Add new TensorCell class #325
Conversation
The new TensorCell class represents a single cell in a TensorMesh, with `center` and `bounds` properties. Add `__iter__` and __getitem__` methods to `TensorMesh`. The latter allows to obtain a single cell in the mesh based on their indices. The former allows to iterate over every cell in the mesh.
Use `numpy.unravel_index` to convert integer indices into tuples, following the right FORTRAN ordering. Simplify `__iter__` by building the iterator with a simple for loop over the whole range of cells, and reply on `__getitem__` for the unravelling of the indices. Add a new test that checks if the int indices return the expected cells.
Add a method to compare two TensorCell objects by their `h` and `origin` attributes.
Add support for slices while indexing cells in a TensorMesh. Add tests for the new feature.
Store the index of the cell in its parent TensorMesh.
This new method return a list of the neighboring cells in the mesh for the current cell. Add tests for the new method.
Sanitize inputs of TensorClass checking that elements of h and origin are numerical values, and elements of index are integers. TODO: test it
|
I'm trying to think if it would be good to just not allow a User to directly instantiate a |
|
It sounds a little too hacky... I doubt that if we don't provide examples that manually instantiate a |
Also, raise errors if any value of h, origin or index are bools (bools are considered to be Number and Integral).
|
Essentially it means you don't explicitly need to do all of the error checking for proper input. There's nothing wrong about creating the object on it's own, it's just that it does really have any tangible meaning without the Mesh class that it came from. |
Allow to get cells through negative single index (ravelled index), through negative indices per dimension and using negative indices in slices.
|
I agree that this new class doesn't introduce anything new if it's isolated from the mesh. That's why I think users won't be manually creating If I understand you correctly, you think it doesn't worth adding all the error raising on invalid inputs because users won't be creating this objects manually, so their proper creation should be covered by tests for the |
Yes, exactly. Unless we are concerned about ourselves programming this incorrectly to start with I don't think we necessarily need all the checks. |
|
Also, if the cell knows the shape of the mesh it came from, it can determine for itself the neighbor, node, face, etc. indices without passing it the mesh. |
Sounds good, that's something I can add. |
The `index` attribute now returns the raveled index. Add a new `index_unraveled` attribute that returns the unraveled indices. The `neighbors` now is a list of raveled indices of neighbouring cells.
Now the `bounds` and `center` properties return Numpy arrays, mimicking the behaviour of `TreeCell.center`.
Codecov Report
@@ Coverage Diff @@
## main #325 +/- ##
==========================================
+ Coverage 85.21% 85.39% +0.18%
==========================================
Files 38 39 +1
Lines 11578 11763 +185
==========================================
+ Hits 9866 10045 +179
- Misses 1712 1718 +6
|
|
@jcapriot I think this is ready to be reviewed. Is there anything else we would like to add to the |
|
For some reason, Windows machines didn't like to check if a variable is an integer with np.issubdtype(type(variable), int)But using np.issubdtype(type(variable), np.integer) |
|
That makes sense, |
There was a problem hiding this comment.
Looks good, only thing left to do is to actually add it to the documentation pages.
We will likely need to re-name the Tree Mesh Cells section in __init__.py to just be Mesh Cells and update the text of that section accordingly (and add the TensorCell) to that list.
Also can you add a simple example to the docs of the TensorCell ?
The new
TensorCellclass represents a single cell in a TensorMesh, withcenter,bounds,neighbors,nodes,edgesandfacesproperties. Add__iter__and__getitem__methods toTensorMesh. The latter allows to obtain a single cell in the mesh based on their indices. The former allows to iterate over every cell in the mesh. Add tests for the new class and methods.