Can't use redbaron in IPython notebook about redbaron

[Psycojoker]

Hello,

There is great chances that this is actually already fixed but unreleased (see: #44). Can you try using RedBaron master to see if this works please?

I'll do a bugfix release if it's the case.

Thanks for reporting :)

[remram44]

I can reproduce the bug on master; zmq's sys.stdout does have fileno(), it just raises when called.

[remram44]

This should be a simple matter of catching UnsupportedOperation, however I do not understand why the isatty() check is here to begin with. __repr__() works differently depending on what sys.stdout is, regardless of whether __repr__() is going there or not?

[Psycojoker]

Yes, in a ipython shell, repr displays the selected code (highlighted with pygment) like BeautifulSoup displays you the selected html. But when running a script you might not know where you are and what you have selected at this point so we display a more common repr (much like any python object).

I'm going to look at it in some minutes.

[remram44]

The problem with that approach is that you have no accurate way of knowing if the result of __repr__ is eventually meant to be printed in the terminal or not. In IPython, the kernel is not attached to a terminal, and yet the result is sent through a ZMQ socket and printed. In contrast, a non-interactive function might still happen in a process which is still connected to a terminal.

The current approach already does weird stuff, for example:

>>> tree.help
<bound method NodeList.help of 0   self.get_input('some_port')
1   u'\n'
2   a = self.force_get_input('something', 2)
3   u'\n'
4   if False:
        self.set_output('port_name', False)

>

AFAIK beautifulsoup doesn't try anything funny here either (link).

I'm not sure what to recommend though because __str__()'s output is indeed useful 😢

[Psycojoker]

Yes, BeautifulSoup always print the html code which is fine. In our case we sometime ends up printing formatting information like white spaces which turns out to be extremely confusing for the user.

So, after one quick google search "how to detect that you are in a python shell" we ends up with this situation that doesn't seems very optimal :/

Right now I don't know have an opinion on the best approach to have. I'll start by fixing it for ipython and catching other exceptions to fall back on a default behavior.

[Psycojoker]

After playing with "def repr_html" of the notebook this end up looking like this:

Should be way more usable. It's pushed on master.

Tell me if you find any more bugs ;)

[Psycojoker]

Closing this issue for now, don't hesitate to complain/reopen it/open other issues if needed.

[remram44]

This is great, thanks 😃