inf-clojure--make-single-line removes consecutive spaces in strings.

[austinhaas]

https://github.com/clojure-emacs/inf-clojure/blob/master/inf-clojure.el#L321

(count "1 5") => 3 ;; Should be 5.

EDIT: There are 3 spaces between 1 and 5 above. Apparently, this markdown is also removing the consecutive spaces.

[arichiardi]

Regexes are bad, I know, I did not cover that use. There are tests though, so a patch + tests should be easy to come up with (I guess?).

[hlolli]

The newlines are also removed from strings

=> "a   b"
"a b"
panaeolus.engine=> "a   
      
      b"
"a     b"
panaeolus.engine=> 

[sogaiu]

For inf-clojure-eval-region, this appears to have adverse affects when working with Arcadia (stalling of some sort).

Calling inf-clojure-eval-region seems to lead to the following chain of calls:

inf-clojure--send-string ->
inf-clojure--sanitize-command ->
inf-clojure--make-single-line

Is it necessary to modify a potentially multi-line region of text to be a single line for inf-clojure-eval-region?

I've been trying a modification where the original string from the region just has a newline appended to it and so far at least the stalling behavior seems better. I'm uncertain of whether the changes are appropriate.

[arichiardi]

Thanks for taking the time for looking into it. Yes newlines evaluate the command in comint so yes, some post processing must be done.

Having said that, it is far from perfect now so patches will be reviewed 😄 The reason is comint and that text-based REPLs are super tricky.

[sogaiu]

I confess to not understanding the explanation :) Are there specific bits of comint (or some document) I might look at to better understand this statement:

newlines evaluate the command in comint

?

I looked through comint.el for instances of newline, but didn't come away feeling any more informed...

[sogaiu]

@arichiardi, #40 seems related to this.

Therein, I see a quote from you:

that change would not work: that function has been added because we cannot send \ns or 
comint will evaluate the string. Each \n evaluates.

I presume that is what you were referring to by:

newlines evaluate the command in comint

There is also this statement:

The rationale of having a single line with no comment was because I could not send 
newlines without evaluating...

For regions, it doesn't appear impossible to send those that contain newlines and have them be evaluated, at least it seems to work with Arcadia. It seems to work when a region is selected from a non-repl buffer and that's sent (w/ a newline appended).

May be I'm not understanding the statement about "newlines evaluating" appropriately. Would you mind elaborating / spelling out some more of the details?

[arichiardi]

@sogaiu I think you understand it right.

Comint is character based and it evaluates as soon as it seems at \n if I remember correctly.

I have briefly checked comint.el and it seems that you can find a very nice explanation in comint-send-input.

Again, any improvement over this is greatly greatly appreciated as I don't have time myself for it at the moment.

[sogaiu]

Thanks for the hints.

Is it possible you mean that pressing the Enter / Return key triggers comint-send-input? Or do you mean something else?

[arichiardi]

It might be, I have very blurred memories about how that all works, your best bet is to probably try directly with ielm. Are you familiar with elisp a bit? If so it should be an easy one.

[sogaiu]

FWIW, I verified with Wireshark that pasting a region into shell mode (connecting to a socket REPL with nc) does not lead to evaluation. I waited some minutes before pressing Enter, which did lead to the entire region being sent by comint-send-input.

Thanks for the tip about ielm -- I am not familiar with it, but will investigate.

[arichiardi]

In that case probably then we wouldn't need to get rid of the \n in the whole string.

[sogaiu]

What would break if inf-clojure--send-string sends string + "\n" via comint-send-string instead of sanitized?

[sogaiu]

With that change, the following:

austinhaas' count - #152 (comment)
hlolli's newline example - #152 (comment)

appear to yield the correct results here.

inf-clojure-eval-region also appears to work so far.

On a side note, the long string example in the original posting of #40 also works properly here -- though this I doubt the environment (Arcadia / Clojure CLR) I'm testing in is the same as that of the original poster.

[arichiardi]

We could try and see how many people report breakage? I seem to remember lumo was breaking but maybe with a patch I can double check if it still true. Definitely a patch to try out would be great.