xref: /freebsd/contrib/byacc/yacc.1 (revision d9a42747950146bf03cda7f6e25d219253f8a57a)
$Id: yacc.1,v 1.42 2022/11/06 17:07:16 tom Exp $

.TH YACC 1 "July 15, 1990"
.UC 6

..
.. Escape single quotes in literal strings from groff's Unicode transform.
Bulleted paragraph
..
YACC 1 "November 6, 2022" "Berkeley Yacc" "User Commands"
NAME
\*N - an LALR(1) parser generator
SYNOPSIS
\*n [ -BdghilLPrtvVy ] [ -b file_prefix ] [ -H defines_file ] [ -o output_file ] [ -p symbol_prefix ] filename
DESCRIPTION
\*N reads the grammar specification in the file filename and generates an LALR(1) parser for it. The parsers consist of a set of LALR(1) parsing tables and a driver routine written in the C programming language. \*N normally writes the parse tables and the driver routine to the file y.tab.c .

The following options are available:

5 -b file_prefix The -b option changes the prefix prepended to the output file names to the string denoted by file_prefix. The default prefix is the character y.

-B create a backtracking parser (compile-time configuration for btyacc).

-d causes the header file y.tab.h to be written. It contains #define's for the token identifiers.

-h print a usage message to the standard error.

-H defines_file causes #define's for the token identifiers to be written to the given defines_file rather than the y.tab.h file used by the -d option.

-g The -g option causes a graphical description of the generated LALR(1) parser to be written to the file y.dot in graphviz format, ready to be processed by dot (1).

-i The -i option causes a supplementary header file y.tab.i to be written. It contains extern declarations and supplementary #define's as needed to map the conventional yacc yy-prefixed names to whatever the -p option may specify. The code file, e.g., y.tab.c is modified to #include this file as well as the y.tab.h file, enforcing consistent usage of the symbols defined in those files.

The supplementary header file makes it simpler to separate compilation of lex- and yacc-files.

-l If the -l option is not specified, \*n will insert #line directives in the generated code. The #line directives let the C compiler relate errors in the generated code to the user's original code. If the -l option is specified, \*n will not insert the #line directives. \fI#line directives specified by the user will be retained.

-L enable position processing, e.g., \*(``%locations\*('' (compile-time configuration for btyacc).

-o output_file specify the filename for the parser file. If this option is not given, the output filename is the file prefix concatenated with the file suffix, e.g., y.tab.c. This overrides the -b option.

-p symbol_prefix The -p option changes the prefix prepended to yacc-generated symbols to the string denoted by symbol_prefix. The default prefix is the string yy.

-P create a reentrant parser, e.g., \*(``%pure-parser\*(''.

-r The -r option causes \*n to produce separate files for code and tables. The code file is named y.code.c , and the tables file is named y.tab.c . The prefix \*(``y.\*('' can be overridden using the -b option.

-s suppress \*(``#define\*('' statements generated for string literals in a \*(``%token\*('' statement, to more closely match original yacc behavior.

Normally when \*n sees a line such as .Ex %token OP_ADD "ADD" .Ee
it notices that the quoted \*(``ADD\*('' is a valid C identifier, and generates a #define not only for OP_ADD, but for ADD as well, e.g., .Ex #define OP_ADD 257

#define ADD 258 .Ee

The original yacc does not generate the second \*(``#define\*(''. The -s option suppresses this \*(``#define\*(''.
POSIX (IEEE 1003.1 2004) documents only names and numbers for \*(``%token\*('', though original yacc and bison also accept string literals.

-t The -t option changes the preprocessor directives generated by \*n so that debugging statements will be incorporated in the compiled code.

\*N sends debugging output to the standard output (compatible with both the original yacc and btyacc), while btyacc writes debugging output to the standard error (like bison).

-v The -v option causes a human-readable description of the generated parser to be written to the file y.output.

-V print the version number to the standard output.

-y \*n ignores this option, which bison supports for ostensible POSIX compatibility.

The filename parameter is not optional. However, \*n accepts a single \*(``-\*('' to read the grammar from the standard input. A double \*(``--\*('' marker denotes the end of options. A single filename parameter is expected after a \*(``--\*('' marker.

EXTENSIONS
\*N provides some extensions for compatibility with bison and other implementations of yacc. It accepts several long options which have equivalents in \*n. The %destructor and %locations features are available only if \*n has been configured and compiled to support the back-tracking (btyacc) functionality. The remaining features are always available:

%code keyword { code } Adds the indicated source code at a given point in the output file. The optional keyword tells \*n where to insert the code:

5 top just after the version-definition in the generated code-file.

5 requires just after the declaration of public parser variables. If the -d option is given, the code is inserted at the beginning of the defines-file.

5 provides just after the declaration of private parser variables. If the -d option is given, the code is inserted at the end of the defines-file.

If no keyword is given, the code is inserted at the beginning of the section of code copied verbatim from the source file. Multiple %code directives may be given; \*n inserts those into the corresponding code- or defines-file in the order that they appear in the source file.

%debug This has the same effect as the \*(``-t\*('' command-line option.

%destructor { code } symbol+ defines code that is invoked when a symbol is automatically discarded during error recovery. This code can be used to reclaim dynamically allocated memory associated with the corresponding semantic value for cases where user actions cannot manage the memory explicitly.

On encountering a parse error, the generated parser discards symbols on the stack and input tokens until it reaches a state that will allow parsing to continue. This error recovery approach results in a memory leak if the YYSTYPE value is, or contains, pointers to dynamically allocated memory.
The bracketed code is invoked whenever the parser discards one of the symbols. Within code, \*(``$$\*('' or \*(``$<tag>$\*('' designates the semantic value associated with the discarded symbol, and \*(``@$\*('' designates its location (see %locations directive).
A per-symbol destructor is defined by listing a grammar symbol in symbol+. A per-type destructor is defined by listing a semantic type tag (e.g., \*(``<some_tag>\*('') in symbol+; in this case, the parser will invoke code whenever it discards any grammar symbol that has that semantic type tag, unless that symbol has its own per-symbol destructor.
Two categories of default destructor are supported that are invoked when discarding any grammar symbol that has no per-symbol and no per-type destructor:

.bP the code for \*(``<*>\*('' is used for grammar symbols that have an explicitly declared semantic type tag (via \*(``%type\*(''); .bP the code for \*(``<>\*('' is used for grammar symbols that have no declared semantic type tag.

%empty ignored by \*n.

%expect number tells \*n the expected number of shift/reduce conflicts. That makes it only report the number if it differs.

%expect-rr number tell \*n the expected number of reduce/reduce conflicts. That makes it only report the number if it differs. This is (unlike bison) allowable in LALR parsers.

%locations tells \*n to enable management of position information associated with each token, provided by the lexer in the global variable yylloc, similar to management of semantic value information provided in yylval.

As for semantic values, locations can be referenced within actions using @$ to refer to the location of the left hand side symbol, and @N (N an integer) to refer to the location of one of the right hand side symbols. Also as for semantic values, when a rule is matched, a default action is used the compute the location represented by @$ as the beginning of the first symbol and the end of the last symbol in the right hand side of the rule. This default computation can be overridden by explicit assignment to @$ in a rule action.
The type of yylloc is YYLTYPE, which is defined by default as: .Ex typedef struct YYLTYPE { int first_line; int first_column; int last_line; int last_column; } YYLTYPE; .Ee
YYLTYPE can be redefined by the user (YYLTYPE_IS_DEFINED must be defined, to inhibit the default) in the declarations section of the specification file. As in bison, the macro YYLLOC_DEFAULT is invoked each time a rule is matched to calculate a position for the left hand side of the rule, before the associated action is executed; this macro can be redefined by the user.
This directive adds a YYLTYPE parameter to yyerror(). If the %pure-parser directive is present, a YYLTYPE parameter is added to yylex() calls.

%lex-param { argument-declaration } By default, the lexer accepts no parameters, e.g., yylex(). Use this directive to add parameter declarations for your customized lexer.

%parse-param { argument-declaration } By default, the parser accepts no parameters, e.g., yyparse(). Use this directive to add parameter declarations for your customized parser.

%pure-parser Most variables (other than yydebug and yynerrs) are allocated on the stack within yyparse, making the parser reasonably reentrant.

%token-table Make the parser's names for tokens available in the yytname array. However, \*n does not predefine \*(``$end\*('', \*(``$error\*('' or \*(``$undefined\*('' in this array.

PORTABILITY
According to Robert Corbett, .Ex Berkeley Yacc is an LALR(1) parser generator. Berkeley Yacc has been made as compatible as possible with AT&T Yacc. Berkeley Yacc can accept any input specification that conforms to the AT&T Yacc documentation. Specifications that take advantage of undocumented features of AT&T Yacc will probably be rejected. .Ee

The rationale in .Ex http://pubs.opengroup.org/onlinepubs/9699919799/utilities/yacc.html .Ee

documents some features of AT&T yacc which are no longer required for POSIX compliance.

That said, you may be interested in reusing grammar files with some other implementation which is not strictly compatible with AT&T yacc. For instance, there is bison. Here are a few differences: .bP Yacc accepts an equals mark preceding the left curly brace of an action (as in the original grammar file ftp.y): .Ex | STAT CRLF = { statcmd(); } .Ee .bP Yacc and bison emit code in different order, and in particular bison makes forward reference to common functions such as yylex, yyparse and yyerror without providing prototypes. .bP Bison's support for \*(``%expect\*('' is broken in more than one release. For best results using bison, delete that directive. .bP Bison has no equivalent for some of \*n's command-line options, relying on directives embedded in the grammar file. .bP Bison's \*(``-y\*('' option does not affect bison's lack of support for features of AT&T yacc which were deemed obsolescent. .bP Yacc accepts multiple parameters with %lex-param and %parse-param in two forms .Ex {type1 name1} {type2 name2} ... {type1 name1, type2 name2 ...} .Ee

Bison accepts the latter (though undocumented), but depending on the release may generate bad code. .bP Like bison, \*n will add parameters specified via %parse-param to yyparse, yyerror and (if configured for back-tracking) to the destructor declared using %destructor. Bison puts the additional parameters first for yyparse and yyerror but last for destructors. Yacc matches this behavior. .
DIAGNOSTICS
If there are rules that are never reduced, the number of such rules is reported on standard error. If there are any LALR(1) conflicts, the number of conflicts is reported on standard error.