Annotation of embedaddon/sqlite3/doc/lemon.html, revision 1.1.1.1

1.1       misho       1: <html>
                      2: <head>
                      3: <title>The Lemon Parser Generator</title>
                      4: </head>
                      5: <body bgcolor=white>
                      6: <h1 align=center>The Lemon Parser Generator</h1>
                      7: 
                      8: <p>Lemon is an LALR(1) parser generator for C or C++.  
                      9: It does the same job as ``bison'' and ``yacc''.
                     10: But lemon is not another bison or yacc clone.  It
                     11: uses a different grammar syntax which is designed to
                     12: reduce the number of coding errors.  Lemon also uses a more
                     13: sophisticated parsing engine that is faster than yacc and
                     14: bison and which is both reentrant and thread-safe.
                     15: Furthermore, Lemon implements features that can be used
                     16: to eliminate resource leaks, making is suitable for use
                     17: in long-running programs such as graphical user interfaces
                     18: or embedded controllers.</p>
                     19: 
                     20: <p>This document is an introduction to the Lemon
                     21: parser generator.</p>
                     22: 
                     23: <h2>Theory of Operation</h2>
                     24: 
                     25: <p>The main goal of Lemon is to translate a context free grammar (CFG)
                     26: for a particular language into C code that implements a parser for
                     27: that language.
                     28: The program has two inputs:
                     29: <ul>
                     30: <li>The grammar specification.
                     31: <li>A parser template file.
                     32: </ul>
                     33: Typically, only the grammar specification is supplied by the programmer.
                     34: Lemon comes with a default parser template which works fine for most
                     35: applications.  But the user is free to substitute a different parser
                     36: template if desired.</p>
                     37: 
                     38: <p>Depending on command-line options, Lemon will generate between
                     39: one and three files of outputs.
                     40: <ul>
                     41: <li>C code to implement the parser.
                     42: <li>A header file defining an integer ID for each terminal symbol.
                     43: <li>An information file that describes the states of the generated parser
                     44:     automaton.
                     45: </ul>
                     46: By default, all three of these output files are generated.
                     47: The header file is suppressed if the ``-m'' command-line option is
                     48: used and the report file is omitted when ``-q'' is selected.</p>
                     49: 
                     50: <p>The grammar specification file uses a ``.y'' suffix, by convention.
                     51: In the examples used in this document, we'll assume the name of the
                     52: grammar file is ``gram.y''.  A typical use of Lemon would be the
                     53: following command:
                     54: <pre>
                     55:    lemon gram.y
                     56: </pre>
                     57: This command will generate three output files named ``gram.c'',
                     58: ``gram.h'' and ``gram.out''.
                     59: The first is C code to implement the parser.  The second
                     60: is the header file that defines numerical values for all
                     61: terminal symbols, and the last is the report that explains
                     62: the states used by the parser automaton.</p>
                     63: 
                     64: <h3>Command Line Options</h3>
                     65: 
                     66: <p>The behavior of Lemon can be modified using command-line options.
                     67: You can obtain a list of the available command-line options together
                     68: with a brief explanation of what each does by typing
                     69: <pre>
                     70:    lemon -?
                     71: </pre>
                     72: As of this writing, the following command-line options are supported:
                     73: <ul>
                     74: <li><tt>-b</tt>
                     75: <li><tt>-c</tt>
                     76: <li><tt>-g</tt>
                     77: <li><tt>-m</tt>
                     78: <li><tt>-q</tt>
                     79: <li><tt>-s</tt>
                     80: <li><tt>-x</tt>
                     81: </ul>
                     82: The ``-b'' option reduces the amount of text in the report file by
                     83: printing only the basis of each parser state, rather than the full
                     84: configuration.
                     85: The ``-c'' option suppresses action table compression.  Using -c
                     86: will make the parser a little larger and slower but it will detect
                     87: syntax errors sooner.
                     88: The ``-g'' option causes no output files to be generated at all.
                     89: Instead, the input grammar file is printed on standard output but
                     90: with all comments, actions and other extraneous text deleted.  This
                     91: is a useful way to get a quick summary of a grammar.
                     92: The ``-m'' option causes the output C source file to be compatible
                     93: with the ``makeheaders'' program.
                     94: Makeheaders is a program that automatically generates header files
                     95: from C source code.  When the ``-m'' option is used, the header
                     96: file is not output since the makeheaders program will take care
                     97: of generated all header files automatically.
                     98: The ``-q'' option suppresses the report file.
                     99: Using ``-s'' causes a brief summary of parser statistics to be
                    100: printed.  Like this:
                    101: <pre>
                    102:    Parser statistics: 74 terminals, 70 nonterminals, 179 rules
                    103:                       340 states, 2026 parser table entries, 0 conflicts
                    104: </pre>
                    105: Finally, the ``-x'' option causes Lemon to print its version number
                    106: and then stops without attempting to read the grammar or generate a parser.</p>
                    107: 
                    108: <h3>The Parser Interface</h3>
                    109: 
                    110: <p>Lemon doesn't generate a complete, working program.  It only generates
                    111: a few subroutines that implement a parser.  This section describes
                    112: the interface to those subroutines.  It is up to the programmer to
                    113: call these subroutines in an appropriate way in order to produce a
                    114: complete system.</p>
                    115: 
                    116: <p>Before a program begins using a Lemon-generated parser, the program
                    117: must first create the parser.
                    118: A new parser is created as follows:
                    119: <pre>
                    120:    void *pParser = ParseAlloc( malloc );
                    121: </pre>
                    122: The ParseAlloc() routine allocates and initializes a new parser and
                    123: returns a pointer to it.
                    124: The actual data structure used to represent a parser is opaque --
                    125: its internal structure is not visible or usable by the calling routine.
                    126: For this reason, the ParseAlloc() routine returns a pointer to void
                    127: rather than a pointer to some particular structure.
                    128: The sole argument to the ParseAlloc() routine is a pointer to the
                    129: subroutine used to allocate memory.  Typically this means ``malloc()''.</p>
                    130: 
                    131: <p>After a program is finished using a parser, it can reclaim all
                    132: memory allocated by that parser by calling
                    133: <pre>
                    134:    ParseFree(pParser, free);
                    135: </pre>
                    136: The first argument is the same pointer returned by ParseAlloc().  The
                    137: second argument is a pointer to the function used to release bulk
                    138: memory back to the system.</p>
                    139: 
                    140: <p>After a parser has been allocated using ParseAlloc(), the programmer
                    141: must supply the parser with a sequence of tokens (terminal symbols) to
                    142: be parsed.  This is accomplished by calling the following function
                    143: once for each token:
                    144: <pre>
                    145:    Parse(pParser, hTokenID, sTokenData, pArg);
                    146: </pre>
                    147: The first argument to the Parse() routine is the pointer returned by
                    148: ParseAlloc().
                    149: The second argument is a small positive integer that tells the parse the
                    150: type of the next token in the data stream.
                    151: There is one token type for each terminal symbol in the grammar.
                    152: The gram.h file generated by Lemon contains #define statements that
                    153: map symbolic terminal symbol names into appropriate integer values.
                    154: (A value of 0 for the second argument is a special flag to the
                    155: parser to indicate that the end of input has been reached.)
                    156: The third argument is the value of the given token.  By default,
                    157: the type of the third argument is integer, but the grammar will
                    158: usually redefine this type to be some kind of structure.
                    159: Typically the second argument will be a broad category of tokens
                    160: such as ``identifier'' or ``number'' and the third argument will
                    161: be the name of the identifier or the value of the number.</p>
                    162: 
                    163: <p>The Parse() function may have either three or four arguments,
                    164: depending on the grammar.  If the grammar specification file request
                    165: it, the Parse() function will have a fourth parameter that can be
                    166: of any type chosen by the programmer.  The parser doesn't do anything
                    167: with this argument except to pass it through to action routines.
                    168: This is a convenient mechanism for passing state information down
                    169: to the action routines without having to use global variables.</p>
                    170: 
                    171: <p>A typical use of a Lemon parser might look something like the
                    172: following:
                    173: <pre>
                    174:    01 ParseTree *ParseFile(const char *zFilename){
                    175:    02    Tokenizer *pTokenizer;
                    176:    03    void *pParser;
                    177:    04    Token sToken;
                    178:    05    int hTokenId;
                    179:    06    ParserState sState;
                    180:    07
                    181:    08    pTokenizer = TokenizerCreate(zFilename);
                    182:    09    pParser = ParseAlloc( malloc );
                    183:    10    InitParserState(&sState);
                    184:    11    while( GetNextToken(pTokenizer, &hTokenId, &sToken) ){
                    185:    12       Parse(pParser, hTokenId, sToken, &sState);
                    186:    13    }
                    187:    14    Parse(pParser, 0, sToken, &sState);
                    188:    15    ParseFree(pParser, free );
                    189:    16    TokenizerFree(pTokenizer);
                    190:    17    return sState.treeRoot;
                    191:    18 }
                    192: </pre>
                    193: This example shows a user-written routine that parses a file of
                    194: text and returns a pointer to the parse tree.
                    195: (We've omitted all error-handling from this example to keep it
                    196: simple.)
                    197: We assume the existence of some kind of tokenizer which is created
                    198: using TokenizerCreate() on line 8 and deleted by TokenizerFree()
                    199: on line 16.  The GetNextToken() function on line 11 retrieves the
                    200: next token from the input file and puts its type in the 
                    201: integer variable hTokenId.  The sToken variable is assumed to be
                    202: some kind of structure that contains details about each token,
                    203: such as its complete text, what line it occurs on, etc. </p>
                    204: 
                    205: <p>This example also assumes the existence of structure of type
                    206: ParserState that holds state information about a particular parse.
                    207: An instance of such a structure is created on line 6 and initialized
                    208: on line 10.  A pointer to this structure is passed into the Parse()
                    209: routine as the optional 4th argument.
