2 /**********************************************************************
8 **********************************************************************/
10 /* The includes/config.h one */
15 #if defined(STDC_HEADERS) || defined(HAVE_STRING_H)
17 /* An ANSI string.h and pre-ANSI memory.h might conflict. */
18 #if !defined(STDC_HEADERS) && defined(HAVE_MEMORY_H)
20 #endif /* not STDC_HEADERS and HAVE_MEMORY_H */
22 #define rindex strrchr
23 #define bcopy(s, d, n) memcpy ((d), (s), (n))
24 #define bcmp(s1, s2, n) memcmp ((s1), (s2), (n))
25 #define bzero(s, n) memset ((s), 0, (n))
26 #else /* not STDC_HEADERS and not HAVE_STRING_H */
28 /* memory.h and strings.h conflict on some systems. */
29 #endif /* not STDC_HEADERS and not HAVE_STRING_H */
32 #include "hsparser.tab.h"
33 #include "constants.h"
36 /* Our substitute for <ctype.h> */
45 static unsigned char CharTable[NCHARS] = {
46 /* nul */ 0, 0, 0, 0, 0, 0, 0, 0,
47 /* bs */ 0, _S, _S, _S, _S, 0, 0, 0,
48 /* dle */ 0, 0, 0, 0, 0, 0, 0, 0,
49 /* can */ 0, 0, 0, 0, 0, 0, 0, 0,
50 /* sp */ _S, 0, 0, 0, 0, 0, 0, 0,
51 /* '(' */ _C, 0, 0, 0, 0, 0, 0, 0,
52 /* '0' */ _D|_H|_O,_D|_H|_O,_D|_H|_O,_D|_H|_O,_D|_H|_O,_D|_H|_O,_D|_H|_O,_D|_H|_O,
53 /* '8' */ _D|_H, _D|_H, _C, 0, 0, 0, 0, 0,
54 /* '@' */ 0, _H|_C, _H|_C, _H|_C, _H|_C, _H|_C, _H|_C, _C,
55 /* 'H' */ _C, _C, _C, _C, _C, _C, _C, _C,
56 /* 'P' */ _C, _C, _C, _C, _C, _C, _C, _C,
57 /* 'X' */ _C, _C, _C, _C, 0, 0, 0, 0,
58 /* '`' */ 0, _H, _H, _H, _H, _H, _H, 0,
59 /* 'h' */ 0, 0, 0, 0, 0, 0, 0, 0,
60 /* 'p' */ 0, 0, 0, 0, 0, 0, 0, 0,
61 /* 'x' */ 0, 0, 0, 0, 0, 0, 0, 0,
63 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
64 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
65 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
66 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
67 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
68 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
69 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
70 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
71 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
72 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
73 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
74 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
75 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
76 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
77 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
78 /* */ 0, 0, 0, 0, 0, 0, 0, 0,
84 return(CharTable[*s]&(_C));
87 /**********************************************************************
93 **********************************************************************/
95 char *input_filename = NULL; /* Always points to a dynamically allocated string */
98 * For my own sanity, things that are not part of the flex skeleton
99 * have been renamed as hsXXXXX rather than yyXXXXX. --JSM
102 static int hslineno = 0; /* Line number at end of token */
103 int hsplineno = 0; /* Line number at end of previous token */
105 static int hscolno = 0; /* Column number at end of token */
106 int hspcolno = 0; /* Column number at end of previous token */
107 static int hsmlcolno = 0; /* Column number for multiple-rule lexemes */
109 int modulelineno = -1; /* The line number where the module starts */
110 int startlineno = 0; /* The line number where something starts */
111 int endlineno = 0; /* The line number where something ends */
113 static BOOLEAN noGap = TRUE; /* For checking string gaps */
114 static BOOLEAN forgetindent = FALSE; /* Don't bother applying indentation rules */
116 static int nested_comments; /* For counting comment nesting depth */
118 /* OLD: Hacky definition of yywrap: see flex doc.
120 If we don't do this, then we'll have to get the default
121 yywrap from the flex library, which is often something
122 we are not good at locating. This avoids that difficulty.
123 (Besides which, this is the way old flexes (pre 2.4.x) did it.)
