Learning about Pratt parsing. (A similar, if not the same, technique is used to parse Mercury terms, and is presumably common to Prolog systems.)
The toy grammar implemented is:
e --> NAME
; INTEGER
; name = e
; e ? e : e
; prefixop e
; e postfixop
; e infixop e
; e ( args )
; ( e )
args --> zero or more comma-separated e
prefixop --> + ; - ; !
postfixop --> !
infixop --> + ; - ; * ; / ; ^ ; < ; >
where < and > are non-associative.
This is a better introduction than most: Pratt Parsers: Expression Parsing Made Easy
Peter Wang [email protected]