                    210: The action routine specified by the grammar for the parser can use
                    211: the ParserState structure to hold whatever information is useful and
                    212: appropriate.  In the example, we note that the treeRoot field of
                    213: the ParserState structure is left pointing to the root of the parse
                    214: tree.</p>
                    215: 
                    216: <p>The core of this example as it relates to Lemon is as follows:
                    217: <pre>
                    218:    ParseFile(){
                    219:       pParser = ParseAlloc( malloc );
                    220:       while( GetNextToken(pTokenizer,&hTokenId, &sToken) ){
                    221:          Parse(pParser, hTokenId, sToken);
                    222:       }
                    223:       Parse(pParser, 0, sToken);
                    224:       ParseFree(pParser, free );
                    225:    }
                    226: </pre>
                    227: Basically, what a program has to do to use a Lemon-generated parser
                    228: is first create the parser, then send it lots of tokens obtained by
                    229: tokenizing an input source.  When the end of input is reached, the
                    230: Parse() routine should be called one last time with a token type
                    231: of 0.  This step is necessary to inform the parser that the end of
                    232: input has been reached.  Finally, we reclaim memory used by the
                    233: parser by calling ParseFree().</p>
                    234: 
                    235: <p>There is one other interface routine that should be mentioned
                    236: before we move on.
                    237: The ParseTrace() function can be used to generate debugging output
                    238: from the parser.  A prototype for this routine is as follows:
                    239: <pre>
                    240:    ParseTrace(FILE *stream, char *zPrefix);
                    241: </pre>
                    242: After this routine is called, a short (one-line) message is written
                    243: to the designated output stream every time the parser changes states
                    244: or calls an action routine.  Each such message is prefaced using
                    245: the text given by zPrefix.  This debugging output can be turned off
                    246: by calling ParseTrace() again with a first argument of NULL (0).</p>
                    247: 
                    248: <h3>Differences With YACC and BISON</h3>
                    249: 
                    250: <p>Programmers who have previously used the yacc or bison parser
                    251: generator will notice several important differences between yacc and/or
                    252: bison and Lemon.
                    253: <ul>
                    254: <li>In yacc and bison, the parser calls the tokenizer.  In Lemon,
                    255:     the tokenizer calls the parser.
                    256: <li>Lemon uses no global variables.  Yacc and bison use global variables
                    257:     to pass information between the tokenizer and parser.
                    258: <li>Lemon allows multiple parsers to be running simultaneously.  Yacc
                    259:     and bison do not.
                    260: </ul>
                    261: These differences may cause some initial confusion for programmers
                    262: with prior yacc and bison experience.
                    263: But after years of experience using Lemon, I firmly
                    264: believe that the Lemon way of doing things is better.</p>
                    265: 
                    266: <h2>Input File Syntax</h2>
                    267: 
                    268: <p>The main purpose of the grammar specification file for Lemon is
                    269: to define the grammar for the parser.  But the input file also
                    270: specifies additional information Lemon requires to do its job.
                    271: Most of the work in using Lemon is in writing an appropriate
                    272: grammar file.</p>
                    273: 
                    274: <p>The grammar file for lemon is, for the most part, free format.
                    275: It does not have sections or divisions like yacc or bison.  Any
                    276: declaration can occur at any point in the file.
                    277: Lemon ignores whitespace (except where it is needed to separate
                    278: tokens) and it honors the same commenting conventions as C and C++.</p>
                    279: 
                    280: <h3>Terminals and Nonterminals</h3>
                    281: 
                    282: <p>A terminal symbol (token) is any string of alphanumeric
                    283: and underscore characters
                    284: that begins with an upper case letter.
                    285: A terminal can contain lowercase letters after the first character,
                    286: but the usual convention is to make terminals all upper case.
                    287: A nonterminal, on the other hand, is any string of alphanumeric
                    288: and underscore characters than begins with a lower case letter.
                    289: Again, the usual convention is to make nonterminals use all lower
                    290: case letters.</p>
                    291: 
                    292: <p>In Lemon, terminal and nonterminal symbols do not need to 
                    293: be declared or identified in a separate section of the grammar file.
                    294: Lemon is able to generate a list of all terminals and nonterminals
                    295: by examining the grammar rules, and it can always distinguish a
                    296: terminal from a nonterminal by checking the case of the first
                    297: character of the name.</p>
                    298: 
                    299: <p>Yacc and bison allow terminal symbols to have either alphanumeric
                    300: names or to be individual characters included in single quotes, like
                    301: this: ')' or '$'.  Lemon does not allow this alternative form for
                    302: terminal symbols.  With Lemon, all symbols, terminals and nonterminals,
                    303: must have alphanumeric names.</p>
                    304: 
                    305: <h3>Grammar Rules</h3>
                    306: 
                    307: <p>The main component of a Lemon grammar file is a sequence of grammar
                    308: rules.
                    309: Each grammar rule consists of a nonterminal symbol followed by
                    310: the special symbol ``::='' and then a list of terminals and/or nonterminals.
                    311: The rule is terminated by a period.
                    312: The list of terminals and nonterminals on the right-hand side of the
                    313: rule can be empty.
                    314: Rules can occur in any order, except that the left-hand side of the
                    315: first rule is assumed to be the start symbol for the grammar (unless
                    316: specified otherwise using the <tt>%start</tt> directive described below.)
                    317: A typical sequence of grammar rules might look something like this:
                    318: <pre>
                    319:   expr ::= expr PLUS expr.
                    320:   expr ::= expr TIMES expr.
                    321:   expr ::= LPAREN expr RPAREN.
