Scheme / Chez Scheme : fix repl "(out)" repetition

[kevung]

I use Conjure with Chez Scheme as a Scheme repl : https://github.com/Olical/conjure/wiki/Quick-start:-Scheme-(stdio)#chez-scheme

When evaluating S-expressions, Petite / Chez Scheme repl output an additional "(out)" which breaks some Conjure features.

For example,

1. I open a simple Scheme file, and display the Repl, I get this

Notice the "(out)" printed twice.

2. I evaluate the form under the cursor, with <LocalLeader>ee, I get this

Again, "(out)" is printed at the end in repl
In comment, I should get 3.

3. I evaluate the form and replace with the result, with <LocalLeader>e!, I get this

I should get 3 instead of "; (out)" when evaluating "(+ 1 2)".

[russtoku]

Thanks for reporting this. Unfortunately, the Scheme client needs a bit of love. Originally written for MIT Scheme, it doesn’t handle the output from Chez or Chicken Scheme very well as you report.

I’ve been thinking about a good way to handle the differences. I wonder if it would be better to have separate client modules for Chez and other Scheme implementations like there is for Guile and Snd/S7.

[Olical]

I'm still thinking the best approach is a much smarter and more complicated wrapper / parser that can turn the stdio output into structured messages like nREPL (or actually nREPL!) that Conjure can consume. But that's not a quick fix.

…

On Tue, 10 Feb 2026 at 09:26, Russ Tokuyama ***@***.***> wrote: *russtoku* left a comment (Olical/conjure#758) <#758 (comment)> Thanks for reporting this. Unfortunately, the Scheme client needs a bit of love. It works well with MIT Scheme but doesn’t handle the output from Chez or Chicken Scheme very well as you report. I’ve been thinking about a good way to handle the differences. I wonder if it would be better to have a separate client modules for Chez and other Scheme implementations like there is for Guile. — Reply to this email directly, view it on GitHub <#758 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACM6XKRCUTLKHVYEG4OGLD4LGP3BAVCNFSM6AAAAACUOWADE2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTQNZWGQYDGNBVGU> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[russtoku]

I'm still thinking the best approach is a much smarter and more
complicated wrapper / parser that can turn the stdio output into structured
messages like nREPL (or actually nREPL!) that Conjure can consume.

Besides dealing with what might appear as random output, each Scheme implementation's REPL has its own replies while responding somewhat alike.

Currently, there are two Conjure clients that handle Scheme REPLs; client.scheme.stdio and client.guile.socket. Both handle the REPL responses by trying to parse the output. I'm hoping to see if there is a way to refactor things slightly so that the individual differences can be compartmentalized into functions that can be configured as needed. Then it should allow tailoring things without a lot of complected logic.

[russtoku]

Just some observations:

• Chicken Scheme's REPL doesn't work with the Scheme stdio client. Sending forms to the REPL in a subprocess doesn't return any response from the REPL.
• Guile Scheme REPL in a subprocess works with the Scheme stdio client but the parsed output is not so good.

[russtoku]

When run in a subprocess, Chicken Scheme's REPL doesn't print out a prompt. That's why the Scheme stdio client doesn't work. It doesn't know when the REPL is done sending its response.

[russtoku]

https://wiki.call-cc.org/emacs#builtin-scheme-support says to run Chicken Scheme this way:

"csi -:c"

I verified that it works with the Scheme stdio client.