2020 Status


p.z.trotter@...
 

Hi, all.

I am thoroughly enjoying going through L.B.Stanza's codebase and interested in contributing.

There appears to be semi-regualr activity in the git repo but fewer regular announcements here and the website news section is now a few years old.

What is the best place to learn what is happening/status/roadmap? orr just to ask questions? (e.g. REPL status for example)
or to discuss areas to contribute/contribute - beyond current github issues?

thanks much.

p.

 

 


 

Hello!

Stanza development is active however the website has definitely fallen behind just because I have been so busy. The short of it is that I founded a company called JITX when I graduated, where we are writing a programming language for designing electronics, and Stanza is our main development language. So development on Stanza is proceeding along because of that. The experimental branch has a REPL now (quite a fancy one too, that I need to write about) and a very convenient built-in build system and test framework.

The main features slated for development are an LLVM backend for Stanza, and Language Server Protocol support. We're very excited for both of them. The LLVM port should allow Stanza to run on just about every platform under the sun, and the LSP support will allow Stanza to be nicely integrated into most of the major IDEs.

This forum is still the best place for discussions/questions.

Cheers, Patrick


p.z.trotter@...
 

Hi Patrick,

Thanks for the quick response - very helpful and very encouraging.

I use LLVM iwith OpenDylan and its proved very useful for C/C++  libray  and platform integration

Stanza is a already a lovely language to work with and I can see LLVM backend and LSP support will  help drive a lot of adoption.

I'll dig into the experimental branch and post to the forums if i have any questions.

thanks much.

P,

 


 

Cool, a Dylan fan! Jonathan Bachrach, my advisor, was one of the early people to work on that system. It influenced Stanza's design greatly.

Sounds good. Just post here if you have any questions.

Patrick


Eric Toombs
 

I asked a few years ago about guarantees for pure functions. Any work
happening there?

On 2020-01-05 13:21, Patrick S. Li wrote:
Cool, a Dylan fan! Jonathan Bachrach, my advisor, was one of the early
people to work on that system. It influenced Stanza's design greatly.

Sounds good. Just post here if you have any questions.

Patrick


 

Hi Eric,

The purity annotations has been deferred until later. One of the primary reasons for adding it is to give the optimizer more information to make more aggressive transformations. However, there's a number of other optimizations that will have more impact that we decided to prioritize first. (Such as the LLVM backend).

Patrick


Eric Toombs
 

The optimisation thing, yeah, that's not too important compared to LLVM.
But knowing a function is guaranteed not to have side effects would be
nice for code reasoning and debugging purposes.

On 2020-01-07 02:10, Patrick S. Li wrote:
Hi Eric,

The purity annotations has been deferred until later. One of the primary
reasons for adding it is to give the optimizer more information to make
more aggressive transformations. However, there's a number of other
optimizations that will have more impact that we decided to prioritize
first. (Such as the LLVM backend).

Patrick


p.z.trotter@...
 

hi Eric ,  re: pure  function annotations

D-lang also has a pure function attribute thats proved very helpful for debugging and thinking about code  side effects -

so agree its very valuable feature for stated reasons, but also agree with Patrick on priority. 

Once the LLVM backend is in place - I  think this should be relatively straightforward to add via existing  LLVM  infrastructure for function attributes. 
(if interested - Seth Goldstien@ Carnegie Melon uni has a good write up of how add function attributes  when going through clang to llvm see http://www.cs.cmu.edu/~seth/llvm/llvmannotation.html )

Not sure at this point, how to port to existing backends - but thats just a matter of getting used to codebase - so I'll have a think on this - since its something I would like as well

Eric  - can you create  an issue on github , so it doesnt get forgotten and add your use case info then we can see if its doable as a community contrib?

 

Patrick - re; Dylan and Jonathan : Yup - I knew of Jonathan's work at Harlequin, and your thesis nicely covers supporting influences and related works.

BTW - Would it be possible to add your thesis to the Stanza website documentation ?- its a good read and might be very helpful for folks discovering Stanza.

 

- Phillip

 


Eric Toombs
 

Ah yeah, here it is, Philip: https://github.com/StanzaOrg/lbstanza/issues/43

*Somebody* closed the feature request :D

As for a full use case, huh, that will take longer. I have struggled
before trying to convince people that hey pure functions are actually
super useful, always thinking the usefulness was self evident.

The biggest resistance I always get is "BUT MUH PRINTF!!!1!" basically.
(see https://github.com/vlang/v/issues/3090)

That and, "I just wanna change my variables!" I get that one a lot from
game devs.

The worst effects are typically only truly unmanageable at scale, so I
will need some time to think, but I will try to cook up some really
good, plausible, but still small-scale bugs that could have been easy to
fix, or perhaps avoid altogether with pure functions. Also, lol, cooking
up bugs. Tasty. Actually, though. Fried mealworms are delicious. But I
digest.

On 2020-01-07 18:37, p.z.trotter@gmail.com wrote:
*hi Eric , * *re: pure  function annotations*

D-lang also has a pure function attribute thats proved very helpful for
debugging and thinking about code  side effects -

so agree its very valuable feature for stated reasons, but also agree
with Patrick on priority. 

Once the LLVM backend is in place - I  think this should be relatively
straightforward to add via existing  LLVM  infrastructure for function
attributes. 
(if interested - Seth Goldstien@ Carnegie Melon uni has a good write up
of how add function attributes  when going through clang to llvm
see http://www.cs.cmu.edu/~seth/llvm/llvmannotation.html )
<http://www.cs.cmu.edu/~seth/llvm/llvmannotation.html>

Not sure at this point, how to port to existing backends - but thats
just a matter of getting used to codebase - so I'll have a think on this
- since its something I would like as well

*Eric*  - can you create  an issue on github , so it doesnt get
forgotten and add your use case info then we can see if its doable as a
community contrib?

 

*Patrick* - *re; Dylan and Jonathan* : Yup - I knew of Jonathan's work
at Harlequin, and your thesis nicely covers supporting influences and
related works.

BTW - Would it be possible to add your thesis to the Stanza website
documentation ?- its a good read and might be very helpful for folks
discovering Stanza.

 

- Phillip

 


 

I do agree that pure functions are super useful if the analysis is accurate enough. The main reason that we deferred this feature was that without a lot of work, our analysis would be way too conservative. Almost everything would be flagged as potentially impure because Stanza is too flexible in what it allows users to do.

I'll add the Stanza thesis to the website. Thanks for the suggestion!