void
yyerror( char *s )
{
if( incp )
printf( "%s: line %d: ", incp->fname, incp->line );
printf( "%s at %s\n", s, symdump( &yylval ) );
++anyerrors;
}
void yyerror( const char * s )
{
/* We use yylval instead of incp to access the error location information as
* the incp pointer will already be reset to 0 in case the error occurred at
* EOF.
*
* The two may differ only if we get an error while reading a lexical token
* spanning muliple lines, e.g. a multi-line string literal or action body,
* in which case yylval location information will hold the information about
* where this token started while incp will hold the information about where
* reading it broke.
*
* TODO: Test the theory about when yylval and incp location information are
* the same and when they differ.
*/
printf( "%s:%d: %s at %s\n", object_str( yylval.file ), yylval.line, s, symdump( &yylval ) );
++anyerrors;
}
int yylex()
{
int c;
char buf[ BIGGEST_TOKEN ];
char * b = buf;
if ( !incp )
goto eof;
/* Get first character (whitespace or of token). */
c = yychar();
if ( scanmode == SCAN_STRING )
{
/* If scanning for a string (action's {}'s), look for the closing brace.
* We handle matching braces, if they match.
*/
int nest = 1;
while ( ( c != EOF ) && ( b < buf + sizeof( buf ) ) )
{
if ( c == '{' )
++nest;
if ( ( c == '}' ) && !--nest )
break;
*b++ = c;
c = yychar();
/* Turn trailing "\r\n" sequences into plain "\n" for Cygwin. */
if ( ( c == '\n' ) && ( b[ -1 ] == '\r' ) )
--b;
}
/* We ate the ending brace -- regurgitate it. */
if ( c != EOF )
yyprev();
/* Check for obvious errors. */
if ( b == buf + sizeof( buf ) )
{
yyerror( "action block too big" );
goto eof;
}
if ( nest )
{
yyerror( "unmatched {} in action block" );
goto eof;
}
*b = 0;
yylval.type = STRING;
yylval.string = object_new( buf );
yylval.file = incp->fname;
yylval.line = incp->line;
}
else
{
char * b = buf;
struct keyword * k;
int inquote = 0;
int notkeyword;
/* Eat white space. */
for ( ;; )
{
/* Skip past white space. */
while ( ( c != EOF ) && isspace( c ) )
c = yychar();
/* Not a comment? */
if ( c != '#' )
break;
/* Swallow up comment line. */
while ( ( ( c = yychar() ) != EOF ) && ( c != '\n' ) ) ;
}
/* c now points to the first character of a token. */
if ( c == EOF )
goto eof;
yylval.file = incp->fname;
yylval.line = incp->line;
/* While scanning the word, disqualify it for (expensive) keyword lookup
* when we can: $anything, "anything", \anything
*/
notkeyword = c == '$';
/* Look for white space to delimit word. "'s get stripped but preserve
* white space. \ protects next character.
*/
while
(
( c != EOF ) &&
( b < buf + sizeof( buf ) ) &&
( inquote || !isspace( c ) )
)
{
if ( c == '"' )
{
/* begin or end " */
inquote = !inquote;
notkeyword = 1;
}
else if ( c != '\\' )
{
/* normal char */
*b++ = c;
}
else if ( ( c = yychar() ) != EOF )
{
/* \c */
if (c == 'n')
c = '\n';
else if (c == 'r')
c = '\r';
else if (c == 't')
c = '\t';
*b++ = c;
notkeyword = 1;
}
else
{
/* \EOF */
break;
}
c = yychar();
}
/* Check obvious errors. */
if ( b == buf + sizeof( buf ) )
{
yyerror( "string too big" );
goto eof;
}
if ( inquote )
{
yyerror( "unmatched \" in string" );
goto eof;
}
/* We looked ahead a character - back up. */
if ( c != EOF )
yyprev();
/* Scan token table. Do not scan if it is obviously not a keyword or if
* it is an alphabetic when were looking for punctuation.