128 /* Essential forward declarations */
130 static void hsnewid PROTO((char *, int));
131 static void layout_input PROTO((char *, int));
132 static void cleartext (NO_ARGS);
133 static void addtext PROTO((char *, unsigned));
134 static void addchar PROTO((char));
135 static char *fetchtext PROTO((unsigned *));
136 static void new_filename PROTO((char *));
137 static int Return PROTO((int));
138 static void hsentercontext PROTO((int));
140 /* Special file handling for IMPORTS */
141 /* Note: imports only ever go *one deep* (hence no need for a stack) WDP 94/09 */
143 static YY_BUFFER_STATE hsbuf_save = NULL; /* Saved input buffer */
144 static char *filename_save; /* File Name */
145 static int hslineno_save = 0, /* Line Number */
146 hsplineno_save = 0, /* Line Number of Prev. token */
147 hscolno_save = 0, /* Indentation */
148 hspcolno_save = 0; /* Left Indentation */
149 static short icontexts_save = 0; /* Indent Context Level */
151 static BOOLEAN etags_save; /* saved: whether doing etags stuff or not */
152 extern BOOLEAN etags; /* that which is saved */
154 extern BOOLEAN nonstandardFlag; /* Glasgow extensions allowed */
156 static int hssttok = -1; /* Stacked Token: -1 -- no token; -ve -- ";"
157 * inserted before token +ve -- "}" inserted before
160 short icontexts = 0; /* Which context we're in */
163 Table of indentations: right bit indicates whether to use
164 indentation rules (1 = use rules; 0 = ignore)
167 push one of these "contexts" at every "case" or "where"; the right bit says
168 whether user supplied braces, etc., or not. pop appropriately (hsendindent).
170 ALSO, a push/pop when enter/exit a new file (e.g., on importing). A -1 is
171 pushed (the "column" for "module", "interface" and EOF). The -1 from the initial
172 push is shown just below.
177 static short indenttab[MAX_CONTEXTS] = {-1};
179 #define INDENTPT (indenttab[icontexts]>>1)
180 #define INDENTON (indenttab[icontexts]&1)
182 #define RETURN(tok) return(Return(tok))
185 #define YY_DECL int yylex1()
187 /* We should not peek at yy_act, but flex calls us even for the internal action
188 triggered on 'end-of-buffer' (This is not true of flex 2.4.4 and up, but
189 to support older versions of flex, we'll continue to peek for now.
191 #define YY_USER_ACTION \
192 if (yy_act != YY_END_OF_BUFFER) layout_input(yytext, yyleng);
196 #define YY_BREAK if (etags) fprintf(stderr,"%d %d / %d %d / %d\n",hsplineno,hspcolno,hslineno,hscolno,startlineno); break;
199 /* Each time we enter a new start state, we push it onto the state stack.
201 #define PUSH_STATE(n) yy_push_state(n)
202 #define POP_STATE yy_pop_state()
207 noyywrap (do not call yywrap on end of file; avoid use of -lfl)
208 never-interactive (to go a bit faster)
209 stack (use a start-condition stack)
213 %option never-interactive
216 /* The start states are:
217 Code -- normal Haskell code (principal lexer)
218 GlaExt -- Haskell code with Glasgow extensions
219 Comment -- Nested comment processing
220 String -- Inside a string literal with backslashes
221 StringEsc -- Immediately following a backslash in a string literal
222 Char -- Inside a character literal with backslashes
223 CharEsc -- Immediately following a backslash in a character literal
225 Note that the INITIAL state is unused. Also note that these states
226 are _exclusive_. All rules should be prefixed with an appropriate
227 list of start states.
230 %x Char CharEsc Code Comment GlaExt UserPragma String StringEsc
232 isoS [\xa1-\xbf\xd7\xf7]
233 isoL [\xc0-\xd6\xd8-\xde]
234 isol [\xdf-\xf6\xf8-\xff]
241 F {N}"."{N}(("e"|"E")("+"|"-")?{N})?
242 S [!#$%&*+./<=>?@\\^|\-~:\xa1-\xbf\xd7\xf7]
244 L [A-Z\xc0-\xd6\xd8-\xde]
245 l [a-z\xdf-\xf6\xf8-\xff]
250 CHAR [ !#$%&()*+,\-./0-9:;<=>?@A-Z\[\]^_`a-z{|}~\xa1-\xff]
259 * Simple comments and whitespace. Normally, we would just ignore these, but
260 * in case we're processing a string escape, we need to note that we've seen
263 * Note that we cater for a comment line that *doesn't* end in a newline.
