Create `sage-shell:restart-sage`

[johanrosenkilde]

I quite often restart the Sage shell. Sometimes its to flush all the globals I've created during experiments to double-check that my functions don't rely on them, and sometimes it's because I accidentally evaluate an infinite loop or something.

For this use case it would be nice with a handy restart function.

[johanrosenkilde]

I tried my hand at creating this function, but it relies very nastily on the time it takes the existing process to finish:

(defun sage-restart (cmd)
  (interactive (list (sage-shell:read-command)))
  (with-current-buffer sage-shell:process-buffer
    (end-of-buffer)
    (ignore-errors
      (progn
        (comint-delchar-or-maybe-eof nil)
        (sleep-for 3)))
    (sage-shell:run cmd nil)
  ))

Note the ugly (sleep-for 3): If that's not there, then sage-shell:run doesn't seem to do anything because it hasn't discovered that the existing process has ended yet.

The advantage of reusing the existing buffer is that history in the buffer is preserved, so you can go back and read what went on previously. The disadvantage is that I don't know how to get back from the process when it has ended, replacing the sleep-for by something robust.

[stakemori]

The advantage of reusing the existing buffer is that history in the buffer is preserved, so you can go back and read what went on previously. The disadvantage is that I don't know how to get back from the process when it has ended, replacing the sleep-for by something robust.

As for the input history, you can set sage-shell:input-history-cache-file. You can reuse input history even if Emacs exits.

For such an asynchronous job, you can use deferred or the macro sage-shell:as-soon-as which needs lexical-binding and is not documented (sorry). Or maybe I should add a hook which runs when the proces exits.

[johanrosenkilde]

As for the input history, you can set sage-shell:input-history-cache-file. You can reuse input history even if Emacs exits.

Yes, but you lose the output history. And the input history is not conveniently listed so you can search backward through it.

For such an asynchronous job, you can use deferred or the macro sage-shell:as-soon-as which needs lexical-binding and is not documented (sorry). Or maybe I should add a hook which runs when the proces exits.

That sounds interesting. I'm afraid I won't have time to look further on this for some days. It sounds like the hook would be an Emacsy and convenient way of accomplishing this.

[stakemori]

I added sage-shell:process-exit-hook and sage-shell:restart-sage in fix-branch. Thanks.

[johanrosenkilde]

Thanks!

[johanrosenkilde]

sage--shell:restart-sage works correctly and with the nice behaviour of staying in the same buffer. But it works only if *Sage* is the current buffer? Perhaps a more convenient semantics is:

• If the current buffer is a Sage buffer, restart that.
• If the current buffer is a sage-shell:sage-mode buffer, restart the connected Sage process.
• Otherwise restart the default Sage buffer.

[stakemori]

Thanks for your suggestion. I modified sage-shell:restart-sage (15ee1e6c2bec83e01b73e1b7e6462831e8e9c5d9).

[johanrosenkilde]

Thanks! I've verified that it works

[johanrosenkilde]

New problem: sage-shell:restart-sage forgets to store the input of the session in the history cache file, even when this is non-nil.

[stakemori]

Thanks. I pushed further fix. I will test commits more carefully.