Skip to content

Conversation

@wfspotz
Copy link
Contributor

@wfspotz wfspotz commented Jun 30, 2014

I was reading through this in detail in order to design a PyTrilinos adaptor, and found some changes I wanted to make.

One question: a 'dim_data' tuple of length greater than one, where one or more of the dimension dictionaries has a 'dist_type' of 'u' would represent a structured array of unstructured dimensions. I am not aware of any uses cases for this configuration. Are you supporting it? I would be OK with a constraint that if 'dist_type' == 'u', then len(dim_data) = 1 must be true.

I was reading through this in detail in order to design a PyTrilinos adaptor, and found some changes I wanted to make.

One question: a 'dim_data' tuple of length greater than one, where one or more of the dimension dictionaries has a 'dist_type' of 'u' would represent a structured array of unstructured dimensions.  I am not aware of any uses cases for this configuration.  Are you supporting it?  I would be OK with a constraint that if 'dist_type' == 'u', then len(dim_data) = 1 must be true.
@kwmsmith
Copy link
Contributor

kwmsmith commented Jul 2, 2014

We do have usecases for arrays with a mixture of unstructured and structured dimensions. And down the road, we may want to allow individual dimensions to be unstructured for use with fancy indexing in a specific dimension.

Why can't PyTrilinos's adapator raise an exception if it gets a distributed array it can't handle?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants