JunoNewParser.mesa (was Parser.mesa)
Coded September 9, 1982 5:57 pm
Last edited January 10, 1983 5:50 pm
Last Edited by: Gnelson, January 10, 1983 5:50 pm
Last Edited by: Stolfi June 11, 1984 9:16:52 pm PDT

SUPERSEDED BY JunoParseUnparse. Stolfi, June 11, 1984 8:22:46 pm PDT

Simple operator precedence parser for operators of the following types: infix, prefix, postfix, matchfix (like "begin" or left grouping parenthesis), subfix (like "[" in Mesa or "(" in f(x)) and closefix (like "end" or ")").

The symbolic expression constructed by the parser is represented as in Lisp, by a nest of ordered pairs; the list cells employed are those of Mesa's LIST OF REF ANY.

DIRECTORY List, Atom, IO, Rope, JunoLexer, JunoParseTable;

JunoNewParser: DEFINITIONS

= BEGIN OPEN PT: JunoParseTable, Lexer: JunoLexer;

Handle: TYPE = REF HandleRec;

HandleRec: TYPE = RECORD
[error: Rope.ROPE, -- initially NIL, set to error message on lexical error
 eof: BOOL, -- initially FALSE, set to TRUE on end of input.
 result: REF ANY, -- contains last thing produced by Parse.
 openCount: INT, -- number of unmatched matchfixs or subfixs in last thing parsed
 in: Lexer.Handle, -- Source of lexemes to be parsed.
-- WARNING The initial value of in.a is relevant; see "Parse" below
table: PT.Handle]; -- associates operator types with operators.

NewHandle: PROC RETURNS [h: Handle];

Parse: PUBLIC PROC[h: Handle];
sets h.result to the List representing the longest prefix of
-- the lexeme stream h.in that is a valid symbolic expression,
-- counting the current value of h.in.a as the first lexeme. Thus
-- the way to start the parser is to call Lex[h.in]; Parse[h].
-- This will parse one expression and leave in h.in.a the first lexeme
-- not used in the parse tree; thus successive calls will work.
-- Sets h.eof if h.in.eof is set
-- Sets h.error if h.in.error is set, or if a parse error occurs.
-- If h.error AND ~ h.in.error AND ~h.in.eof then the input can be reconstructed
-- by Unparse[h.result]; print h.in.a; print all of h.in, except that
-- in the unparse you should skip the NIL that appears at the right-most
-- position in h.result.

END.