On the evaluation formalism in Red (Rebol, Spry, ...)

[dumblob]

@meijeru I think to increase credibility of a language and the overall understanding, more formalism would undoubtedly be beneficial.

Recently I came across Call By Push Value which seems to be the underlying central building block of Rebol/Red/Spry/Levy/... languages.

Could you take a look and add it to the docs?

[meijeru]

I have some difficulty with your statement " Call By Push Value [...] seems to be the underlying central building block of Rebol/Red/Spry/Levy/... languages". Can you explain why you think CBPV is applicable to (the evaluation mechanism of) Red?

[9214]

Thanks for the link @dumblob, that's an interesting paradigm. But I'm afraid it's not related to Red and Rebol, can't say about other languages.

As I understood from a brief reading, CBPV provides a unifying formalism for λ-calculus into which both call-by-value and call-by-name reduction strategies can be converted, which is useful for PLDI research — instead of 2 different semantics you use just 1.

Red's interpreter uses call-by-value by copying arguments (value slots) onto evaluation stack, but some values contain references to external data outside them, thus giving the ability to mutate said data in-place (unless you explicitly copy it), which looks like a call-by-reference to the uninitiated.

Other dialects (eDSLs and general-purpose languages) can use entirely different calling conventions and implement their own runtimes around what you call "underlying central building block" † — the common data format (which Red is, by the virtue of homoiconicity).

So, yes, in that sense Red can subsume call-by-value and call-by-name just like call-by-push-value, but IMO that's too far-fetched. "Functional/imperative synthesis" also somewhat resonates with Red, but that's debatable too.

In our previous encounter, I gave you some pointers on how evaluation in Red and Rebol works. Please review them once more and ask questions in our community chat if something needs further clarification.

† It's really a block!, no pun intended.