Hi,
I'm creating VDB files using the Python API. I'm running the Python standalone (i.e., not within Houdini) to generate the VDBs. These VDBs are based on data from climate simulations. The Python code runs fine and generates VDBs. I'm able to view them in vdb_view and vdb_print (which come standard in a OpenVDB installation). But when I try to load the VDB into Houdini, I get an error:

Unable to read file /path/to/file/CLOUD_20101210_04.vdb
Expected array object (near byte offset 109, line 1, column 10)
Backend IO

Here's a link to a vdb file that has this problem:
https://svs.gsfc.nasa.gov/prepub/greg/vdb/ [svs.gsfc.nasa.gov]

I was occasionally able to get some of the files to load, but mostly not... How can I create VDB files using the Python API that will reliably load into Houdini? Is there a special tag that is needed?

Thanks,
--Greg < SVS >

For others who may have this problem... we've figured this out. Apparently, Houdini 18.5.351 (and likely prior versions) is not compatible with recent VDB formats. We had to back-rev way back to OpenVDB 3 in order to get files written out using pyopenvdb to load properly into Houdini.

It sure would be nice if Houdini recognized more recent VDB formats - or at least let the user know that the VDB that you're attempting to load is incompatible, rather than displaying a cryptic message about byte offsets.

I've found that VDB files written with pyopenvdb 223 works with Houdini, whereas 224 (which recent versions of OpenVDB come with) does not. You can find which version you have by running the following on the command line:

% python
Python 2.7.5 (default, Nov 16 2020, 22:23:17)
on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import pyopenvdb
>>> pyopenvdb.FILE_FORMAT_VERSION
223

--Greg < SVS >

I'm not seeing the same behaviour ...



-Drew

you can also check here which version is supported in latest houdini
https://www.sidefx.com/docs/houdini/licenses/index [www.sidefx.com]

currently it seems to support OpenVDB 7.2.2