************************************************************************* * * * Compiler Construction Tool Box * * ============================== * * * * Version 9208 * * * * Copyright (c) 1989, 1990, 1991, 1992 by * * * * Gesellschaft fuer Mathematik und Datenverarbeitung * * (German National Research Center for Computer Science) * * Forschungsstelle fuer Programmstrukturen * * an der Universitaet Karlsruhe * * * * All rights reserved. GMD assumes no responsibility for the use * * or reliability of its software. * * * ************************************************************************* Direct requests, comments, questions, and error reports to: Josef Grosch GMD Forschungsstelle Vincenz-Priessnitz-Str. 1 D-7500 Karlsruhe 1 Phone: +721-662226 Email: grosch@karlsruhe.gmd.de Distribution Format: -------------------- The compiler construction tool box is distributed via anonymous ftp in compressed tar format or in plain tar format on the following media: - DC300/600 data cartridge (streamer tape) - TK 50 - Exabyte - 1/2" magnetic tape (1600 bpi) To read a tape use: tar -xvfb /dev/rst0 20 or tar -xvb 20 or similar commands The directories and their contents are as follows: -------------------------------------------------- directory contents ------------------------------------------------------------------------ README this file Makefile compilation and installation of the tools doc.ps documentation in postscript format doc.me documentation in troff format, me macros man manual pages in troff format, man macros rex Scanner Generator lalr LALR(1) Parser Generator ell LL(1) Recursive Descent Parser Generator bnf Transforms Grammars from Extended BNF to Plain BNF front Common Front-End of Lalr, Ell, and Bnf reuse Library of Reusable Modules common Library for estra and ell specs Example Specifications for the Above Tools cg Common Program implementing Ast and Ag Ast = Generator for Abstract Syntax Trees Ag = Attribute Evaluator Generator puma Transformation Tool based on Pattern Matching l2r Transforms Lex input to Rex input y2l Transforms Yacc input to Lalr input r2l Transforms Rex input to Lex input rpp Rex PreProcessor: rpp + cg extract most of a scanner specification out of a parser specification estra Transformation of attributed trees (prototype) hexa contains the scanner and parser tables of Rex and Front (= front-end of Lalr and Bnf) converted from binary to ascii hexadecimal representation bin shell scripts (my version) lib executables, table and data files (for SUN 3/SunOS 4.0) (mtc Modula-2 to C translator) The names of the subdirectories indicate the following types of information: sub directory contents ------------------------------------------------------------------------ src source files in Modula-2 m2c source files in C (generated from the Modula-2 sources) c source files in C (hand-written) lib data files, module skeletons test test environment for a tool Documentation: -------------- The directories doc.ps and doc.me contain documentation in postscript format and in troff format (me macros). The document entitled "Toolbox Introduction" in the files intro.ps or intro.me gives an overview and introduces into the toolbox. It should be read first. The following documents are available: Filename Title ------------------------------------------------------------------------ intro Toolbox Introduction toolbox A Tool Box for Compiler Construction werkzeuge Werkzeuge fu"r den U"bersetzerbau reuse Reusable Software - A Collection of Modula-2-Modules prepro Preprocessors rex Rex - A Scanner Generator scanex Selected Examples of Scanner Specifications scangen Efficient Generation of Table-Driven Scanners lalr-ell The Parser Generators Lalr and Ell lalr Lalr - A Generator for Efficient Parsers ell Efficient and Comfortable Error Recovery in Recursive Descent Parsers highspeed Generators for High-Speed Front-Ends autogen Automatische Generierung effizienter Compiler ast Ast - A Generator for Abstract Syntax Trees toolsupp Tool Support for Data Structures ag Ag - An Attribute Evaluator Generator ooags Object-Oriented Attribute Grammars estra Spezifikation und Implementierung der Transformation attributierter Ba"ume puma Puma - A Generator for the Transformation of Attributed Trees trafo Transformation of Attributed Trees Using Pattern Matching (minilax Specification of a MiniLAX-Interpreter) (begmanual BEG - a Back End Generator - User Manual) References: ----------- 1. J. Grosch, `Generators for High-Speed Front-Ends', LNCS, 371, 81-92 (Oct. 1988), Springer Verlag. 2. H. Emmelmann, F. W. Schroeer, Rudolf Landwehr, ` BEG - a Generator for Efficient Back Ends', Sigplan Notices, 24, 227-237 (Jul. 1989) 3. W. M. Waite, J. Grosch and F. W. Schroeer, `Three Compiler Specifications', GMD-Studie Nr. 166, GMD Forschungsstelle an der Universitaet Karlsruhe, Aug. 1989. 4. J. Grosch, `Efficient Generation of Lexical Analysers', Software-Practice & Experience, 19, 1089-1103 (Nov. 1989). 5. J. Grosch, `Efficient and Comfortable Error Recovery in Recursive Descent Parsers', Structured Programming, 11, 129-140 (1990). 6. J. Grosch, H. Emmelmann, `A Tool Box for Compiler Construction', LNCS, 477, 106-116 (Oct. 1990), Springer Verlag. 7. J. Grosch, `Object-Oriented Attribute Grammars', in: Proceedings of the Fifth International Symposium on Computer and Information Sciences (ISCIS V) (Eds. A. E. Harmanci, E. Gelenbe), Cappadocia, Nevsehir, Turkey, 807-816, (Oct. 1990). 