                    322:   expr ::= VALUE.
                    323: </pre>
                    324: </p>
                    325: 
                    326: <p>There is one non-terminal in this example, ``expr'', and five
                    327: terminal symbols or tokens: ``PLUS'', ``TIMES'', ``LPAREN'',
                    328: ``RPAREN'' and ``VALUE''.</p>
                    329: 
                    330: <p>Like yacc and bison, Lemon allows the grammar to specify a block
                    331: of C code that will be executed whenever a grammar rule is reduced
                    332: by the parser.
                    333: In Lemon, this action is specified by putting the C code (contained
                    334: within curly braces <tt>{...}</tt>) immediately after the
                    335: period that closes the rule.
                    336: For example:
                    337: <pre>
                    338:   expr ::= expr PLUS expr.   { printf("Doing an addition...\n"); }
                    339: </pre>
                    340: </p>
                    341: 
                    342: <p>In order to be useful, grammar actions must normally be linked to
                    343: their associated grammar rules.
                    344: In yacc and bison, this is accomplished by embedding a ``$$'' in the
                    345: action to stand for the value of the left-hand side of the rule and
                    346: symbols ``$1'', ``$2'', and so forth to stand for the value of
                    347: the terminal or nonterminal at position 1, 2 and so forth on the
                    348: right-hand side of the rule.
                    349: This idea is very powerful, but it is also very error-prone.  The
                    350: single most common source of errors in a yacc or bison grammar is
                    351: to miscount the number of symbols on the right-hand side of a grammar
                    352: rule and say ``$7'' when you really mean ``$8''.</p>
                    353: 
                    354: <p>Lemon avoids the need to count grammar symbols by assigning symbolic
                    355: names to each symbol in a grammar rule and then using those symbolic
                    356: names in the action.
                    357: In yacc or bison, one would write this:
                    358: <pre>
                    359:   expr -> expr PLUS expr  { $$ = $1 + $3; };
                    360: </pre>
                    361: But in Lemon, the same rule becomes the following:
                    362: <pre>
                    363:   expr(A) ::= expr(B) PLUS expr(C).  { A = B+C; }
                    364: </pre>
                    365: In the Lemon rule, any symbol in parentheses after a grammar rule
                    366: symbol becomes a place holder for that symbol in the grammar rule.
                    367: This place holder can then be used in the associated C action to
                    368: stand for the value of that symbol.<p>
                    369: 
                    370: <p>The Lemon notation for linking a grammar rule with its reduce
                    371: action is superior to yacc/bison on several counts.
                    372: First, as mentioned above, the Lemon method avoids the need to
                    373: count grammar symbols.
                    374: Secondly, if a terminal or nonterminal in a Lemon grammar rule
                    375: includes a linking symbol in parentheses but that linking symbol
                    376: is not actually used in the reduce action, then an error message
                    377: is generated.
                    378: For example, the rule
                    379: <pre>
                    380:   expr(A) ::= expr(B) PLUS expr(C).  { A = B; }
                    381: </pre>
                    382: will generate an error because the linking symbol ``C'' is used
                    383: in the grammar rule but not in the reduce action.</p>
                    384: 
                    385: <p>The Lemon notation for linking grammar rules to reduce actions
                    386: also facilitates the use of destructors for reclaiming memory
                    387: allocated by the values of terminals and nonterminals on the
                    388: right-hand side of a rule.</p>
                    389: 
                    390: <h3>Precedence Rules</h3>
                    391: 
                    392: <p>Lemon resolves parsing ambiguities in exactly the same way as
                    393: yacc and bison.  A shift-reduce conflict is resolved in favor
                    394: of the shift, and a reduce-reduce conflict is resolved by reducing
                    395: whichever rule comes first in the grammar file.</p>
                    396: 
                    397: <p>Just like in
                    398: yacc and bison, Lemon allows a measure of control 
                    399: over the resolution of paring conflicts using precedence rules.
                    400: A precedence value can be assigned to any terminal symbol
                    401: using the %left, %right or %nonassoc directives.  Terminal symbols
                    402: mentioned in earlier directives have a lower precedence that
                    403: terminal symbols mentioned in later directives.  For example:</p>
                    404: 
                    405: <p><pre>
                    406:    %left AND.
                    407:    %left OR.
                    408:    %nonassoc EQ NE GT GE LT LE.
                    409:    %left PLUS MINUS.
                    410:    %left TIMES DIVIDE MOD.
                    411:    %right EXP NOT.
                    412: </pre></p>
                    413: 
                    414: <p>In the preceding sequence of directives, the AND operator is
                    415: defined to have the lowest precedence.  The OR operator is one
                    416: precedence level higher.  And so forth.  Hence, the grammar would
                    417: attempt to group the ambiguous expression
                    418: <pre>
                    419:      a AND b OR c
                    420: </pre>
                    421: like this
                    422: <pre>
                    423:      a AND (b OR c).
                    424: </pre>
                    425: The associativity (left, right or nonassoc) is used to determine
                    426: the grouping when the precedence is the same.  AND is left-associative
                    427: in our example, so
                    428: <pre>
                    429:      a AND b AND c
                    430: </pre>
                    431: is parsed like this
                    432: <pre>
                    433:      (a AND b) AND c.
