Difference between STDIN and FILE

[ceving]

This is ack.scm:

(define (ack m n)
  (cond ((= m 0) (+ n 1))
        ((= n 0) (ack (- m 1) 1))
        (else (ack (- m 1) (ack m (- n 1))))))

(display (= (ack 3 12) 32765))
(newline)

scheme-rs stackoverflows if the input is STDIN (might be ok):

$ time bash -c "scheme-rs < ack.scm"
 
thread '<unknown>' (9199) has overflowed its stack
fatal runtime error: stack overflow, aborting
bash: Zeile 1:  9198 Abgebrochen             scheme-rs < ack.scm
 
real    3m1,355s
user    4m38,267s
sys     0m44,885s

But throws an error if the input is a FILE:

$ time bash -c "scheme-rs ack.scm"
Error: Exception(#<compound-condition #<&syntax form: #<syntax:ack.scm:1:0 (define (ack m n) (cond ((= m 0) (+ n 1)) ((= n 0) (ack (- m 1) 1)) (else (ack (- m 1) (ack m (- n 1))))))> subform: Some(#<syntax:ack.scm:1:0 define>)> #<&message "undefined variable">>)
 
real    0m0,010s
user    0m0,007s
sys     0m0,004s

$ time bash -c "scheme-rs <(cat ack.scm)"
Error: Exception(#<compound-condition #<&syntax form: #<syntax:63:1:0 (define (ack m n) (cond ((= m 0) (+ n 1)) ((= n 0) (ack (- m 1) 1)) (else (ack (- m 1) (ack m (- n 1))))))> subform: Some(#<syntax:63:1:0 define>)> #<&message "undefined variable">>)
 
real    0m0,010s
user    0m0,010s
sys     0m0,000s

[xnacly]

With the changes in #228 I can make out the define seems to be the issue:

Uncaught exception:
 - Message: undefined variable

--> test.scm:1:0:
001 | (define (ack m n)
    | ~ here
002 |   (cond ((= m 0) (+ n 1))
003 |         ((= n 0) (ack (- m 1) 1))
004 |         (else (ack (- m 1) (ack m (- n 1))))))

Prefixing your snippet with (import (rnrs)):

+ (import (rnrs))
  (define (ack m n)
    (cond ((= m 0) (+ n 1))
          ((= n 0) (ack (- m 1) 1))
          (else (ack (- m 1) (ack m (- n 1))))))
  
  (display (= (ack 3 12) 32765))
  (newline)

Which is automatically imported in main (only for repls tho):

scheme-rs/src/main.rs

Lines 69 to 70 in 7bcf36a

	maybe_await!(repl.import("(library (rnrs))".parse().unwrap()))
	.expect("Failed to import standard library");

Makes it run for ever on my hardware. Overflows on my hardware too:

thread '<unknown>' (86691) has overflowed its stack
fatal runtime error: stack overflow, aborting
fish: Job 1, 'cargo run -- test.scm' terminated by signal SIGABRT (Abort)

The repl imports rnrs, while the file execution context doesnt, maybe this needs to be changed to either both importing it or none.

[maplant]

This is expected behavior, although perhaps slightly obtuse and not exactly conformant in terms of what error is presented.
scheme-rs interprets files from the command line as top-level programs, which have the following form:

 A top-level program is a delimited piece of text, typically a file, that has the following form:
<import form> <top-level body>

An <import form> has the following form:
(import <import spec> ...)

A <top-level body> has the following form: 

You are not importing a library, and therefore scheme-rs has no imported symbols for your program.
In R6RS, define, cond, =, -, display, and newline are not intrinsic parts of the language as they are in other languages. Instead, they are namespaced to libraries. This means you have to import (rnrs) to use them.

The REPL automatically imports this library because it would be pretty annoying to have to import it every time. This is behavior that's seen in other scheme implementations, such as Chez Scheme.

The issue with the stack overflow is kind of expected but probably could be converted to an OOM. There's probably a DFS in the garbage collector that needs to be converted from recursive to iterative.

I'll leave this issue open for a while until there's better error messages.

[ceving]

This is behavior that's seen in other scheme implementations, such as Chez Scheme.

Chez recognizes define regardless of whether the program comes from STDIN or from a file.

$ time bash -c "scheme < ack.scm"
Chez Scheme Version 10.0.0
Copyright 1984-2024 Cisco Systems, Inc.
 
> > #t> 
> 
 
real    0m1,164s
user    0m1,134s
sys     0m0,018s

$ time bash -c "scheme --script ack.scm"
#t
 
real    0m1,153s
user    0m1,112s
sys     0m0,040s

It's the same with Racket and Guile.

According to R7RN appendix A define is in (scheme base) together with cond and else. Does it make any sense to omit the import of "(scheme base)"?

[maplant]

This is behavior that's seen in other scheme implementations, such as Chez Scheme.

Chez recognizes define regardless of whether the program comes from STDIN or from a file.

$ time bash -c "scheme < ack.scm"
Chez Scheme Version 10.0.0
Copyright 1984-2024 Cisco Systems, Inc.
 
> > #t> 
> 
 
real    0m1,164s
user    0m1,134s
sys     0m0,018s

$ time bash -c "scheme --script ack.scm"
#t
 
real    0m1,153s
user    0m1,112s
sys     0m0,040s

It's the same with Racket and Guile.

In order for Chez to interpret your program as an RNRS top-level program (rather than as a "Scheme shell script"), you need to pass the "--program" option:

map@fedora:~$ scheme --program ack.scm
Exception: invalid top-level program import subform (define (ack m n) (cond ((...) (...)) ((...) (...)) (else (...)))) at line 1, char 1 of ack.scm

According to R7RN appendix A define is in (scheme base) together with cond and else. Does it make any sense to omit the import of "(scheme base)"?

scheme-rs is R6RS, not R7RS, but I think it's reasonable to expect that that if all imports are omitted then (rnrs) is imported.