Yet Another LR Implementation

This is another implementation of the classic LR(1) parsing algorithm, along with Knuth’s LR(1) table generation algorithm.

Rather than follow the model of tools like yacc and require external code generation, this library is driven dynamically from Python code.

An obvious question is, “why?”

Two reasons:

• firstly, to experiment with surface syntax for expressing grammars in Python code in as uninvasive a way as possible.

  The package “ptk” takes a similar approach here.

• secondly, to provide a playground to look at other “cutting-edge” parsing algorithms, as a basis for improving my own understanding of them.

By “cutting-edge” I’m largely referring to the late 1970s.

The current implementation is largely inspired by yacc, including yacc’s various deficiencies:

• there’s a linearisation used for reduce/reduce conflict resolution;

• there’s a linearisation used for token precedence.

In practice, it’s easy to get bitten by both of these - as it is for the default shift/reduce resolution to mask errors. A more flexible partial ordering is certainly feasible.

Credits

Some of the example grammars included in the tests are derived from the (Red) Dragon Book, which I’ve had since I was about 16; it’s a book which definitely shaped my life.

I’ve no idea how the reference numbers line up with later additions.

Compilers: Principles, Techniques, and Tools (1986) by Alfred V. Aho, Ravi Sethi and Jeffrey D. Ullman