Chapel: Sparse Arrays/Domains Meta-issue

[e-kayrakli]

This is a metaissue to track sparse array/domain issues.

Functionality

• [ ] Slicing (#8023, and many others linked from that)
• [ ] Iterators in subset of dimensions (#7374)
• [x] Buffered bulk index addition (#12672)
• [x] Local index addition (#12673)
• [x] Slicing a dense array with a sparse domain (#12704)
• [ ] Creating a dense array from a sparse one (#13441)
• [ ] Bulk index removal (No issue yet, probably low priority)

Performance

• [ ] Sparse Array assignment (#9344 and to some extent #9364)

Design Questions

Note: The issues below mainly need a community discussion and decision-making in terms of what the desired behavior is. Lack of implementation is not the primary concern.

• [ ] Cross-type assignments for arrays (#7788)

• [ ] Zippering sparse arrays/domains with other iterables. (#12424, #11504, #9437, and many others. Oldest issues I could see: #6577)

• [ ] Dense followers (#13452)

Miscellaneous

• [ ] Investigate an extension to DSI to support fast sparse (and associative?) domain assignment
• [ ] Revise distributed sparse domain assignment
• [ ] Add more sparse microbenchmarks
  
  
  
  • [ ] Same-type and cross-type assignments
  
  

[e-kayrakli]

@ben-albrecht suggested creating this issue both to track the issues and help those who are interested in contributing to the sparse domain/array support in Chapel.

Anyone with the right permissions should feel free to modify the contents of the original post.

[himbhadani]

Thanks @e-kayrakli and @ben-albrecht Sir,
This meta-issue is surely going to help me a lot.:)

[bradcray]

Only loosely related, but I wanted to mention that I'm putting in an effort to look into what it would take to slice dense arrays with sparse domains.

[e-kayrakli]

@bradcray -- I am trying to track what happens on github but, if you happen to open an issue capturing what you think about that, IMHO it should be listed in this issue as well. Such support would definitely increase the usability of sparse domains.

Feel free to ping me up if you open an issue and I miss it.

[bradcray]

I think you're right that I haven't opened up an issue for this case yet. I was hoping to just get it done before ever finding the time to open the issue. :)

[bradcray]

(in fact, I just tried it for the first time on the slicing branch I've been developing independently of the sparse domain slicing question, and it seems to work! :) ).

[e-kayrakli]

A PR is always better than an issue, I guess :)

[bradcray]

The branch is still going well, but I opened issue #12704 all the same.