I've patched in the code I have on top of the commit and created
https://github.com/cffi/cffi/pull/186
Including some changed doc. Please review and feel free to edit. I'd appreciate stackmith's comments too - this may be more complex than stacksmiths code, and I'm a bit doubtful about that.
* Luís Oliveira luismbo@gmail.com CAB-HnLT_fAE5tQCx-bqYWxZjFW4dHeDCZ9tsz=AUCmUZtB+Tbw@mail.gmail.com Wrote on Sun, 5 Sep 2021 17:26:00 +0100
I agree with your proposed improvement. Could you prepare a pull request? On Sun, 5 Sep 2021 at 04:16, Madhu enometh@meer.net wrote:
I notice this commit to CFFI:
- commit 743a90251e5a4407147a20bd8354df03a87ca46d
Author: stacksmith stacksmith@users.noreply.github.com AuthorDate: Thu Aug 19 06:34:28 2021 -0700 Commit: GitHub noreply@github.com CommitDate: Thu Aug 19 14:34:28 2021 +0100
Allow local names in WITH-FOREIGN-SLOTS It remains compatible with existing code, but accepts additionalformats for bindings: 1) (name slot-name) - just like WITH-SLOTS; 2) (name :pointer slot-name) - a pointer version.
On Thu, 16 Apr 2020 15:24:52 +0530 I had propsed a similar patch to Luis With a slightly different signature.
| https://github.com/enometh/cffi/commit/023676a25128786174bbaa8a5df7fdc1d27c3... | It seems a natural extension to me, though I was not able to | document it to my satisfaction and I don't know if it would be too | confusing to others. WDYT - (could clean it up and push it on a | separate branch if you had comments.)
The doctring from that patch is reproduced here:
-- (with-foreign-slots (bindings ptr type) body) - Now each binding can be one of these forms:
SLOT-NAME -- binds SLOT-NAME to (FOREIGN-SLOT-VALUE SLOT-NAME)
(:POINTER SLOT-NAME) -- binds SLOT-NAME to (FOREIGN-SLOT-POINTER SLOT-NAME)
(VAR-NAME SLOT-NAME) -- binds VAR-NAME to (FOREIGN-SLOT-VALUE SLOT-NAME)
(:POINTER (VAR-NAME SLOT-NAME)) -- binds VAR-NAME to (FOREIGN-SLOT-POINTER SLOT-NAME)
(VAR-NAME (:POINTER SLOT-NAME)) -- binds VAR-NAME to (FOREIGN-SLOT-POINTER SLOT-NAME)
This proposed syntax is different from what has been implemented by stacksmith: The last two forms above correspond to the form
(name :pointer slot-name)
Common Lisp bindings follow a standard syntax LHS RHS and should always be destructurable to a LHS RHS
I believe introducing a nonstandard binding form with 3 elements is not in the spirit of common lisp syntax and it will only necessitate unnecessary special-casing for further macrology
I would encourage stacksmith to consider this point and try to adopt the syntax I proposed for extending with-foreign-slots
Or would it be possible to support both these forms in addition to the new 3 element binding form
It is still early and perhaps that the syntax can be fixed before it is cast in stone - I'm hoping Luis can cooperate