*/
*b = 0;
yylval.type = ARG;
if ( !notkeyword && !( isalpha( *buf ) && ( scanmode == SCAN_PUNCT ) ) )
for ( k = keywords; k->word; ++k )
if ( ( *buf == *k->word ) && !strcmp( k->word, buf ) )
{
yylval.type = k->type;
yylval.keyword = k->word; /* used by symdump */
break;
}
if ( yylval.type == ARG )
yylval.string = object_new( buf );
}
if ( DEBUG_SCAN )
printf( "scan %s\n", symdump( &yylval ) );
return yylval.type;
eof:
/* We do not reset yylval.file & yylval.line here so unexpected EOF error
* messages would include correct error location information.
*/
yylval.type = EOF;
return yylval.type;
}
int
yylex()
{
int c;
char buf[BIGGEST_TOKEN];
char *b = buf;
if( !incp )
goto eof;
/* Get first character (whitespace or of token) */
c = yychar();
if( scanmode == SCAN_STRING )
{
/* If scanning for a string (action's {}'s), look for the */
/* closing brace. We handle matching braces, if they match! */
int nest = 1;
while( c != EOF && b < buf + sizeof( buf ) )
{
if( c == '{' )
nest++;
if( c == '}' && !--nest )
break;
*b++ = c;
c = yychar();
/* turn trailing "\r\n" sequences into plain "\n"
* for Cygwin
*/
if (c == '\n' && b[-1] == '\r')
--b;
}
/* We ate the ending brace -- regurgitate it. */
if( c != EOF )
yyprev();
/* Check obvious errors. */
if( b == buf + sizeof( buf ) )
{
yyerror( "action block too big" );
goto eof;
}
if( nest )
{
yyerror( "unmatched {} in action block" );
goto eof;
}
*b = 0;
yylval.type = STRING;
yylval.string = newstr( buf );
yylval.file = incp->fname;
yylval.line = incp->line;
}
else
{
char *b = buf;
struct keyword *k;
int inquote = 0;
int notkeyword;
/* Eat white space */
for( ;; )
{
/* Skip past white space */
while( c != EOF && isspace( c ) )
c = yychar();
/* Not a comment? Swallow up comment line. */
if( c != '#' )
break;
while( ( c = yychar() ) != EOF && c != '\n' )
;
}
/* c now points to the first character of a token. */
if( c == EOF )
goto eof;
yylval.file = incp->fname;
yylval.line = incp->line;
/* While scanning the word, disqualify it for (expensive) */
/* keyword lookup when we can: $anything, "anything", \anything */
notkeyword = c == '$';
/* look for white space to delimit word */
/* "'s get stripped but preserve white space */
/* \ protects next character */
while(
c != EOF &&
b < buf + sizeof( buf ) &&
( inquote || !isspace( c ) ) )
{
if( c == '"' )
{
/* begin or end " */
inquote = !inquote;
notkeyword = 1;
}
else if( c != '\\' )
{
/* normal char */
*b++ = c;
}
else if( ( c = yychar()) != EOF )
{
/* \c */
*b++ = c;
notkeyword = 1;
}
else
{
/* \EOF */
break;
}
c = yychar();
}
/* Check obvious errors. */
if( b == buf + sizeof( buf ) )
{
yyerror( "string too big" );
goto eof;
}
if( inquote )
{
yyerror( "unmatched \" in string" );
goto eof;
}
/* We looked ahead a character - back up. */
if( c != EOF )
yyprev();
/* scan token table */
/* don't scan if it's obviously not a keyword or if its */
/* an alphabetic when were looking for punctuation */
*b = 0;
yylval.type = ARG;
if( !notkeyword && !( isalpha( *buf ) && scanmode == SCAN_PUNCT ) )
{
for( k = keywords; k->word; k++ )
if( *buf == *k->word && !strcmp( k->word, buf ) )
{
yylval.type = k->type;
yylval.string = k->word; /* used by symdump */
break;
}
}
if( yylval.type == ARG )
yylval.string = newstr( buf );
}
if( DEBUG_SCAN )
printf( "scan %s\n", symdump( &yylval ) );
return yylval.type;
eof:
yylval.file = "end-of-input"; /* just in case */
yylval.line = 0;
yylval.type = EOF;
return yylval.type;
}
