This is an interesting idea. IIUC current implementation implies bindings are evaluated once per plait instance, right? Like:

(plait [a (do (prn "a evaluated") 1)
        b (inc a)]
  (plait [b (+ 2 a)])) ;; 3
;; Prints
;; "a evaluated"
;; "a evaluated"

This is correct! We haven't had the use case yet but you could even use it to trigger an extra evaluation by passing the empty vector (plait [] ...)

So the strategy is very naive to avoid any surprises.

Another thing worth mentioning is that all bindings need to be declared at the top level. This to make the ordering explicit.

Do you have the opposite use case? Eval bindings the minimum amount of times. Like statically tracking symbol usage to prune unused bindings and avoid binding repetition when an ancestor is shared:

(plait [a (do (prn "a evaluated") 1) ; a has 2 children
        b (prn "b shadowed " (inc a)) ; b is shadowed, no need to run (inc a)
        ]
  (plait [b (+ 2 a)]
    (prn "b " b) ; b must be evaluated here, so a must be too
    )
  (plait [b (+ 3 a)]
    (prn "b " b) ; b must be evaluated here, a is already evaluated
    ))
;; Prints
;; "a evaluated"
;; "b 3"
;; "b 4"


;;  (do (prn "a evaluated" 1)
;;   -----------------------
;;         |
;;         a
;;         |\
;;        /  \_
;;       /      \
;; (+ 2 a)  (+ 3 a)
;; ------   ------
;;    |       |
;;    b       b
;;    |        \
;;    |         |
;; (prn "b " b) (prn "b " b)

Ah no! That would be more in line with rspec from Ruby where a traversal of the directed acyclical graph means only the minimum code is called. We also experimented with this approach but were unable to get satisfactory let-style bindings without re-implementing let and therefore abandoned the approach. But it would be interesting to see! We experimented with prismatics fnk for that approach.

That's excellent news, congrats!

Thanks @U04V15CAJ! Looking forward to seeing dev-tools/scripts built on top of this: schema inferrers, coercers, pretty printers etc.

Very excited about this! Thank you for the work and effort bringing these together!

brilliant!

Not released on clojars, but copy-paste a file into one's project?

Otherwise, looks cool!

it doesn't really, it is using the transient as a mutable collection, bashing on it in place, which the docs specifically say not to do, and will not work on larger collections (transient operations may existing or return new values)

yeah, that's broken

@U03JDCVV1MX I can set aside some time if you need help publishing to clojars or maven central. You would probably want to select unique coordinates other than clojure.heap and java.core for the namespaces. Also porting to deps.edn would allow you to have it available as a git dep

@U3JH98J4R @UEENNMX0T Thank you for the kind suggestions. The project is on Clojars.

> Also porting to deps.edn would allow you to have it available as a git dep FYI, libs like this can be used as a gitlib by overriding tools.deps manifest detection, like so:

com.github.gnarroway/hato {:git/tag "v0.8.2" :git/sha "e75c194" :deps/manifest :deps}

Many thanks for the suggestions, much appreciated. We're looking into it.