Love the Python REPL?
Hate that you can't write real software in a Jupyter Notebook?
Well do I have the project for you.

Pete and Repeat (pnrp) brings all the goodness of the REPL and Jupyter Notebooks to your regular old Python projects. It kind of works like Observable but for Python from the CLI.

To get started run,

 $ pip install pnrp
 $ python -m pnrp <my_script>.py

pnrp will run your script to completion just as the regular Python interpreter. Instead of exiting though, it'll monitor every line of code executed for changes. When a line of code changes, it will rerun ONLY the statements that have changed, and their dependent statements. We are able to do this by monitoring the files of any imported module, analyzing the AST of the code to be executed, and tracking dependencies between statements. This is of course no silver bullet, Python is a highly dynamic language and there are situations when this won't work. That is why we'll provide you an escape hatch to rerun from specific points in your script onward.

Suppose we have a python program

# Script at time t

data = open('some-big-file.txt', 'r').read()
print(len(data))

We then modify the script,

# Script at time t+1
import re

expr = re.compile('Pete (and)? repeat')

data = open('some-big-file.txt', 'r').read()
lines = data.split('\n')
print(len([x for x in lines if expr.match(x)]))

Rerunning the script normally would cause our machine to reread some-big-file.txt on each iteration. With pnrp, we intelligently look at your code to see that line 6 is unaltered so we cache the computation and only run the lines that were changed.

• Multi-file / import support
• AST Based statement parsing
• AST Dependency based rerun
• Explore sys.settrace for code dependencies
• Pretty printing
• Server / Client model ?? (to better support interactivity, interrupts, etc.)