The DCPL branch #140
The DCPL branch #140
Conversation
licenses/lzf.txt
Outdated
| @@ -0,0 +1,33 @@ | |||
| Copyright Notice and Statement for LZF filter | |||
There was a problem hiding this comment.
I wonder now if we have to include licenses for lzf/blosc/h5py: because we no longer include C sources for lzf/blosc (which, I think, was the case early on) - as we rely on blosc-sys and lzf-sys. So I think the entire licenses/ can be now probably removed.
There was a problem hiding this comment.
I think the rework of automatic chunking also means we now no longer have any direct code from h5py. We also comply with the hdf5 license by including this in the hdf5-src archive. We should be safe to remove licenses (but I am not a lawyer).
There was a problem hiding this comment.
Yes, I think they are safe to remove now.
| } | ||
| } | ||
|
|
||
| macro_rules! impl_builder_stuff { |
There was a problem hiding this comment.
Wonder - are the generated docs ok with these macros?
There was a problem hiding this comment.
The docs are replicated on all relevant types, this is duplication, but it all shows as a separate impl block. I've added a small comment on the impl blocks to describe how they are all common between dataset-builders
|
@mulimoen I've went through the code once again and (the lack of docs and docstrings aside, and also the complete lack of tests aside) it's pretty good to go. I left a few minor comments here and there. (I really think we should write some docs and tests after this is released and is considered stable api-wise.) |
mulimoen
force-pushed
the
feature/dcpl-2021
branch
from
949a274 to
092bab3
Compare
|
I've opened some new GH issues for some of the remaining tasks. This PR is enourmous as it is, and the other issues can be merged and reviewed separately. |
|
Good point re: separate tickets (array trait, docs, avoiding panics), they can be handled separately. Maybe also a ticket for adding full tests for |
|
At this point I'm somewhat scared to press the green button because this PR is so big my browser sometimes hangs when viewing diffs :) Great work @mulimoen in resurrecting it and finishing all the unfinished parts! (Before the release, as I've said, I think we should consider merging at least the ownership/IDs PR) |
|
(Added a note that this fixes #47 as well) |
Only halfway, one would need to dispatch based on some fixed sizes. |
Hm? The question was to create a dataset with runtime-known string size. This is now doable via |
|
I think what @pmarks is trying is creating the type from an |
|
Thank you very much for reviewing and finishing this @aldanor! Hopefully future PRs won't be as massive as this one! |
|
Yea - well we solved it halfway by allowing runtime descriptors, that's already something that wasn't previously possible, so that's something. Also, that 'giant match' can now be implemented via const generics. Finally, with strings... I have a ticket with a hacky solution for Rust types like strings - maybe we'll get to that at some point as well. |
Yea, let's try to keep it simple from now on :) I may open a few PRs in the nearby future with some docs updates. |
This is a continued effort on the
dcpl-branch of @aldanor. This adds a lot more functionality and makes the API much nicer to use for selections.See mulimoen#2 for some earlier comments and discussions regarding this branch.
Fixes #45
Fixes #88
Fixes #126