Feature Request: Roswell integration (`ros install cando`)

[ItsMeForLua]

I'd like to propose adding Roswell integration for Cando, enabling users to install it with a simple ros install cando command. This would significantly lower the barrier to entry for computational chemistry researchers interested in trying Cando and help expand adoption within the broader Common Lisp community.

Motivation

Current Installation Challenges

• Cando has complex build dependencies (Clasp, LLVM-14, boost, clang, etc.)
• New users face a steep learning curve just to get started
• The computational chemistry community often prefers straightforward installation methods(as far as I'm aware atleast)
• Manual builds can be time-consuming and error-prone across different environments

Benefits of Roswell

• Simplified onboarding: ros install cando vs complex manual build process
• Environment management: Roswell handles multiple Lisp implementations cleanly
• Wider adoption: Makes Cando accessible to Lisp developers who might contribute
• Scientific reproducibility: Consistent installation across research environments
• Integration with Lisp ecosystem: Natural fit with other ros install packages

---

I'm currently working on fixing the underlying ros install clasp functionality (which is currently broken due to archived dependencies). This would need to be resolved first, as Cando depends on Clasp.

See here
and here

Possible Approach

The Roswell integration could handle the full dependency chain:

1. Install/verify Clasp with required LLVM version
2. Download and build Cando within the appropriate clasp/projects directory
3. Set up environment variables and path configuration
4. Verify the installation with basic smoke tests

Potential Considerations

I understand this is more complex than typical Roswell packages due to:

• Heavy system dependencies (LLVM, boost, etc.)
• Domain-specific build requirements
• Active development potentially changing build processes

My Offer

I'm willing to contribute the development work for this integration, including:

• Writing the Roswell installation scripts
• Testing across different Linux distributions
• Helping main the integration as Cando inevitably grows
• I can write documentation for users and maintainers(I like writing)

Proposed Approach

1. Phase 1: Get ros install clasp working reliably (in progress)
2. Phase 2: Create proof-of-concept cando installer script
3. Phase 3: Test across multiple environments and edge cases
4. Phase 4: Submit to Roswell maintainers for inclusion

Timeline

I'm happy to work on this at whatever pace makes sense for the project. Given the complexity, I'd expect this to be a multi-month effort done in collaboration with the Cando team.

Additional Context

This aligns with the broader trend of making scientific computing tools more accessible. Languages like Python have benefited enormously from simple installation methods (pip install), and the Lisp scientific computing ecosystem could benefit similarly from ros install integration.

I believe Cando represents some of the most exciting work happening in computational chemistry. I personally think that better accessibility could help it reach a MUCH wider audience of both users and contributors.

[yitzchak]

Do you know about the packaged binaries? We can add other build platforms. https://github.com/clasp-developers/clasp/wiki/Installing-Using-a-Package-Manager

Although I haven't used Roswell for a while, and I suppose it could have improved, my last opinions of its code and status weren't high.

[ItsMeForLua]

@yitzchak Honestly, I completely understand. For me, my perspective on Roswell is not really in regards to its current status, but more-so its potential.

I do think the binaries y'all offer are great, but they require some manual configuration for me to get them compatible with my Roswell environment, since the system packages install Quicklisp standalone while Roswell manages Quicklisp within its own environment. This creates some friction when trying to use CANDO alongside other Lisp projects I'm managing through Roswell. I also find using Roswell for CI/CD or really any automated setup for lisp environments quite easy once you get the right dockerfile, and/or Makefile.

That's ultimately what brought me to the idea of having it available from ros install , because it would solve the environment isolation issue for users already invested in the Roswell ecosystem.

But I completely understand where you're coming from. Maintenance is obviously slow on Roswell, and there are a lot of lackluster features, but I do believe it's a good canvas/foundation that could be built up further with the right maintainers and contributors.

Would you be open to me exploring both approaches? I could look into improving the existing package integration while also investigating the Roswell path as a longer-term option.

If you're not comfortable with the Roswell idea, i could maybe fork and try and setup a pseudo Roswell install via a makefile.

[yitzchak]

What about starting with updating the roswell install for Clasp so that it actually uses the prebuilt binaries? Avoiding having a C++ build environment might be a bigger help.

[ItsMeForLua]

Thanks for the feedback, that's actually a great point. The sbcl installation model is an example of using that approach. Since ros install sbcl installs the binary. However, there are situations where the binary is lacking very specific features, that need to be installed via a make flag when compiling from source. So they offer: ros install sbcl-source

Which could be a valid Addition, but I agree that starting with just installing the binary is the best option, and standard.

Instead of a full source build, I'd like to propose we first get a ros install cando (and ofc a ros install clasp) command working that uses the pre-built binaries you've already made available. This would immediately provide a fast, easy, and reliable installation experience for most users and directly address the friction of the manual setup I mentioned.

For more advanced users or contributors who need to work with the source, we could then explore a second option: a ros install cando-source option. This would require fixing the ros install clasp to handle the full build process, which I am still committed to developing.

Does this seem like a reasonable way forward? It would allow us to get a win for new users quickly while keeping the door open for a more flexible solution down the line.

[yitzchak]

Yes, that makes sense. In any case, building from source should be much easier then the old method used with roswell install clasp since we have koga versus waf now. We also have source packages that include all dependencies. See this Arch PKGBUILD for how they are used.

https://aur.archlinux.org/cgit/aur.git/tree/PKGBUILD?h=cando

[ItsMeForLua]

Perfect

All that will make things much more straightforward