dfbrowse

Python library that provides spreadsheet-style browsing of a Pandas dataframe within a terminal/curses environment.

It was borne out of my love for curses-style/modal interfaces, and a frustration with how difficult it is to really navigate/browse/inspect a large dataframe while you're operating on it within IPython.

Installation

pip install dfbrowse

Getting started

dfbrowse is generally intended to be used inside IPython. Start by launching IPython, then import the IPython magics:

import dfbrowse.ipython_main

If you want to jump straight into browsing, load a parquet or CSV directly:

%csv ~/work/project1/employees.csv emp
%pq ~/work/project2/my_data.parquet data

If you've already loaded a dataframe, browse it like so:

import pandas as pd

emp = pd.read_parquet('~/work/project2/my_data.parquet')

%show emp

Inside the browser

Once inside the browser, you can navigate with arrow keys and the mouse.

Press ? to find out about some basic browsing commands. You can resize columns, hide them, reorder them, and even search the row contents, with a few keystrokes.

There are many other commands that you can discover in keybindings.py.

There are even more commands that can be discovered by typing / and then pressing TAB. These are pre-registered functions, such as str_match and query, that perform some operation on a dataframe. The returned dataframe will be rendered in the browser.

In theory it is possible to write and register your own dataframe functions as well, but this is a topic for a much larger document and I'm not sure how well it's working anyway. If you want to do complex transforms on your dataframe, you're probably better off to drop back into the IPython 'shell', do your work, and then jump back into the browser with one of the options above.

quitting and reentering

You can use q on your keyboard to go back to the shell/ipython session.

Re-enter the browser via %show <df_name>, or %fg to foreground the most recent dataframe.

bugs

There are lots. I even know about some of them. Please feel free to report any you find on GitHub.