264 * This is incorrect, strictly speaking, but seems like the right thing
265 * to do. Reported by Rajiv Mirani. (WDP 95/08)
267 * Hackily moved up here so that --<<EOF>> will match -- SOF 5/97
271 <Code,GlaExt,StringEsc>"--"[^\n\r]*{NL}?{WS}* |
272 <Code,GlaExt,UserPragma,StringEsc>{WS}+ { noGap = FALSE; }
276 * Special GHC pragma rules. Do we need a start state for interface files,
277 * so these won't be matched in source files? --JSM
283 /* I believe the next rule is not ever matched.
285 The '#line ' rule is un-cool, recognising a cpp directive inside hs source.
286 Driver has now been modified to output `standard' {-# LINE ..-} pragmas
287 where possible, so the lexer should now never see cpp directives
288 like '# ' and '#line'.
292 <Code,GlaExt>^"# ".*{NL} {
293 char tempf[FILENAME_SIZE];
294 sscanf(yytext+1, "%d \"%[^\"]", &hslineno, tempf);
296 hsplineno = hslineno; hscolno = 0; hspcolno = 0;
299 <Code,GlaExt>^"#line ".*{NL} {
300 char tempf[FILENAME_SIZE];
301 sscanf(yytext+5, "%d \"%[^\"]", &hslineno, tempf);
303 hsplineno = hslineno; hscolno = 0; hspcolno = 0;
308 <Code,GlaExt>"{-# LINE ".*"-}"{NL} {
309 /* partain: pragma-style line directive */
310 char tempf[FILENAME_SIZE];
311 sscanf(yytext+9, "%d \"%[^\"]", &hslineno, tempf);
313 hsplineno = hslineno; hscolno = 0; hspcolno = 0;
316 <Code,GlaExt>"{-#"{WS}*"INTERFACE" {
317 PUSH_STATE(UserPragma);
318 RETURN(INTERFACE_UPRAGMA);
320 <Code,GlaExt>"{-#"{WS}*"SPECIALI"[SZ]E {
321 PUSH_STATE(UserPragma);
322 RETURN(SPECIALISE_UPRAGMA);
324 <Code,GlaExt>"{-#"{WS}*"INLINE" {
325 PUSH_STATE(UserPragma);
326 RETURN(INLINE_UPRAGMA);
328 <Code,GlaExt>"{-#"{WS}*"NOINLINE" {
329 PUSH_STATE(UserPragma);
330 RETURN(NOINLINE_UPRAGMA);
332 <Code,GlaExt>"{-#"{WS}*"MAGIC_UNFOLDING" {
333 PUSH_STATE(UserPragma);
334 RETURN(MAGIC_UNFOLDING_UPRAGMA);
336 <Code,GlaExt>"{-#"{WS}*"GENERATE_SPECS" {
337 /* these are handled by hscpp */
341 <Code,GlaExt>"{-#"{WS}*"OPTIONS" {
342 /* these are for the driver! */
346 <Code,GlaExt>"{-#"{WS}*"SOURCE"{WS}*"#"?"-}" {
347 /* these are used by `make depend' and the
348 compiler to indicate that a module should
349 be imported from source */
351 RETURN(SOURCE_UPRAGMA);
354 <Code,GlaExt>"{-#"{WS}*[A-Z_]+ {
355 fprintf(stderr, "%s:%d: Warning: Unrecognised pragma '",
356 input_filename, hsplineno);
357 format_string(stderr, (unsigned char *) yytext, yyleng);
358 fputs("'\n", stderr);
362 <UserPragma>"#-}" { POP_STATE; RETURN(END_UPRAGMA); }
366 * Haskell keywords. `scc' is actually a Glasgow extension, but it is
367 * intentionally accepted as a keyword even for normal <Code>.
