cel-spec VS sci

My employer uses a combination of Protocol Buffers (for the config schema definition) and Bazel/Starlark (for concrete instantiations). Configs are validated at build time and runtime using CEL (https://github.com/google/cel-spec).

The Common Expression Language (CEL) implements common semantics for expression evaluation, enabling different applications to more easily interoperate. https://github.com/google/cel-spec

Looks like Google invented a specification for a simple "expression language." -> https://github.com/google/cel-spec/blob/master/doc/langdef.md. Writing the expressions feels like writing Java, C++, Go, or TypeScript code. Google then released C++ and Go versions of this langauge as a library.

https://github.com/google/cel-spec unfortunately, it's in Go or C++. Of course I can write a binding to them. But is there any other similar that you would know of for Java? My other course of action would be to offload computation to another service using this library in Go, or Jsonnet or Open Policy Agent/Rego based evaluation, which I'd prefer not to. Executing JS in Java via Nashorn also an option but it'd be heavy weight.

Having a standard way to share expressions does seem quite useful, especially when it's multilingual.

[0]: https://github.com/google/cel-spec

There's more information about CEL and its specifications here

Especially when dealing with more complex match rules, I personally much prefer Caddy's matchers over building some weird-ass if constructs in Nginx. It also supports CEL for request matching, giving you access to extremely powerful logic, if you need it.

I've tried researching this myself and can't find too much. There's this project metaes which is an mci for JS, and there's the SCI module of the Clojure babashka project, but that's about it. I also saw Triska's video on mci but it was pretty theoretical.

https://github.com/babashka/SCI if it's a requirement for proper sandboxing

Since cherry is a compiler, the code generally runs faster than with SCI which is an interpreter. For many cases SCI is fast enough, but numerical computations in a hot loop isn't one of its strenghts:

Startup and sustained performance are absolutely implementation issues. For example, SBCL will take its sweet time to make machine code out of Common Lisp, but CLISP will interpret and generate bytecode. Both are useful, and both implement the same language. Clojure on the JVM takes also takes plenty of time to start up, so some use an interpreter instead. Furthermore neither of these languages has a cost model, so the cost of anything is an implementation issue.

So far we have relied on an increasing number of home-grown integration points to our platform, where relevant combined with the excellent SCI (so we can write some Clojure-code when adhoc data conversions / calculations / tweaking is required).

Also take a look at SCI, https://github.com/babashka/sci/blob/master/doc/libsci.md

You probably want to take a look at sci if you are creating a DSL or want to use Clojure itself as your DSL.

Built with the lovely SCI library (https://github.com/babashka/sci) + GraalVM, probably the most useful GraalVM project I've seen in the wild so far.

Also, Babashka will probably always support more features than ClojureRS could ever, particularly the interop with the various Java classes/functions, as that'd be very hard to achieve in ClojureRS.