                    434: </pre>
                    435: The EXP operator is right-associative, though, so
                    436: <pre>
                    437:      a EXP b EXP c
                    438: </pre>
                    439: is parsed like this
                    440: <pre>
                    441:      a EXP (b EXP c).
                    442: </pre>
                    443: The nonassoc precedence is used for non-associative operators.
                    444: So
                    445: <pre>
                    446:      a EQ b EQ c
                    447: </pre>
                    448: is an error.</p>
                    449: 
                    450: <p>The precedence of non-terminals is transferred to rules as follows:
                    451: The precedence of a grammar rule is equal to the precedence of the
                    452: left-most terminal symbol in the rule for which a precedence is
                    453: defined.  This is normally what you want, but in those cases where
                    454: you want to precedence of a grammar rule to be something different,
                    455: you can specify an alternative precedence symbol by putting the
                    456: symbol in square braces after the period at the end of the rule and
                    457: before any C-code.  For example:</p>
                    458: 
                    459: <p><pre>
                    460:    expr = MINUS expr.  [NOT]
                    461: </pre></p>
                    462: 
                    463: <p>This rule has a precedence equal to that of the NOT symbol, not the
                    464: MINUS symbol as would have been the case by default.</p>
                    465: 
                    466: <p>With the knowledge of how precedence is assigned to terminal
                    467: symbols and individual
                    468: grammar rules, we can now explain precisely how parsing conflicts
                    469: are resolved in Lemon.  Shift-reduce conflicts are resolved
                    470: as follows:
                    471: <ul>
                    472: <li> If either the token to be shifted or the rule to be reduced
                    473:      lacks precedence information, then resolve in favor of the
                    474:      shift, but report a parsing conflict.
                    475: <li> If the precedence of the token to be shifted is greater than
                    476:      the precedence of the rule to reduce, then resolve in favor
                    477:      of the shift.  No parsing conflict is reported.
                    478: <li> If the precedence of the token it be shifted is less than the
                    479:      precedence of the rule to reduce, then resolve in favor of the
                    480:      reduce action.  No parsing conflict is reported.
                    481: <li> If the precedences are the same and the shift token is
                    482:      right-associative, then resolve in favor of the shift.
                    483:      No parsing conflict is reported.
                    484: <li> If the precedences are the same the the shift token is
                    485:      left-associative, then resolve in favor of the reduce.
                    486:      No parsing conflict is reported.
                    487: <li> Otherwise, resolve the conflict by doing the shift and
                    488:      report the parsing conflict.
                    489: </ul>
                    490: Reduce-reduce conflicts are resolved this way:
                    491: <ul>
                    492: <li> If either reduce rule 
                    493:      lacks precedence information, then resolve in favor of the
                    494:      rule that appears first in the grammar and report a parsing
                    495:      conflict.
                    496: <li> If both rules have precedence and the precedence is different
                    497:      then resolve the dispute in favor of the rule with the highest
                    498:      precedence and do not report a conflict.
                    499: <li> Otherwise, resolve the conflict by reducing by the rule that
                    500:      appears first in the grammar and report a parsing conflict.
                    501: </ul>
                    502: 
                    503: <h3>Special Directives</h3>
                    504: 
                    505: <p>The input grammar to Lemon consists of grammar rules and special
                    506: directives.  We've described all the grammar rules, so now we'll
                    507: talk about the special directives.</p>
                    508: 
                    509: <p>Directives in lemon can occur in any order.  You can put them before
                    510: the grammar rules, or after the grammar rules, or in the mist of the
                    511: grammar rules.  It doesn't matter.  The relative order of
                    512: directives used to assign precedence to terminals is important, but
                    513: other than that, the order of directives in Lemon is arbitrary.</p>
                    514: 
                    515: <p>Lemon supports the following special directives:
                    516: <ul>
                    517: <li><tt>%code</tt>
                    518: <li><tt>%default_destructor</tt>
                    519: <li><tt>%default_type</tt>
                    520: <li><tt>%destructor</tt>
                    521: <li><tt>%extra_argument</tt>
                    522: <li><tt>%include</tt>
                    523: <li><tt>%left</tt>
                    524: <li><tt>%name</tt>
                    525: <li><tt>%nonassoc</tt>
                    526: <li><tt>%parse_accept</tt>
                    527: <li><tt>%parse_failure </tt>
                    528: <li><tt>%right</tt>
                    529: <li><tt>%stack_overflow</tt>
                    530: <li><tt>%stack_size</tt>
                    531: <li><tt>%start_symbol</tt>
                    532: <li><tt>%syntax_error</tt>
                    533: <li><tt>%token_destructor</tt>
                    534: <li><tt>%token_prefix</tt>
                    535: <li><tt>%token_type</tt>
                    536: <li><tt>%type</tt>
                    537: </ul>
                    538: Each of these directives will be described separately in the
                    539: following sections:</p>
                    540: 
                    541: <h4>The <tt>%code</tt> directive</h4>
                    542: 
                    543: <p>The %code directive is used to specify addition C/C++ code that
                    544: is added to the end of the main output file.  This is similar to
                    545: the %include directive except that %include is inserted at the
                    546: beginning of the main output file.</p>
                    547: 
                    548: <p>%code is typically used to include some action routines or perhaps
                    549: a tokenizer as part of the output file.</p>
                    550: 
                    551: <h4>The <tt>%default_destructor</tt> directive</h4>
                    552: 
                    553: <p>The %default_destructor directive specifies a destructor to 
                    554: use for non-terminals that do not have their own destructor
                    555: specified by a separate %destructor directive.  See the documentation
                    556: on the %destructor directive below for additional information.</p>
                    557: 
                    558: <p>In some grammers, many different non-terminal symbols have the
                    559: same datatype and hence the same destructor.  This directive is
                    560: a convenience way to specify the same destructor for all those
                    561: non-terminals using a single statement.</p>
                    562: 
                    563: <h4>The <tt>%default_type</tt> directive</h4>
                    564: 
                    565: <p>The %default_type directive specifies the datatype of non-terminal
                    566: symbols that do no have their own datatype defined using a separate
                    567: %type directive.  See the documentation on %type below for addition
                    568: information.</p>
                    569: 
                    570: <h4>The <tt>%destructor</tt> directive</h4>
                    571: 
                    572: <p>The %destructor directive is used to specify a destructor for
                    573: a non-terminal symbol.
                    574: (See also the %token_destructor directive which is used to
                    575: specify a destructor for terminal symbols.)</p>
                    576: 
                    577: <p>A non-terminal's destructor is called to dispose of the
                    578: non-terminal's value whenever the non-terminal is popped from
                    579: the stack.  This includes all of the following circumstances:
                    580: <ul>
                    581: <li> When a rule reduces and the value of a non-terminal on
                    582:      the right-hand side is not linked to C code.
                    583: <li> When the stack is popped during error processing.
                    584: <li> When the ParseFree() function runs.
                    585: </ul>
                    586: The destructor can do whatever it wants with the value of
                    587: the non-terminal, but its design is to deallocate memory
                    588: or other resources held by that non-terminal.</p>
                    589: 
                    590: <p>Consider an example:
                    591: <pre>
                    592:    %type nt {void*}
                    593:    %destructor nt { free($$); }
                    594:    nt(A) ::= ID NUM.   { A = malloc( 100 ); }
                    595: </pre>
                    596: This example is a bit contrived but it serves to illustrate how
                    597: destructors work.  The example shows a non-terminal named
                    598: ``nt'' that holds values of type ``void*''.  When the rule for
                    599: an ``nt'' reduces, it sets the value of the non-terminal to
                    600: space obtained from malloc().  Later, when the nt non-terminal
                    601: is popped from the stack, the destructor will fire and call
                    602: free() on this malloced space, thus avoiding a memory leak.
                    603: (Note that the symbol ``$$'' in the destructor code is replaced
                    604: by the value of the non-terminal.)</p>
                    605: 
                    606: <p>It is important to note that the value of a non-terminal is passed
                    607: to the destructor whenever the non-terminal is removed from the
                    608: stack, unless the non-terminal is used in a C-code action.  If
                    609: the non-terminal is used by C-code, then it is assumed that the
                    610: C-code will take care of destroying it if it should really
                    611: be destroyed.  More commonly, the value is used to build some
                    612: larger structure and we don't want to destroy it, which is why
                    613: the destructor is not called in this circumstance.</p>
                    614: 
                    615: <p>By appropriate use of destructors, it is possible to
                    616: build a parser using Lemon that can be used within a long-running
                    617: program, such as a GUI, that will not leak memory or other resources.
                    618: To do the same using yacc or bison is much more difficult.</p>
                    619: 
                    620: <h4>The <tt>%extra_argument</tt> directive</h4>
                    621: 
                    622: The %extra_argument directive instructs Lemon to add a 4th parameter
                    623: to the parameter list of the Parse() function it generates.  Lemon
                    624: doesn't do anything itself with this extra argument, but it does
                    625: make the argument available to C-code action routines, destructors,
                    626: and so forth.  For example, if the grammar file contains:</p>
                    627: 
                    628: <p><pre>
                    629:     %extra_argument { MyStruct *pAbc }
                    630: </pre></p>
                    631: 
                    632: <p>Then the Parse() function generated will have an 4th parameter
                    633: of type ``MyStruct*'' and all action routines will have access to
                    634: a variable named ``pAbc'' that is the value of the 4th parameter
                    635: in the most recent call to Parse().</p>
                    636: 
                    637: <h4>The <tt>%include</tt> directive</h4>
                    638: 
                    639: <p>The %include directive specifies C code that is included at the
                    640: top of the generated parser.  You can include any text you want --
                    641: the Lemon parser generator copies it blindly.  If you have multiple
                    642: %include directives in your grammar file the value of the last
                    643: %include directive overwrites all the others.</p.
                    644: 
                    645: <p>The %include directive is very handy for getting some extra #include
                    646: preprocessor statements at the beginning of the generated parser.
                    647: For example:</p>
                    648: 
                    649: <p><pre>
                    650:    %include {#include &lt;unistd.h&gt;}
                    651: </pre></p>
                    652: 
                    653: <p>This might be needed, for example, if some of the C actions in the
                    654: grammar call functions that are prototyed in unistd.h.</p>
                    655: 
                    656: <h4>The <tt>%left</tt> directive</h4>
                    657: 
                    658: The %left directive is used (along with the %right and
                    659: %nonassoc directives) to declare precedences of terminal
                    660: symbols.  Every terminal symbol whose name appears after
                    661: a %left directive but before the next period (``.'') is
                    662: given the same left-associative precedence value.  Subsequent
                    663: %left directives have higher precedence.  For example:</p>
                    664: 
                    665: <p><pre>
                    666:    %left AND.
                    667:    %left OR.
                    668:    %nonassoc EQ NE GT GE LT LE.
                    669:    %left PLUS MINUS.
                    670:    %left TIMES DIVIDE MOD.
                    671:    %right EXP NOT.
                    672: </pre></p>
                    673: 
                    674: <p>Note the period that terminates each %left, %right or %nonassoc
                    675: directive.</p>
                    676: 
                    677: <p>LALR(1) grammars can get into a situation where they require
                    678: a large amount of stack space if you make heavy use or right-associative
                    679: operators.  For this reason, it is recommended that you use %left
                    680: rather than %right whenever possible.</p>
                    681: 
                    682: <h4>The <tt>%name</tt> directive</h4>
                    683: 
                    684: <p>By default, the functions generated by Lemon all begin with the
                    685: five-character string ``Parse''.  You can change this string to something
                    686: different using the %name directive.  For instance:</p>
                    687: 
                    688: <p><pre>
                    689:    %name Abcde
                    690: </pre></p>
                    691: 
                    692: <p>Putting this directive in the grammar file will cause Lemon to generate
                    693: functions named
                    694: <ul>
                    695: <li> AbcdeAlloc(),
                    696: <li> AbcdeFree(),
                    697: <li> AbcdeTrace(), and
                    698: <li> Abcde().
                    699: </ul>
                    700: The %name directive allows you to generator two or more different
                    701: parsers and link them all into the same executable.
                    702: </p>
                    703: 
                    704: <h4>The <tt>%nonassoc</tt> directive</h4>
                    705: 
                    706: <p>This directive is used to assign non-associative precedence to
                    707: one or more terminal symbols.  See the section on precedence rules
                    708: or on the %left directive for additional information.</p>
                    709: 
                    710: <h4>The <tt>%parse_accept</tt> directive</h4>
                    711: 
                    712: <p>The %parse_accept directive specifies a block of C code that is
                    713: executed whenever the parser accepts its input string.  To ``accept''
                    714: an input string means that the parser was able to process all tokens
                    715: without error.</p>
                    716: 
                    717: <p>For example:</p>
                    718: 
                    719: <p><pre>
                    720:    %parse_accept {
                    721:       printf("parsing complete!\n");
                    722:    }
                    723: </pre></p>
                    724: 
                    725: 
                    726: <h4>The <tt>%parse_failure</tt> directive</h4>
                    727: 
                    728: <p>The %parse_failure directive specifies a block of C code that
                    729: is executed whenever the parser fails complete.  This code is not
                    730: executed until the parser has tried and failed to resolve an input
                    731: error using is usual error recovery strategy.  The routine is
                    732: only invoked when parsing is unable to continue.</p>
                    733: 
                    734: <p><pre>
                    735:    %parse_failure {
                    736:      fprintf(stderr,"Giving up.  Parser is hopelessly lost...\n");
                    737:    }
                    738: </pre></p>
                    739: 
                    740: <h4>The <tt>%right</tt> directive</h4>
                    741: 
                    742: <p>This directive is used to assign right-associative precedence to
                    743: one or more terminal symbols.  See the section on precedence rules
                    744: or on the %left directive for additional information.</p>
                    745: 
                    746: <h4>The <tt>%stack_overflow</tt> directive</h4>
                    747: 
                    748: <p>The %stack_overflow directive specifies a block of C code that
                    749: is executed if the parser's internal stack ever overflows.  Typically
                    750: this just prints an error message.  After a stack overflow, the parser
                    751: will be unable to continue and must be reset.</p>
                    752: 
                    753: <p><pre>
                    754:    %stack_overflow {
                    755:      fprintf(stderr,"Giving up.  Parser stack overflow\n");
                    756:    }
                    757: </pre></p>
                    758: 
                    759: <p>You can help prevent parser stack overflows by avoiding the use
                    760: of right recursion and right-precedence operators in your grammar.
                    761: Use left recursion and and left-precedence operators instead, to
                    762: encourage rules to reduce sooner and keep the stack size down.
                    763: For example, do rules like this:
                    764: <pre>
                    765:    list ::= list element.      // left-recursion.  Good!
                    766:    list ::= .
                    767: </pre>
                    768: Not like this:
                    769: <pre>
                    770:    list ::= element list.      // right-recursion.  Bad!
                    771:    list ::= .
