Rescale to data range sometimes produces garbage bounds with XDMF HDF5 file

Hello Paraview community, this is my first post!

I’ve been experiencing some issues with visualizing a model output HDF5 data set using XDMF3. The problem: intermittently, the “Rescale to Data Range” tool sets the colormap bounds to seemingly garbage values, often on the order of ± 1E30 but it varies with every load; the data set values all lie in [0,1]. Interestingly, when this happens, the “Rescale to Visible Data Range” tool sets the colormap correctly.

About the data: The .h5 file is created and written to in Python using h5py, and I have written code that manually writes the .xmf file. I load the data in Paraview with the XDMF Reader. The data exists as cell data in a 2D curvilinear grid, with dimensions stored as [time, spatial dim1, spatial dim2]. In the example file attached, the cell dimensions are [5, 600, 10] and the grid point dimensions are [601, 11]. I use the “hyperslab” XDMF item type to get slices of the cell data at each time value – I’ve seen elsewhere that hyperslab doesn’t always play nicely with Paraview, so maybe this is where the problem lies?

The most recent time that I loaded this file, the automatic colormap scale for time=1 was [-1.4E-70, 7.7E291]. When I reset the session and reload the same exact file, the colormap auto scaled correctly to [0, 0.78]. But then I reset/reload again, and the auto colormap is [-8.4E298, 0.0] !

I have triple quadruple checked the dataset and am confident that the entire .h5 cell array is written to, so I don’t think that it contains garbage data.

Any ideas as to what’s going on? All help is appreciated.

OS: Windows 10 v 20H2
Paraview version: 5.9.1 downloaded from Paraview website
Graphics hardware: Intel UHD Graphics 620

Thank you,

p.s. This hasn’t mattered for other .h5 files I’ve visualized but just in case it’s important: the .h5 and .xmf files are stored in the WSL2 file system, Paraview installed on the Windows side.

cells.h5 (288.0 KB)
cells.xmf (4.8 KB)