ess-synchronize-dirs causes error

[BrendanHalpin]

ess-synchronize-dirs causes an error because (ess-command ess-getpwd-command) fails. That fails because Stata echoes the command in the *ess-get-words* buffer used by ess-command.

Two possibilities:
1: remove ess-synchronize-dirs from ess-idle-timer-functions and do without the functionality
2: edit ess-command to check comint-process-echoes and behave accordingly.

Option 1 is easy. Option 2 is better but more difficult, as the existing code seems to be written without taking the possibility of echoed input into account.

[lionel-]

I'm not in love with ess-synchronize-dirs running on a timer but it is important that ess-command works for stata so I think we should fix (2) even if we do end up doing (1) as well.

One possibility would be to debug and test echoed inputs with the R REPL. The echo global option is ignored in interactive session but we could create a simple repl_echo() R function for this purpose. This would be a great way of covering echoing REPLs in our unit tests.

Or maybe it's possible to configure stata not to echo inputs?

The other thing to consider is whether ess-get-words-from-vector should be (or include) a customisation point for specific REPLs.

[BrendanHalpin]

I've e-mailed Stata tech support to see if Stata can be told not to echo inputs (it behaves differently in "normal" terminals).

I agree that fixing ess-command would be best.

What seems to me the simplest way to fix it would be to add code (as exists elsewhere) that does "if comintp-process-echoes and looking-at input, delete input" at the right moment. But there might be better solutions than duplicating code.

[BrendanHalpin]

Problems with ess-synchronize-dirs are not (just) from ess-command. I have "fixed" the latter to delete the input, but ess-get-words-from-vector fails. This appears to be because Stata returns an unquoted string in response to (ess-command ess-getwd-command) whereas R and Julia return quoted strings.

[lionel-]

That was my guess, it looks like we need to make g-w-f-v generic.

[BrendanHalpin]

Stata Corp tell me it's the terminal's behaviour, not Stata's.

I think they may be mostly right, and the problem is that ESS isn't consistently honoring comint-process-echoes.

In an xterm, Stata doesn't duplicate the input.
In an Emacs shell buffer, it does, but if I set comint-process-echoes to t, it stops duplicating the output.
In ESS, it duplicates the input and doesn't respond to changing comint-process-echoes.

[lionel-]

Could it be that we just need to set comint-process-echoes to nil in the customize-alist?

Or maybe the same inferior initialisation as in R's iESS mode?

(setq-local comint-process-echoes (eql ess-eval-visibly t))

[BrendanHalpin]

I don't think it's going to be as simple as that. You've seen the other thread at #1 (comment).

[lionel-]

What happens if you add (setq-local comint-process-echoes (eql ess-eval-visibly t)) to stata-mode's iESS init?

[BrendanHalpin]

No change. comint-process-echoes remains nil (since ess-eval-visibly is nowait). If I explicitly set comint-process-echoes to t, there is still no change.