#342: Try to recover as intelligently as possible if user calls magic(). #352

wants to merge 3 commits into


None yet

2 participants


This was annoying me from the perspective of force-of-habit typing quit() (system install at the lab of IPython 0.8.3 (!!!!!) requires it) but in general, people typing some_magic() with automagic on and expecting results might be awfully surprised by the NameError that results.

This attempts to address the problem in the following way:

  • Check if whatever precedes the first paren in line_info.ifun is a magic, and of course is not shadowed by something in the user namespace
  • Throw out everything after the first paren, as well as the contents of line_info.the_rest (trying to sensibly convert function call syntax to magic-calling syntax in a general way doesn't seem prudent or possible)
  • Visually rewrite as ----> %some_magic so that the user has some clue that what he typed was interpreted as a magic but that the subsequent parenthesized arguments were ignored.

The result looks something like this:

In [6]: quit(5, 4, 3)
------> %quit

Ideally then, individual magics, if they require arguments, will complain about not receiving them, and in the process demonstrate their usage with the correct %foo bar baz way of calling them.

This pull request replaces #345.

IPython member

How is this pull related to #345?


How is this pull related to #345?

It replaces it and solves the problem more generally. I meant to close the old pull request and in fact clicked the button to do so but was apparently the victim of a network hiccup. The Canadian inter-tubes are still thawing out from the winter, I guess...

IPython member

OK, looking at the code I think I understand better. But I'm not convinced we should add more input handling logic of this kind. Basically, I think if someone types foo(), then they mean a python call, not a magic. Granted, it may take a little bit to realize that ipython offers two overlapping systems, one for "commands" (magics) and one for pure python code, but anyone using a tool like ipython should figure that out pretty quickly.

Now, as for #342, I think we can address that one slightly differently, but I need to play with some ideas.

But I'm reluctant to add even more text input special-casing, as much of our code on that front is already fairly brittle. What do you think?

IPython member

BTW, I'm on IRC (freenode, #ipython) if you want to chat live on these two for a moment... That may be more efficient.


I'm not entirely sure, either. @tk suggested that magic() ought to map to %magic for more than just %quit and %exit, and I went further with magic(...) mapping to %magic, but that last step might be too far.

I'll hop on IRC, just need to download a client on this machine. :)


After discussing it with Fernando it was decided this was overkill, and #352 will be reopened.

@dwf dwf closed this Apr 10, 2011
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment