oppen-pretty-printer

This project is a faithful translation, to python, of the pretty-printing algorithm developed by DEREK C. OPPEN at Stanford University in the late 1970s. The appendix of PRETTY PRINTING, Derek C. Oppen, 1979 contains the original prettyprint program written in MESA

A much quoted, more recent algorithm, A prettier printer, Philip Wadler 2003 refers to Oppen, saying, "The pretty printer presented here uses an algorithm equivalent to Oppen’s, but presented in a functional rather than an imperative style". Many of todays pretty printer implementations are based on the Wadler algorithm. So, in a way, the road leads back to Derek C. Oppen, 1979.

Two python versions are included. pretty_oppen.py is, as far as possible, a line-by-line translation of the Oppen algorithm from MESA to Python. pretty.py is functionally identical to pretty_oppen.py but written using a more Pythonic idiom.

Pytest tests and example output are included for both versions.

Usage

pip install oppen-pretty-printer

Example:

from oppen_pretty_printer import pprint, Tokens as T

tokens = [
    T.BEGIN(),
    T.STRING("begin"),T.BREAK(),
    T.STRING("x"),
    T.BREAK(), T.STRING(":="), T.BREAK(),
    T.STRING("40"),
    T.BREAK(), T.STRING("+"), T.BREAK(),
    T.STRING("2"),T.BREAK(),
    T.STRING("end"),
    T.END(),
    T.EOF()]

pprint(tokens)

Result:

"begin x := 40 + 2 end"

For more complex examples see the tests folder.

Overview

The pretty printer processes a sequence of the tokens BEGIN, END, STRING, BREAK, EOF. The tokens BEGIN..END define a block that can contain any number of STRING and BREAK tokens together with nested BEGIN..END sub-blocks. The pretty printer attempts to keep the content of blocks together on the same line. If this is not possible then the block is automatically folded and indented at a suitable BREAK.

The STRING token holds a sequence of printable characters that will never be broken. The BRAKE token is a replacement for a sequence of non-printing, white-space, characters. A BRAKE token is nominally printed as a single-space or, if folding takes place, new-line-with-indent-and-offset.

This example from Oppen's paper shows how the pretty printer deals with the same input for different line widths.

width=75

        begin
          x := f(x); y := f(y); z := f(z); w := f(w);
          end;

width=24

        begin
          x := f(x); y := f(y);
          z := f(z); w := f(w);
          end;

width=75, BreakType.consistent

        begin
          x := f(x);
          y := f(y);
          z := f(z);
          w := f(w);
          end;

BLOCK and BREAK tokens can be parameterized, if needed, to modify the default folding behaviour. The parameters are used to control indent, offset and the consistency of folding.

To create a pretty printer for a new language grammar it is normal to write a tree visitor using the visitor design-pattern. The visitor emits pretty-printer tokens as is traverses the program AST. The token sequence is passed to the pretty printer for printing.

See odata-pretty-printer for an example of traversing an AST and emitting oppen-pretty-printer tokens. For other examples of usage see the pytest examples in this projects the ./tests folder.

Development

pip install -r requirements.txt

pytest

To run pytest on pretty_oppen.py use:

pytest --oppen

publish:

invoke publish 0.1.2 --pypi

Links: (copies in the docs folder)

• PRETTY PRINTING, Derek C. Oppen, 1979
• A prettier printer, Philip Wadler 2003
• MESA