mypython

A Python REPL the way I like it.

What? Why?

I was unsatisfied with all the existing Python interpreter options. None quite worked the way I liked them to, even after attempting to configure them (often a frustrating experience).

So instead I wrote my own. Turns out it’s really easy.

prompt-toolkit is a nice framework for creating custom REPLs. It took under an hour to create a basic Python REPL, and only a few days to add most of the features I like to it. REPLs created with prompt-toolkit are really nice. They support advanced features like syntax highlighting, multiline editing, and popup tab completion without any additional work (unlike REPLs based on readline).

Features

• Fancy emoji numbered prompts.
• Multiline input, which “just works”
• Keyboard shortcuts configured exactly the way I like them.
• History search (M-Up/Down Arrow) can be used to combine multiple things from the history.
• Up/down automatically go to the top/bottom of multiline inputs from the history. Up+Enter intelligently executes.
• Outputs saved as _1, _2, etc.. Previous outputs saved in _, __, and ___.
• Support for bracketed paste (pasting stuff in the terminal “just works” without the need for any %paste magic or a special “paste” mode).
• prompts are stripped automatically from pasted text.
• Emacs command editing keybindings.
• Automatic syntax highlighting.
• Automatic highlighting of errors via pyflakes.
• On syntax error, the cursor is moved to the error.
• Matching and mismatching parentheses highlighting.
• Tracebacks for stuff defined interactively show the code line.
• Tab completion using Jedi and dir().
• Per-terminal history.
• A nice theme (the same one I use in emacs, called “1am”, based on XCode’s “midnight”).
• stuff? shows the help for stuff. Works even if stuff is a complex expression. Does the right thing for NumPy ufuncs.
• stuff?? shows the source for stuff. Works even if stuff was defined interactively.
• %time and %timeit magic commands.
• %doctest mode to emulate standard Python REPL (for copy-paste purposes).
• %sympy magic (works like sympy.init_session().
• %pyinstrument magic to run code with pyinstrument.
• %line_profiler magic to run code with line_profiler.
• %pudb magic to run code in PuDB.
• Debugging functions defined interactively works.
• Integration as a custom shell for PuDB.
• SymPy objects automatically pretty print.
• Shell integration with iTerm2.
• GUI Matplotlib plots on macOS work correctly.

And some other stuff that I haven’t implemented yet.

Most of this stuff either comes for free from prompt-toolkit or was really easy to implement, in some cases by modifying some code from other libraries (ptpython, ipython, sympy, the Python standard library).

Installation

I haven’t packaged it yet. For now you can clone the repo and run ./bin/mypython.

It requires the following packages, which can be installed from conda-forge:

• prompt_toolkit
• pygments
• iterm2_tools (on Mac)
• parso
• jedi
• pyflakes
• setproctitle

As well as some optional dependencies that are needed for some %magic commands:

• pudb
• pyinstrument
• matplotlib
• sympy

You can install them using the requirements.txt file.

Configuration

There is none. It’s already configured the way I like it.

Features and bugs and stuff

I wrote this for me. You can request features, submit PRs, and report bugs, but be aware that I won’t accept any PR unless it’s a feature I want. I’m really not designing this to be used by other people.

Acknowledgments

Thanks to Jonathan Slenders for prompt-toolkit and ptpython (which I borrowed some of the more tricky things like Python multiline and Jedi completion from). Basically all the fancy stuff here is coming from prompt-toolkit. I just combined it together into a REPL that I like. Thanks to the IPython guys for figuring out the matplotlib eventloop stuff (which I could never do on my own), and for inspiring many of the features I’ve implemented.

License

MIT.

Licenses for code taken and modified from ptpython, IPython, prompt-toolkit, SymPy, and the Python standard library are at the top of the respective files.