search

n-riesco / ijavascript

IJavascript is a javascript kernel for the Jupyter notebook
Other
2.19k stars 185 forks source link

feature request: more detailed kernel control #49

Open jacksonloper opened 9 years ago

jacksonloper commented 9 years ago

Interrupting the kernel appears to actually be equivalent to restarting the kernel. That is,

>> x=3;
>> sleep.sleep(5)
((interrupt)
>> x
ReferenceError: x is not defined

Also, the kernel doesn't appear to succesfully inform jupyter that the kernel is busy (by making the little circle in the upper-right into a filled circle).

n-riesco commented 9 years ago

Yes, currently the kernel can't be interrupted (the kernel is restarted instead).

At the moment, I don't know how to implement a kernel interruption. I could send a SIGINT signal, but the SIGINT handler won't be run until the running function is stopped.

Instead and ideally, what I would like is for IJavascript to be able to save and restore sessions (like the R shell does). Unfortunately Node.js only provides .save and .load to save and load the list of executed commands:

> .help
.break  Sometimes you get stuck, this gets you out
.clear  Alias for .break
.exit   Exit the repl
.help   Show repl options
.load   Load JS from a file into the REPL session
.save   Save all evaluated commands in this REPL session to a file

PS: I've created issue #52 to track the bug of kernel restart requests not publishing the kernel status.

n-riesco commented 9 years ago

An option (that depends on node-gyp): https://github.com/tjanczuk/tripwire https://github.com/nodejs/node/issues/1268

jacksonloper commented 9 years ago

Eek! I see this is much trickier than I had hoped. We need to find somebody willing to reopen nodejs/node#1268 and get it solved :). I fear it might involve very tricky changes to node core itself... ok... not happening soon.

If tripwire could work, that would be a great stopgap. Set the default timeout to 2 seconds, put $$tripwire$$ in the global namespace so that users can set the timeout to something else if they need to run a long piece of code. But I'm not sure if tripwire will actually allow you to recover from a tripwire exception. I raised a new issue around this question at tjanczuk/tripwire#8. It may well be that I'm not understanding something very fundamental though... :)