, tangent)) | ||
| parallel_parts = np.array(parallel_parts) |
There was a problem hiding this comment.
Isn't this the same as
| parallel_parts = [] | |
| for dir, tangent in zip(normal_dirs, self.tangents): | |
| parallel_parts.append(np.dot(dir, tangent)) | |
| parallel_parts = np.array(parallel_parts) | |
| parallel_parts = np.matmul(normal_dirs, self.tangents.transpose()) |
There was a problem hiding this comment.
Almost. We'd then need to take the diagonal of the resultant matrix (parallel_parts is supposed to be an array of scalars and np.matmul(normal_dirs, self.tangents.transpose()) results in an NxN matrix where N is the number of points on the filament), but that's easy enough to do.
There was a problem hiding this comment.
Maybe I described the transpose backwards then? It should definitely be possible to get a vector of N scalars from multiplying a 3xN matrix times an Nx3 matrix, right?
connoramoreno
force-pushed
the
magnets-cq
branch
from
9cb3464 to
3fbecaf
Compare
| parallel_parts = np.diagonal( | ||
| np.matmul(outward_dirs, self.tangents.transpose()) | ||
| ) |
There was a problem hiding this comment.
See my comment elsewhere... would this get you a vector if you swapped the transpose?
| parallel_parts = np.diagonal( | |
| np.matmul(outward_dirs, self.tangents.transpose()) | |
| ) | |
| parallel_parts = p.matmul(outward_dirs.transpose(), self.tangents) |
They are both 3 x N matrices, right? so there must be some way to arrange them such that a matrix multiplication results in a 1 x N vector without needing to take the diagonal...
There was a problem hiding this comment.
Perhaps I'm not seeing the same thing as you but I don't believe there's a way for a matrix product to be smaller in a dimension than the smallest dimension of the original matrices. I.e., for two Nx3 matrices, we can produce a 3x3 matrix (first matrix transposed) or a NxN matrix (second matrix transposed). That said, there may be some NumPy operation capable of producing what we want but I can't seem to find one...
There was a problem hiding this comment.
I really am getting old.... and would probably fail my qual... I'll stop questioning these things when they work...
Models the magnets using CadQuery instead of Coreform Cubit. Furthermore, restricts magnet cross-section to rectangular, removing the option for circular magnets.
Closes #138