Behavior of (import ...) in R7RS define-library forms

Behavior of (import ...) in R7RS define-library forms

[J. Vincent Toups]

I have two r7rs-style libraries, one of which depends on the other,
located in the following filesystem structure:

src/lib/shadchen-helpers.scm
src/lib/shadchen.scm

In principle, shadchen-helpers isn't useful on its own - it provides
some helper material for the shadchen library.

The library definitions look like this:

$ find src/ -iname "*.scm" | grep lib | xargs head -n 2
==> src/lib/shadchen-helpers.scm <==
(define-library
    (lib shadchen-helpers)

==> src/lib/shadchen.scm <==
(define-library
    (lib shadchen)

So if I start a repl like this:

$ java -jar lib/kawa.jar -Dkawa.import.path="./src/*.scm"

Kawa can find the libraries if I do an import:

#|kawa:1|# (import (lib shadchen-helpers))

The above works fine. However, in a fresh interpreter:

#|kawa:1|# (import (lib shadchen))
/home/toups/work/games/jmonkey3d-lacraw/src/lib/shadchen.scm:9:4:
unknown library (lib shadchen-helpers)
/home/toups/work/games/jmonkey3d-lacraw/src/lib/shadchen.scm:110:6:
warning - no declaration seen for pattern-bindings-equal?
/home/toups/work/games/jmonkey3d-lacraw/src/lib/shadchen.scm:117:6:
warning - no declaration seen for all-identifiers?

The import fails because the library reference to shadchen-helpers
fails to resolve. That reference in `shadchen.scm` looks like this:

  (import (except (kawa base) match)
      (kawa lib std_syntax)
      (srfi 28)
      (rnrs sorting (6))
      (rnrs hashtables (6))
      (lib shadchen-helpers))

I have noticed that importing shadchen-helpers at the repl first and
then importing shadchen works. I would guess because importing the
shadchen-helpers library compiles that library and then the import
form in the r7rs define-library can then find the compiled artifact.

But doesn't it make sense for the import form in shadchen to trigger
the compilation of shadchen-helpers so that it behaves the same as a
top-level import? Is there any way I can get this behavior? It seems
limiting to interactive development to have to have either a build
process or a script to import all libraries despite an explicit
dependency graph existing in the form of the import statements in the
library definitions?

I'm relatively new to Kawa and a little rusty in Java (long time
Schemer, though) so maybe I'm missing something.

-V

Re: Behavior of (import ...) in R7RS define-library forms

[spellcard199]

Hello. I'm far from being an expert (and not using Java nor Kawa atm)
but after trying to reproduce your project structure I think it's a
problem of how "kawa.import.path" is resolved.

2 things seem to work:

1) If you can, instead of -Dkawa.import.path="./src/*.scm" use an
absolute path (e.g.):
  -Dkawa.import.path="/home/user/development/project_name/src/*.scm"

2) If you must use a relative path:
   1. Use: -Dkawa.import.path="./*.scm"
   2. Start repl from inside the src directory

I just skimmed rapidly through it and I may be wrong but I think the
section "19.6.2 Searching for source files" in the Kawa manual expains
how this works.

‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
On Friday 5 February 2021 15:21, J. Vincent Toups via Kawa <kawa@sourceware.org> wrote:

> I have two r7rs-style libraries, one of which depends on the other,
> located in the following filesystem structure:
>
> src/lib/shadchen-helpers.scm
> src/lib/shadchen.scm
>
> In principle, shadchen-helpers isn't useful on its own - it provides
> some helper material for the shadchen library.
>
> The library definitions look like this:
>
> $ find src/ -iname "*.scm" | grep lib | xargs head -n 2
> ==> src/lib/shadchen-helpers.scm <==
> (define-library
> (lib shadchen-helpers)
>
> ==> src/lib/shadchen.scm <==
> (define-library
> (lib shadchen)
>
> So if I start a repl like this:
>
> $ java -jar lib/kawa.jar -Dkawa.import.path="./src/*.scm"
>
> Kawa can find the libraries if I do an import:
>
> #|kawa:1|# (import (lib shadchen-helpers))
>
> The above works fine. However, in a fresh interpreter:
>
> #|kawa:1|# (import (lib shadchen))
> /home/toups/work/games/jmonkey3d-lacraw/src/lib/shadchen.scm:9:4:
> unknown library (lib shadchen-helpers)
> /home/toups/work/games/jmonkey3d-lacraw/src/lib/shadchen.scm:110:6:
> warning - no declaration seen for pattern-bindings-equal?
> /home/toups/work/games/jmonkey3d-lacraw/src/lib/shadchen.scm:117:6:
> warning - no declaration seen for all-identifiers?
>
> The import fails because the library reference to shadchen-helpers
> fails to resolve. That reference in `shadchen.scm` looks like this:
>
> (import (except (kawa base) match)
> (kawa lib std_syntax)
> (srfi 28)
> (rnrs sorting (6))
> (rnrs hashtables (6))
> (lib shadchen-helpers))
>
> I have noticed that importing shadchen-helpers at the repl first and
> then importing shadchen works. I would guess because importing the
> shadchen-helpers library compiles that library and then the import
> form in the r7rs define-library can then find the compiled artifact.
>
> But doesn't it make sense for the import form in shadchen to trigger
> the compilation of shadchen-helpers so that it behaves the same as a
> top-level import? Is there any way I can get this behavior? It seems
> limiting to interactive development to have to have either a build
> process or a script to import all libraries despite an explicit
> dependency graph existing in the form of the import statements in the
> library definitions?
>
> I'm relatively new to Kawa and a little rusty in Java (long time
> Schemer, though) so maybe I'm missing something.
>
> -V