                    772: </pre>
                    773: 
                    774: <h4>The <tt>%stack_size</tt> directive</h4>
                    775: 
                    776: <p>If stack overflow is a problem and you can't resolve the trouble
                    777: by using left-recursion, then you might want to increase the size
                    778: of the parser's stack using this directive.  Put an positive integer
                    779: after the %stack_size directive and Lemon will generate a parse
                    780: with a stack of the requested size.  The default value is 100.</p>
                    781: 
                    782: <p><pre>
                    783:    %stack_size 2000
                    784: </pre></p>
                    785: 
                    786: <h4>The <tt>%start_symbol</tt> directive</h4>
                    787: 
                    788: <p>By default, the start-symbol for the grammar that Lemon generates
                    789: is the first non-terminal that appears in the grammar file.  But you
                    790: can choose a different start-symbol using the %start_symbol directive.</p>
                    791: 
                    792: <p><pre>
                    793:    %start_symbol  prog
                    794: </pre></p>
                    795: 
                    796: <h4>The <tt>%token_destructor</tt> directive</h4>
                    797: 
                    798: <p>The %destructor directive assigns a destructor to a non-terminal
                    799: symbol.  (See the description of the %destructor directive above.)
                    800: This directive does the same thing for all terminal symbols.</p>
                    801: 
                    802: <p>Unlike non-terminal symbols which may each have a different data type
                    803: for their values, terminals all use the same data type (defined by
                    804: the %token_type directive) and so they use a common destructor.  Other
                    805: than that, the token destructor works just like the non-terminal
                    806: destructors.</p>
                    807: 
                    808: <h4>The <tt>%token_prefix</tt> directive</h4>
                    809: 
                    810: <p>Lemon generates #defines that assign small integer constants
                    811: to each terminal symbol in the grammar.  If desired, Lemon will
                    812: add a prefix specified by this directive
                    813: to each of the #defines it generates.
                    814: So if the default output of Lemon looked like this:
                    815: <pre>
                    816:     #define AND              1
                    817:     #define MINUS            2
                    818:     #define OR               3
                    819:     #define PLUS             4
                    820: </pre>
                    821: You can insert a statement into the grammar like this:
                    822: <pre>
                    823:     %token_prefix    TOKEN_
                    824: </pre>
                    825: to cause Lemon to produce these symbols instead:
                    826: <pre>
                    827:     #define TOKEN_AND        1
                    828:     #define TOKEN_MINUS      2
                    829:     #define TOKEN_OR         3
                    830:     #define TOKEN_PLUS       4
                    831: </pre>
                    832: 
                    833: <h4>The <tt>%token_type</tt> and <tt>%type</tt> directives</h4>
                    834: 
                    835: <p>These directives are used to specify the data types for values
                    836: on the parser's stack associated with terminal and non-terminal
                    837: symbols.  The values of all terminal symbols must be of the same
                    838: type.  This turns out to be the same data type as the 3rd parameter
                    839: to the Parse() function generated by Lemon.  Typically, you will
                    840: make the value of a terminal symbol by a pointer to some kind of
                    841: token structure.  Like this:</p>
                    842: 
                    843: <p><pre>
                    844:    %token_type    {Token*}
                    845: </pre></p>
                    846: 
                    847: <p>If the data type of terminals is not specified, the default value
                    848: is ``int''.</p>
                    849: 
                    850: <p>Non-terminal symbols can each have their own data types.  Typically
                    851: the data type  of a non-terminal is a pointer to the root of a parse-tree
                    852: structure that contains all information about that non-terminal.
                    853: For example:</p>
                    854: 
                    855: <p><pre>
                    856:    %type   expr  {Expr*}
                    857: </pre></p>
                    858: 
                    859: <p>Each entry on the parser's stack is actually a union containing
                    860: instances of all data types for every non-terminal and terminal symbol.
                    861: Lemon will automatically use the correct element of this union depending
                    862: on what the corresponding non-terminal or terminal symbol is.  But
                    863: the grammar designer should keep in mind that the size of the union
                    864: will be the size of its largest element.  So if you have a single
                    865: non-terminal whose data type requires 1K of storage, then your 100
                    866: entry parser stack will require 100K of heap space.  If you are willing
                    867: and able to pay that price, fine.  You just need to know.</p>
                    868: 
                    869: <h3>Error Processing</h3>
                    870: 
                    871: <p>After extensive experimentation over several years, it has been
                    872: discovered that the error recovery strategy used by yacc is about
                    873: as good as it gets.  And so that is what Lemon uses.</p>
                    874: 
                    875: <p>When a Lemon-generated parser encounters a syntax error, it
                    876: first invokes the code specified by the %syntax_error directive, if
                    877: any.  It then enters its error recovery strategy.  The error recovery
                    878: strategy is to begin popping the parsers stack until it enters a
                    879: state where it is permitted to shift a special non-terminal symbol
                    880: named ``error''.  It then shifts this non-terminal and continues
                    881: parsing.  But the %syntax_error routine will not be called again
                    882: until at least three new tokens have been successfully shifted.</p>
                    883: 
                    884: <p>If the parser pops its stack until the stack is empty, and it still
                    885: is unable to shift the error symbol, then the %parse_failed routine
                    886: is invoked and the parser resets itself to its start state, ready
                    887: to begin parsing a new file.  This is what will happen at the very
                    888: first syntax error, of course, if there are no instances of the 
                    889: ``error'' non-terminal in your grammar.</p>
                    890: 
                    891: </body>
                    892: </html>

FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>