Read in nonuniform grid #1115
Read in nonuniform grid #1115
Conversation
| @@ -332,12 +332,12 @@ namespace aspect | |||
| /** | |||
| * Interpolation functions to access the data. | |||
| */ | |||
| std::vector<Functions::InterpolatedUniformGridData<dim> *> data; | |||
| std::vector<Functions::InterpolatedTensorProductGridData<dim> *> data; | |||
There was a problem hiding this comment.
Can you instead just make it a std::vector<Function<dim>*>, for reasons that will become clear below?
There was a problem hiding this comment.
(Note: Function<dim> is a base class for both the uniform and tensor product classes.)
anne-glerum
force-pushed
the
uniform_to_nonuniform_asciidata
branch
from
95fad9e
to
5533e5d
Compare
|
/run-tests |
| coordinate_values[i].push_back(temp_coord); | ||
|
|
||
| // The grid spacing | ||
| double first_grid_spacing = 0, grid_spacing = 0; |
There was a problem hiding this comment.
We usually split these declarations into two lines, and I do not think you need to initialize these variables (this is my mandatory comment, which proofs that I have looked through the code 😄).
|
Very nice and useful addition. I assume you have tested the result? Could you maybe take one of the ascii_data tests and modify them to use a non-uniform grid? One test and one changed file should be enough. |
|
@gassmoeller I've incorporated your comments and added a test. I've tested the result by changing existing test data files, now I'll just do one more test with a data file I've generated. I'll let you know :) |
anne-glerum
force-pushed
the
uniform_to_nonuniform_asciidata
branch
from
e585465
to
87d9505
Compare
|
@gassmoeller I've changed the tolerance used for comparing the grid spacing and added output when the data is found to be nonuniform. |
| std::numeric_limits<double>::epsilon()*std::abs(grid_spacing+first_grid_spacing)*2) | ||
| // Compare current grid spacing with first grid spacing, | ||
| // taking into account roundoff of the read-in coordinates | ||
| if (grid_spacing / first_grid_spacing < 0.995 || grid_spacing / first_grid_spacing > 1.005) |
There was a problem hiding this comment.
I think the previous version of this check was more intuitive. Is your current one not equivalent to replacing std::numeric_limits<double>::epsilon() by 0.005? I would be favouring that if you need a larger tolerance.
|
Thanks! Apart from my one stylistic comment this is ready to go. Care to add an entry in changes.h about the new possibility? |
Remove unnecessary functions and add assert Let file format dictate uniform or nonuniform GridData Improve comments Indent
Add test Use a different magic number Add output and increase tolerance Update test and reduce refinement Astyle
anne-glerum
force-pushed
the
uniform_to_nonuniform_asciidata
branch
from
87d9505
to
a80efdf
Compare
This way we can also read in data that is not equally spaced in a direction, for example crust1.0.