8. J. Grosch, `Lalr - a Generator for Efficient Parsers', Software-Practice & Experience, 20, 1115-1135 (Nov. 1990). 9. J. Grosch, `Tool Support for Data Structures', Structured Programming, 12, 31-38 (1991). Machine Dependencies: --------------------- All machine dependent code is isolated in the file System.c which is written in C. This file is set up to work under UNIX. There are three copies of this file in the following directories: reuse/c reuse/src reuse/m2c The UNIX command 'install' is used during installation. Unfortunately, this command is not as standard as it should be. If 'install' is missing on your machine or it complains about the calls then the shell script in the file hexa/install can simulate the desired behaviour. Installation: ------------- The Makefile at the global level controls the compilation and installation of the individual tools. It activates the tool specific Makefiles. Several tools use binary data files called Scan.Tab and Pars.Tab whose internal representation depends on whether the machine is little-endian or big-endian. All machines store integer numbers in a sequence of bytes with increasing adresses. An integer number usually consists of four bytes where the bytes at both ends are termed most significat byte (MSB) and least significant byte (LSB). Big-endian machines store the MSB at the lowest address and the LSB at the highest address. Little-endian machines store the bytes the other way round. Big-endian are e. g. MC 680x0, SUN/3, SUN/4, SPARC, little-endian are e. g. VAX, DEC Station. Initially the binary data files are configured for big-endian machines. To find out in which class your machine is in execute: make endian To convert the binary files from big-endian to little-endian or vice versa execute: make bin.conv Edit the first couple of lines in the Makefile to accomodate your needs. To compile the programs execute: make To install the programs execute: make install Recent Changes -------------- Version 9208: - The scanner generator 'rex' and the parser generators 'lalr' and 'ell' allow to chose arbitrary names for the generated modules. Therefore, it is possible to have several scanners and parsers in one program. - The length of a token and the lookahead in scanners generated by 'rex' is no longer restricted to 256 characters. Both, tokens and lookahead can be of arbitrary length. A restriction in the size of the tables generated by 'lalr' has been removed. Now it is possible to generate rather huge parsers. - The attribute grammar tool 'ag' has been extended to generate attribute evaluators for well-defined attribute grammars (WAGs). The program checks grammars whether they obey this property. It is possible to access non-local attributes and to compute attributes on a restricted form of graphs. - The auxiliary modules 'Errors' and 'Source' have been included into the library of reusable modules called 'reuse'. The 'Errors' module has been extended to support messages with a string argument. It allows to store the messages and print them sorted by the source position. An extra module named 'Positions' has been introduced in 'reuse', too, for the handling of source positions. - The program 'cg' which implements 'ast' and 'ag' accepts several input files. Instead of one file that communicates a tree definition to 'puma' with the fixed name 'TREE.TS' it is possible to produce several of those with different names. This is of interest if different "views" have to be communicated. - The extern declarations for malloc, free, and exit have been removed from the generated C code. - All tools do not generate # line directives by default, only upon request. - In case of fatal errors during the execution of generated modules a user defined exception routine can be called instead of the predefined 'exit (1)'. Version 9202: - The Toolbox contains a new tool for the transformation of attributed trees called 'puma'. It is based on pattern-matching. It replaces its predecessor 'estra' and comes with documentation in English. - The tool for abstract syntax trees 'ast' has been extended from single to multiple inheritance. So-called "subunits" allow the implementation of one abstract tree by several compilation units. The concept of "views" makes it possible to derive from a common specification several abstract syntax trees which represent subsets. - The attribute grammar tool 'ag' has analogously been extended to process object-oriented attribute grammars with multiple inheritance. It supports the generation of several attribute evaluators that run one after the other. - The error handling module for the parser generators 'lalr' and 'ell' are independent of the parser module or the grammar. The manual for these parser generators has been completely rewritten. - The Modula-2 to C translator 'mtc' has a new code-generator. This brings big efficiency improvements: 30% smaller program, 10% faster, 75% less dynamic memory consumption. It generates ANSI C as well as K&R C. - The sources of all tools are in ANSI C as well as in K&R C. - All tools generate modules in the target languages C (ANSI + K&R), C++, and Modula-2. - The interface to the operating system has been redesigned. It allows to switch IO operations between UNIX system calls and C library calls. This should assure much better portability. Compiler Construction Tool Box ============================== Rex (Regular EXpression tool) is a scanner generator whose specifications are based on regular expressions and arbitrary semantic actions written in one of the target languages C or Modula-2. As scanners sometimes have to consider the context to unambiguously recognize a token the right context can be speci- fied by an additional regular expression and the left context can be handled by so-called start states. The generated scanners automatically compute the line and column position of the tokens and offer an efficient mechanism to normalize identifiers and keywords to upper or lower case letters. The scanners are table- driven and run at a speed of 180,000 to 195,000 lines per minute on a MC 68020 processor. Lalr is a LALR(1) parser generator accepting grammars writ- ten in extended BNF notation which may be augmented by semantic actions expressed by statements of the target language. The gen- erator provides a mechanism for S-attribution, that is syn- thesized attributes can be computed during parsing. In case of LR-conflicts unlike other tools Lalr provides not only informa- tion about an internal state consisting of a set of items but it prints a derivation tree which is much more useful to analyze the problem. Conflicts can be resolved by specifying precedence and associativity of operators and productions. The generated parsers include automatic error recovery, error messages, and error repair. The parsers are table-driven and run at a speed of 560,000 lines per minute. Currently parsers can be generated in the target languages C and Modula-2. Ell is a LL(1) parser generator accepting the same specifi- cation language as Lalr except that the grammars must obey the LL(1) property. It is possible to evaluate an L-attribution during parsing. The generated parsers include automatic error recovery, error messages, and error repair like Lalr. The parsers are implemented following the recursive descent method and reach a speed of 810,000 lines per minute. The possible tar- get languages are again C and Modula-2. Ast - A Generator for Abstract Syntax Trees - generates abstract data types (program modules) to handle trees - the trees may be attributed - besides trees graphs are handled as well - nodes may be associated with arbitrary many attributes of arbitrary type - specifications are based on extended context-free grammars - common notation for concrete and abstract syntax - as well as for attributed trees and graphs - an extension mechanism provides single inheritance - trees are stored as linked records - generates efficient program modules - generates modules in Modula-2 or C - provides many tree operations (procedures): - node constructors combine aggregate notation and storage management - ascii graph reader and writer - binary graph reader and writer - reversal of lists - top down and bottom up traversal - interactive graph browser Ag - An Attribute Evaluator Generator - processes ordered attribute grammars (OAGs) - processes higher order attribute grammars (HAGs) - operates on abstract syntax - is based on tree modules generated by Ast - the tree structure is fully known - terminals and nonterminals may have arbitrary many attributes - attributes can have any target language type - allows tree-valued attributes - differentiates input and output attributes - allows attributes local to rules - allows to eliminate chain rules - offers an extension mechanism (single inheritance) - attributes are denoted by unique selector names instead of nonterminal names with subscripts - attribute computations are expressed in the target language - attribute computations are written in a functional style - attribute computations can call external functions - non-functional statements and side-effects are possible - allows to write compact, modular, and readable specifications - AGs can consist of several modules - the context-free grammar is specified only once - checks an AG for completeness of the attribute computations - checks for unused attributes - checks an AG for the classes SNC, DNC, OAG, LAG, and SAG - the evaluators are directly coded using recursive procedures - generates efficient evaluators - generates evaluators in Modula-2 (or C) Puma - Transformation Tool based on Pattern Matching - last but not least A comparison of the above tools with the corresponding UNIX tools shows that significant improvements in terms of error handling as well as efficiency have been achieved: Rex generated scanners are 4 times faster than those of LEX. Lalr generated parsers are 2-3 times faster than those of YACC. Ell generated parsers are 4 times faster than those of YACC. The input languages of the tools are improvements of the LEX and YACC inputs. The tools also understand LEX and YACC syntax with the help of the preprocessors l2r and y2l. The tool box is publicly copyable. It has been developed since 1987. It has been tested by generating scanners and parsers for e. g. Pascal, Modula, Oberon, Ada and found stable. The tool box is implemented in Modula-2. It has been developed using our own Modula-2 compiler called MOCKA on a MC 68020 based UNIX workstation. It has been ported to the SUN workstation and been compiled successfully using the SUN Modula-2 compiler. The tools also run on VAX/BSD UNIX and VAX/ULTRIX machines. This should assure a reasonable level of portability for the Modula-2 code. Meanwhile the sources exist in C, too.