[Feature request] Feature parity with IPython

[malmaud]

IJulia is truly awesome, but it's still missing some of the nice things that IPython has for making interactive computing nicer. I thought I'd enumerate some of them here and get a sense for if people think these are important to have or are too low-priority to work on now (or ever) or require substantial changes in Julia itself to implement.

• The ? operator

? on a variable produces richer information than Julia's current ?, include its nicely formatted current value and other metadata:

In [3]: x=[1,2]
In [4]: ?x
Type:       list
String Form:[1, 2]
Length:     2
Docstring:
list() -> new empty list
list(iterable) -> new list initialized from iterable's items

• ?? operator for seeing the source of a function

In [5]: def f(x):
   ...:     y=x+1
   ...:     return y
   ...: 

In [6]: ??f
Type:       function
String Form:<function f at 0x10317baa0>
File:       /Users/malmaud/tmp/<ipython-input-5-9d6105d7ff7a>
Definition: f(x)
Source:
def f(x):
    y=x+1
    return y

• pylab mode

Automatically import commonly-used scientific packages to the global namespace. In our case, that would probably mean

using HDF5
using JLD # for @save and @load
using PyPlot

and perhaps also Distributions and nlopt, which roughly accord with scipy functionality.

• Redefinition of composite types in the global module

Currently it's convenient to define types in a Julia notebook, since it has to be wrapped in a module or will throw an error when the cell is re-evaluated.

And commonly-used magics:

• autoreload mode

Every time a command is executed as the REPL, first reload all (or a specified subset) of imported modules from disk. That makes the style of development where you are editing a julia package in a standard editor while testing it in the REPL more convenient.

• Command-line functions
  cd, ls, etc. typed at the REPL operate as run(ls).

[stevengj]

• Since ? just uses Julia's help, improvement should be requested as a Julia issue, not as IJulia. Better help ultimately depends on associating data with functions, modules, and globals julia#3988 being resolved.
• Similarly, implemeting "show source code" functionality with ?? means resolving store the actual source code of a method julia#2625 first.
• What would a "pylab mode" do that using PyPlot does not? We already have inline plots.
• Automatic importing should really be in a .juliarc file or similar, so that it is up to the user; I'm reluctant to hard-code this in IJulia. In any case, for performance reasons it may need to wait for WIP: static compile julia julia#3892 and Improving package load times  julia#4373
• Profblems redefining types are a Julia issue, not IJulia.
• For developing modules, unfortunately a kernel restart is often required after changes, e.g. due to Reloading functions, but not constants, from packages? julia#4008, so I'm skeptical of the utility of an autoreload mode at this point.
• You can already do shell commands at the REPL by prefixing them with ;. e.g. ; ls.

[malmaud]

Thanks for the reply - looks like it's really Julia enhancements that I'm after.

[stevengj]

Note that IJulia now reads .juliarc.jl at startup (#64) if you want to automatically import some packages.

(The only problem is that this make it hard for other people to use your notebooks if you forget to add using statements to the notebook.)