Lengthy detailed GIT and ASDF LISP story

(defsystem "nud"
  :depends-on (:nud/user)
  :class :package-inferred-system)

(uiop:define-package :nud/user
    (:mix :nud/main :cl))

Note that packages are referred to by the path to their source file within the system, not their registered package name. If you want to register a differe package name, you could =(register-system-packages :nud/user '(:nud-user))= in your .asd file. However, you still refer to :nud/user (i.e. its path) in :depends-on, :mix, :use and so forth. However the package name it is Truly Registered Under will be :nud-user rather than the default, :nud/user. Actually, I never do this, since the old-fashioned-style nicknames just take a whole package name as an alias for the package. So I don't, but I could

(uiop:define-package :nud/user
    (:mix :nud/main :cl)
  (:nicknames :nud-user))

And this is similar to having register-system-packages :nud/user to :nud-user, except that I can also refer to the package (truly) using :nud/user, and I would get this when I introspect the package's name I guess.

(slime)

(require :nud)

(in-package :nud/user)

(uiop:define-package :nud/main
    (:mix :cl)
  (:export #:gird #:move-in #:move-on #:look
           #:} #:}} #:{ #:reg-act #:cursor-set
           #:] #:]] #:[ #:[[
           #:&))

my naming decisions notwithstanding, I have forgotten to export =#:@=, which I added to work like =@= in VI (it executes whatever is in a register, as vi expressions/macros).

But, I forgot to export it.

(uiop:define-package :nud/main
    (:mix :cl)
  (:export #:gird #:move-in #:move-on #:look
           #:} #:}} #:{ #:reg-act #:cursor-set
           #:] #:]] #:[ #:[[
           #:& #:@))

(asdf:load-system :nud/main :force t)

the =:force t= is necessary, because asdf's default behaviour is to leave packages that are already loaded alone. Forcing lets it know to overwrite the already-loaded package in places that changed.

Well, now I can use the exported symbol (it's a function)

(nud/main:@ g)

; @ performs an operation on a particularly structured cons tree conventionally named g.

But since I only recompiled =:nud/main=, I don't have this newly exported symbol mixed into =:nud/user=, which I would like it to be.

(asdf:load-system :nud/user :force t)

Now, =:nud/user= gets recompiled (now mixing =:nud/main:@=), but =:nud/main= did not need to be recompiled. And =@= is mixed into =:nud/user=.

(@ g)

. In this case I could have just =(asdf:load-system :nud/user :force t)= which would have found and recompiled =:nud/main='s changes, but not doing that was a learning experience.

Note that variables and things unique to the =:nud/user= and =:nud/main= package instances loaded into the repl were preserved, unless they were explicitly overwritten by a source change, so you don't have to interrupt what you were doing at all. I would call this hot-recompiling dependencies.

(defsystem "nud"
  :depends-on (:nud/user)
  :class :package-inferred-system)

(uiop:define-package :nud/user
    (:mix :nud/main :cl))

(uiop:define-package :nud/main
    (:mix :cl)
  (:export #:gird #:move-in #:move-on #:look
           #:} #:}} #:{ #:reg-act #:cursor-set
           #:] #:]] #:[ #:[[
           #:& #:@))

(in-package :nud/main)

(uiop:define-package :nud/user
    (:import-from :arrokoth/marklib)
  (:mix :nud/main :cl))