I hold that it's far easier to read and step through Danny Yoo's BrainF interpreter than to try and piece together your one-line pythonic monstrosity.

Of course, if you're arguing that your Python version is somehow better than the longer idiomatic Racket implementation, I think we both agree that leveling the playing field would create a more fair comparison. Heavens, wouldn't it make more sense to compare a one-line BF interpreter written in Racket to your one-line BF interpreter written in Python? ;)

Two can play this game:

  #lang racket (require racket/match) (let* ([data "++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>."] [move-left (λ (left cur right) (let ([right (if (null? right) '(0) right)]) (list (cons cur left) (car right) (cdr right))))] [move-right (λ (left cur right) (let ([left (if (null? left) '(0) left)]) (list (cdr left) (car left) (cons cur right))))] [inc (λ (left cur right) (list left (add1 cur) right))] [dec (λ (left cur right) (list left (sub1 cur) right))] [consume-one-char (λ (char state) (match char [#\+ (apply inc state)] [#\- (apply dec state)] [#\< (apply move-left state)] [#\> (apply move-right state)] [#\. (begin (display (integer->char (cadr state))) state)] [#\, (list (car state) (char->integer (read-char)) (caddr state))] [else state]))]) (let loop ([counter 0] [prev-counters '()] [state (list '() 0 '())]) (if (= counter (string-length data)) state (let ([char (string-ref data counter)]) (match char [#\[ (loop (add1 counter) (cons counter prev-counters) state)] [#\] (if (zero? (cadr state)) (loop (add1 counter) (cdr prev-counters) state) (loop (car prev-counters) (cdr prev-counters) state))] [else (loop (add1 counter) prev-counters (consume-one-char char state))])))))

Both this example and your one-line python example are similarly incomprehensible. While we're on the topic of meaningless benchmarks, please allow me to point out that my racket version is 163 bytes shorter than your python version, without trying to abbreviate variable names and without smashing whitespace. That's even including the "Hello World" BF program near the beginning of the source.

Ok, yes, you have a point -- the article I cited earlier is long and focuses on Racket's awkward tooling. However, my point is that we're comparing apples to oranges here. It would be better to compare Danny's article to a Python article that implements a BF interpreter, covers some optimizations, and describes how to package it as an Egg and upload it to PyPi. None of these topics are trivial, especially to a beginner, hence Danny's thurough coverage.