TP Lex and Yacc Version 3.0 as of April 1991, Version 3.0a as of May 1992
Copyright (c) 1990-92 by Albert Graef, Schillerstrasse 18, 6509 Schornsheim,
Germany (ag@muwiinfa.geschichte.uni-mainz.de)
All rights reserved

This disk contains TP Lex and Yacc, compiler generator tools for Borland's
Turbo Pascal (4.0 or later), running under MS-DOS (3.0 or later).

(Turbo Pascal is a trademark of Borland International.)

The TP Lex and Yacc software is licensed free of charge and may be copied
and redistributed for non-profit purposes. There is no warranty of any kind.
In particular, I do not guarantee that the software is free of bugs, or fits
some particular purpose, and I take no responsibility for damages or any
other consequences of its use.

The manual is contained in the file TPLY.DOC on the disk. It can be printed
using the DOS command `copy tply.doc prn'. (A TeX'ed version of the manual
is contained in the file TPLY.TEX.) The manual provides a rather terse
description of TP Lex and Yacc and you should follow the references given
there for further reading.

As distributed, the TP Lex and Yacc library units (LEXLIB.TPU and
YACCLIB.TPU) are compiled with Turbo Pascal 6.0; to use them with other
Turbo Pascal versions you probably have to recompile these units (sources
LEXLIB.PAS and YACCLIB.PAS on the distribution disk).

Older versions of TP Lex and Yacc are no longer supported. Most programs
written for TP Lex and Yacc 2.0 should not to be too difficult to adapt
to the new version. The major differences are:

- The TP Lex library routine `start' is now used to set the start state of
  the lexical analyzer (in V2.0, this routine was named `begin_').

- TP Lex no longer supports character tables (%T).

- TP Lex now provides an optimization option which causes TP Lex to generate
  minimal DFA's, and a verbose option to generate a readable description
  (.LST file) of the constructed DFA.

- The TP Yacc error message printing routine is now named `yyerror' (as in
  the UNIX version); in V2.0 it was named `yymsg'. The V2.0 routine named
  `yyerror' (which triggered error recovery in the parser) is now named
  `yyerrlab'.

- TP Yacc no longer supports %union definitions; instead, value types are
  specified directly using tags of the form <type-identifier> in %token
  and %type definitions. TP Yacc automatically generates a suitable
  variant record type (YYSType) from the type definitions in the grammar.

- TP Yacc no longer generates a separate header (.H) file; all definitions
  for the lexical analyzer (token numbers, yylval variable) are put into
  the .PAS output file. Hence the lexical analyzer should be included right
  behind the parser (e.g., in the auxiliary procedures section of the Yacc
  grammar); it cannot be put into a separate unit any more. Although this
  might seem a little restrictive, I decided to use this scheme in order to
  facilitate communication between parser and lexical analyzer, and I think
  that it actually simplifies matters for the programmer.

- TP Yacc provides an option to generate the parser description (.LST file)
  and an option to trigger debugging output of the parser.

- The code templates for the TP Lex and Yacc output files are now supplied
  as ASCII files. This gives you the possibility to adapt the yylex and
  yyparse routines to your applications.

- The DFA and LALR parser construction algorithms have been reimplemented
  from scratch, and they are now far more efficient than in the 2.0
  version.

Version 3.0a fixes the following known bugs of the 3.0 version:

- TP Lex 3.0 sometimes loses transitions to the same state, e.g. in the
  following rule:
    a(b|c|d)*           return(1);

- TP Yacc 3.0 erases type and precedence information in multiple definitions
  of the same token, e.g.:
    %token <Integer> OP
    %left OP

- TP Yacc 3.0 loses rule precedence information in case of actions within
  a rule following the rightmost terminal of the rule, e.g.:
    %token OP
    %%
    expr : OP { act; } expr ;

The following incompatibility between TP and UNIX Lex is known, but will
NOT be fixed:

- The TP Lex library unit starts counting line numbers at 0, incrementing
  the count BEFORE a line is read (in contrast, UNIX Lex initializes
  yylineno to 1 and increments it AFTER the line end has been read). This
  is motivated by the way in which TP Lex maintains the current line,
  and will not affect your programs unless you explicitly reset the
  yylineno value (e.g., when opening a new input file). In such a case
  you should set yylineno to 0 rather than 1.

5-13-92 Albert Graef