An autoformatter should be three things:

• Fast: Run quickly so that you can afford to run the autoformatter on a large code base without ever having to wait for the autoformat to complete.
• Consistent: Running the autoformatter a second time won't change the output. Code will run the same before and after the autoformatter is run. Trivial whitespace changes will be reverted.
• Pretty: Output looks nice and preferably follows PEP8.

PFA chooses to be Fast and Consistent, but sacrifices Pretty output in favor of speed.

To create the executable file pfa, run

make

Afterwards you can copy it into PATH, say to /usr/bin or ~/bin/.

There are two ways to run the program. If the executable file name does not end in "i", i.e. with

pfa that_python_script.py

then the formatted file will be dumped to standard output. If you make a symlink from pfa to pfai and run the latter like

pfai that_python_script.py scriptus_secundus.py

then all files listed as arguments will be formatted in place.

• Why is PFA written in C? The startup time for the Python interpreter is often longer than it takes to run pfa on a 2000 line file.

• How fast is it? The other popular Python formatters are yapf and autopep8. Formatting about 120KB of python code in place for the second time in a row with the following commands,

    yapf -i bx.py
    autopep8 -i bx.py
    pfai bx.py
  

  one finds that yapf takes 14.0 seconds; autopep8 takes 1.8 seconds; and pfai completes in 0.020 seconds, less than the time it takes to press Enter.

• I have a change to contribute. Will it be accepted?: Yes, as long as pfa still runs in O(n) for even pathological input.