Clojurians Log v2

Any particular use case you have in mind for a tool like codeq?

Universal semantic code repository.

Content addressable something like unison language or convex project's orthogonal persistence.

All code and versions can be merged into one graph.

Then build out tooling on top of that.

Have you thought about how a codeq’s identity does or does not change w.r.t. the codeqs it depends on? Eg if codeq-a depends on codeq-b and codeq-b has a bug fixed in it, does codeq-a now transitively get a rev bump as well? Its the difference between identity based on code as text or identity being based on code-as-text+the codeqs it depends on. If you were to build, say, a repository of generative test seeds and results, and wanted to only rerun tests for codeqs that changed, the transitive dep being part of a codeqs identity matters.

content hashes and merkle trees

so if code changes it gets a new hash

similar to this https://www.unison-lang.org/docs/the-big-idea/

What if you don’t have the source of a fn, just the lib version it came from?

not sure. seems like blobs would not be able to participate semantically

could just use the hash of the blob in that case

eg. if the lib is in nix packages for example, it would already have a deterministic reproducible hash

The coordinates, right? “com.foo/lib mvn/version 123, containing codeq-b” + hash of codeq-a

no coordinates needed just a hash

coordinates could just be pointers to a hash

more like tags

so the idea is a bit bigger than codeq. more like a new thing that combines code repos and package management etc.

the convex project has a notion of orthogonal persistence also using merkle trees https://docs.convex.world/docs/tutorial/convex-lisp/convex-lisp

could be a product that is broader than just clojure

mostly though i just wanted semantic version control that would let you merge all repos into one and build tooling on that

was curious about datomic / codeq as an inspiration or base for that

curious if you have any thoughts

So many thoughts haha Codeq is one of my favorite ideas and I’ve thought about it for many years. I'm on mobile now and running errands but lets keep the convo going async. I will look at the convex project

thanks! doing some hiking in a bit but async works

@dustingetz codeq discussion in case you are interested

> if you were to build, say, a repository of generative test seeds and results, and wanted to only rerun tests for codeqs that changed, the transitive dep being part of a codeqs identity matters. Btw, I did build this (based on codeq), never released it

tests could certainly be optimized around when merkle hashes change

polylith seems to get some mileage out of testing only modules that change. a more fine grained capability would be even more powerful

That whole work grew out of the idea of specs and generative testing as a core. You don't want to check specs at runtime, you want to exhaustively (slowly) test each unit of code generatively and you do not need to retest things you have not changed

sometimes you have oddball third parties though where you can't predict what they will send but that is more at the service api layer

I am also interested in this. Think there are a lot of opportunities for something like codeq or related tools