Thanks Larry. I'll have that PR for a first pass submitted in a few hours.
The only one that's come up so far is the geterror() mechanism (returning a
string, which means caching it with a thread_local in the C layer) but I'll
write more about that later.

Cheers,
Anders

On Thu, 22 Oct 2020 at 17:13, Larry Gritz <[email protected]> wrote:

> I just want to make sure you know...
>
> If you see something in the APIs that is "too C++" in the sense that it is
> extremely reliant on very C++-specific idioms, not only in expression but
> also conceptually, and therefore difficult to map to C or the way any other
> language would see things, then by all means make suggestions or even add
> something new on the C++ API to make the C wrappers easier. Don't feel like
> the C++ API is 100% set in stone and a sacred item that can't be changed.
> Your design space can extend into the existing interfaces if there are ways
> to improve them.
>
>
> On Oct 20, 2020, at 8:33 PM, Anders Langlands <[email protected]>
> wrote:
>
> Sounds good Scott. Once I’ve sketched out something and we all agree on a
> general approach let’s split the bindings into small chunks we can each
> tackle and PR piece wise and do the same for the Rust crate?
>
> On Wed, 21 Oct 2020 at 14:49, Scott Wilson <[email protected]> wrote:
>
>> That works for me. I'll start contributing bindings once Anders sends in
>> the first PR.
>>
>> On Tue, Oct 20, 2020 at 5:39 PM Larry Gritz <[email protected]> wrote:
>>
>>> On Oct 20, 2020, at 1:56 PM, Anders Langlands <[email protected]>
>>> wrote:
>>>
>>> Ok so then what I think I’ll do is port the C part of my rust crate over
>>> to oiio proper, implementing just enough to add a simple round-trip test,
>>> is load an image, add an attribute to the header, write it out as an exr.
>>> Then I’ll make a PR so you can comment. Sound good?
>>>
>>>
>>> Awesome.
>>>
>>> And I would recommend that we approach the C bindings piece by piece.
>>> I'm a fan of breaking it up into many small PRs so they can each be
>>> reviewed quickly, one class at a time. Start with the easy ones and along
>>> the way I'm sure we'll hash out the design issues and have a consensus by
>>> the time we get to the big hairy classes. If we try to do it all at once,
>>> we could end up going down the wrong road and only discover when in the
>>> middle of reviewing and integrating an enormous amount of code that we
>>> wished we had made some different basic decision.
>>>
>>> -- lg
>>>
>>>
> --
> Larry Gritz
> [email protected]
>
>
>
>
> _______________________________________________
> Oiio-dev mailing list
> [email protected]
> http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org
>
_______________________________________________
Oiio-dev mailing list
[email protected]
http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org

Reply via email to