Difference between Output and Returned Values

[yewscion]

Hello,

Is there a way to differentiate between values sent to stdout via a BQN program (using Show or similar) and values returned to the user at the REPL? If not, could such a functionality be added as a flag or similar?

I am most of the way through implementing Org Babel support for BQN, but my current workaround for capturing output breaks if the final command given in a block is an output that contains a newline, because CBQN will repeat the last return value at the REPL.

This is fine and makes sense, but if it could be suppressed with a special flag or marked as being a value instead of output in some way, then implementing the difference between output and value in Org Babel (and similar systems that might only care about one or the other) would be made much easier.

[dzaima]

There's currently no way, and adding some would be quite complicated over a single text pipe. If you want to not get REPL output, you can prepend )r to the source (i.e. )escaped ")r •Out 'a'+↕5"), but it won't help you if you want both.

Half-relatedly, making the REPL not have implicit output if the expression ends with a call to •Show or •Out has been on my to-do list for a while.

Some more programmatic REPL interface is also a thing that would be useful (including things like inspectable stacktraces, actual info about when a thing's done running, escaped messages about each output call), but is non-trivial amounts of work.

If CBQNs •ReBQN supported system value redefinition, redefining •Show & •Out to include info about their length would be an option, but alas CBQNs •ReBQN is incomplete too.

(btw, you want to escape " to \" in )escaped (and \ to \\ if you're not doing that already); it only works currently because I was lazy with my error checking)

[yewscion]

)r was exactly what I needed, actually. If there is no output, there's no reason to strip the value. If there is, )r will strip the return value. Sorry for the simple question; I must have missed the definition in docs/.

As for the escapes, I've implemented the ones You've mentioned, but escaping (for instance) \n as \\n prevents CBQN from interpreting that as a newline. Is this intended behavior? Should I remove the blanket escape of \ in favor of one that does not escape \n, \t, etc?

I'll need to brush up on C before I can look at the code for CBQN, but once I've finished that and learning BQN itself, maybe I can contribute to some of the goals above. Either way, thank You for this awesome project and Your quick responses to questions!

[dzaima]

The backslash escaping is the same as C/JS/Python/pretty much every traditional language, i.e. first you replace \ with \\, and only then replace "→\" and newline → \n. So

("hello \backslash
world")

is (\"hello \\backslash\nworld\")

(and the only escapes needed are for backslashes, quotes, and newlines (both \n and \r))

[yewscion]

Copy that, makes sense.

I was mostly asking for the edge case of a string having a newline \n inside of it before the escapes are applied. Right now, as in my image above, such a string will have that \n escaped to \\n, and therefore the output will have a literal \n string instead of a newline. This will prevent formatting output using the \n escape string in this interface, as it will always appear literally in the output.

I suppose excluding \n and \r from the escapes would then allow the output to behave as a user might expect (expanding into a literal newline), but if that's an error maybe I can manually apply the expansion afterwards instead.

[dzaima]

BQN's string literals themselves don't parse backslash escapes at all. "a\nb" in BQN is a 4-character string, and interfaces should keep it as such. (there is double-quote escaping by typing it twice, i.e. "a""b" is a 3-char string, but that's unrelated)

[yewscion]

Copy that, will do that then. Thanks!