uob-positron-imaging-centre / up4 Goto Github PK

I successfully ran up4 on windows using the rust-standalone version.
One needs to install HDF5 (link) on windows and make sure that the bin folder is in PATH (which it should be automatically).

The main function in src/main.rs runs flawlessly and also plots the rotating drum (after some tweaks).

installing it in python works, however for some reason the HDF5-rust crate is not able to write a file.

pyo3_runtime.PanicException: Unable to create hdf5 File: Invalid file access property list id: 792633534417207322

@DanW97 can you try to reproduce this error? if so, i will raise an issue on HDF5-rust as I can not find anything online

Hi, trying to use up4 to visualise velocity fields, noticed the tutorial code doesn't work, even with the test file provided. I think one of the commands has been changed from vector_field to vectorfield, and an index is required for the command vec_field_plotter.unit_vector_plot, but when i make those two changes, I still get an error saying argument 'axis': 'builtins.VectorGrid' object cannot be interpreted as an integer.

I am thinking of a way to compare datasets easier.
A function that allows the qualitative comparison between any function implemented in uPPPP with any dataset.
Example:

import uppp
simulation = uppp.Data("simulation.hdf5")
experiment = uppp.Data("experiment.hdf5")
experiment.compare(function = uppp.velocityfield, dataset=simulation, kwargs=...)

Return type/plot of this would be a cell by cell coomparison of a field (1D, 2D, or 3D) like this plot:

where both axis show the value of the individual cell (velocity for example) and a straight line would mean that the cells are the same.

Problem would be that experiments normally do not overlap spatially with the simulation. hence we would need to translate the whole dataset and we would need to automatically do that.

1. Implement global, min-max and elementwise scaling of u and v data
2. Remove hardcoded axis formatting
3. Add option/enum structs for titles, axis labels and ranges
4. Investigate colourbar addition
5. Enable square plots

the ndarray crate uses the Dim type(?) (or it's an enum) and we also do, which leads to some very confusing type/syntax hinting from rust-analyzer. It would be more ergonomic to rename the enum to provide a clearer rust API for the user

We found that the HSM_Glass_2l_250.hdf5 file lacked the required "Dimension" attribute which broke a lot of the testing aspects of the vtk converter. Raising this issue so that we can keep aware of this in case there is an underlying problem to hdf5 file generation.

The attached files show a velocity vector and a velocity scalar field in the same plane at the same location. As seen by the heatmap, the velocities are flipped about the centre of the image. This is true for all data in any axis view.

Old version of uPPPP (pyAnalyser) does not define behaviour for NaN -positions/velocities given from experimental data such as PEPT. Result is a invalid field for example here:

There are valid positions in the sections without data, however a single NaN makes the field invalid.
We must check for NaNs when implementing the old functions in the new version

The latest update of the software has had an issue where the scalar fields have become offset. Image 1 is a new plot, image 2 is the old plot - both of scalar velocity in the XY plane. The data seems to be offset in both of the axes. Furthermore, the data seems to show different colours at some cells - perhaps indicating a change in the way velocity has been calculated?

There is a lot confusion going on whether the name should be uPPPP or uPPP or what else.
The name of the framework will be Universal Post-Processor for Particulate Processes, hence 4 Ps make sense. however that is a bit much to type in python. (e.g import upppp)
Hence I propose "up4" or, which sounds even better "up3" (Universal Post-Processor for Particulate Processes)

Please let me know what your thoughts about that !

Following code returns a field of NaNs

import up4
import numpy as np

data = up4.Data("test.hdf5")
grid = up4.Grid(data)
velfield = data.velocityfield(grid)

gran_temp_field = data.granular_temperature(field)
gran_temp_2d= gran_temp_field.collapse(0)

print(np.isnan(gran_temp_2d).all())

## returns True

While it is not the way the grid should be used, this bug should not happen.
gran_temp_field contains numbers that are not NaN, the grid.collapse() function should be able to produce a proper field. (slices work)