Dennis Heimbigner

Ward- are you in a position to get me that stack trace?

You are probably better off calling nc_get_vara and then doing the stride yourself in-memory.

Using nc_get_vars is very slow compared to nc_get_vara. So you get to choose performance vs python code complexity. I will have to let you decide the trade-off.

https://github.com/jswhit is correct. Sorry for being unclear. The tradeoff is that instead of using nc_get_vars, you use nc_get_vara. This will read in extra data, but is much faster. Once in...

Impressive. I need to see if I can implement this same strategy directly in the netcdf-c library.

Ok, let me see what I can do.

This PR was apparently made moot by other, later PRs.

Can you add a test case showing a failure in the existing code that is fixed with your change?

Ward- here is some additional information: https://stackoverflow.com/questions/46539084/what-is-the-block-size-in-hdf5

I guess I am dense, but I cannot see the purpose behind the block concept. Can't we just set block to be a vector of ones?