371 <Code,GlaExt>"case" { RETURN(CASE); }
372 <Code,GlaExt>"class" { RETURN(CLASS); }
373 <Code,GlaExt,UserPragma>"data" { RETURN(DATA); }
374 <Code,GlaExt>"default" { RETURN(DEFAULT); }
375 <Code,GlaExt>"deriving" { RETURN(DERIVING); }
376 <Code,GlaExt>"do" { RETURN(DO); }
377 <Code,GlaExt>"else" { RETURN(ELSE); }
378 <Code,GlaExt>"if" { RETURN(IF); }
379 <Code,GlaExt>"import" { RETURN(IMPORT); }
380 <Code,GlaExt>"in" { RETURN(IN); }
381 <Code,GlaExt>"infix" { RETURN(INFIX); }
382 <Code,GlaExt>"infixl" { RETURN(INFIXL); }
383 <Code,GlaExt>"infixr" { RETURN(INFIXR); }
384 <Code,GlaExt,UserPragma>"instance" { RETURN(INSTANCE); }
385 <Code,GlaExt>"let" { RETURN(LET); }
386 <Code,GlaExt>"module" { RETURN(MODULE); }
387 <Code,GlaExt>"newtype" { RETURN(NEWTYPE); }
388 <Code,GlaExt>"of" { RETURN(OF); }
389 <Code,GlaExt>"then" { RETURN(THEN); }
390 <Code,GlaExt>"type" { RETURN(TYPE); }
391 <Code,GlaExt>"where" { RETURN(WHERE); }
393 <Code,GlaExt>"as" { RETURN(AS); }
394 <Code,GlaExt>"hiding" { RETURN(HIDING); }
395 <Code,GlaExt>"qualified" { RETURN(QUALIFIED); }
397 <Code,GlaExt>"_scc_" { RETURN(SCC); }
398 <GlaExt>"_ccall_" { RETURN(CCALL); }
399 <GlaExt>"_ccall_GC_" { RETURN(CCALL_GC); }
400 <GlaExt>"_casm_" { RETURN(CASM); }
401 <GlaExt>"_casm_GC_" { RETURN(CASM_GC); }
402 <GlaExt>"foreign" { RETURN(FOREIGN); }
403 <GlaExt>"export" { RETURN(EXPORT); }
404 <GlaExt>"label" { RETURN(LABEL); }
405 <GlaExt>"unsafe" { RETURN(UNSAFE); }
406 <GlaExt>"_stdcall" { RETURN(STDCALL); }
407 <GlaExt>"_ccall" { RETURN(C_CALL); }
408 <GlaExt>"_pascal" { RETURN(PASCAL); }
409 <GlaExt>"_fastcall" { RETURN(FASTCALL); }
410 <GlaExt>"dynamic" { RETURN(DYNAMIC); }
414 * Haskell operators: special, reservedops and useful varsyms
418 <Code,GlaExt,UserPragma>"(" { RETURN(OPAREN); }
419 <Code,GlaExt,UserPragma>")" { RETURN(CPAREN); }
420 <Code,GlaExt,UserPragma>"[" { RETURN(OBRACK); }
421 <Code,GlaExt,UserPragma>"]" { RETURN(CBRACK); }
422 <Code,GlaExt>"{" { RETURN(OCURLY); }
423 <Code,GlaExt>"}" { RETURN(CCURLY); }
424 <Code,GlaExt,UserPragma>"," { RETURN(COMMA); }
425 <Code,GlaExt>";" { RETURN(SEMI); }
426 <Code,GlaExt>"`" { RETURN(BQUOTE); }
427 <Code,GlaExt>"_" { RETURN(WILDCARD); }
429 <Code,GlaExt>".." { RETURN(DOTDOT); }
430 <Code,GlaExt,UserPragma>"::" { RETURN(DCOLON); }
431 <Code,GlaExt,UserPragma>"=" { RETURN(EQUAL); }
432 <Code,GlaExt>"\\" { RETURN(LAMBDA); }
433 <Code,GlaExt>"|" { RETURN(VBAR); }
434 <Code,GlaExt>"<-" { RETURN(LARROW); }
435 <Code,GlaExt,UserPragma>"->" { RETURN(RARROW); }
436 <Code,GlaExt>"-" { RETURN(MINUS); }
437 <Code,GlaExt>"+" { RETURN(PLUS); }
439 <Code,GlaExt,UserPragma>"=>" { RETURN(DARROW); }
440 <Code,GlaExt>"@" { RETURN(AT); }
441 <Code,GlaExt>"!" { RETURN(BANG); }
442 <Code,GlaExt>"~" { RETURN(LAZY); }
446 * Integers and (for Glasgow extensions) primitive integers. Note that
447 * we pass all of the text on to the parser, because flex/C can't handle
448 * arbitrary precision numbers.
452 <GlaExt>("-")?"0"[Oo]{O}+"#" { /* octal */
453 yylval.uid = xstrndup(yytext, yyleng - 1);
456 <Code,GlaExt>"0"[Oo]{O}+ { /* octal */
457 yylval.uid = xstrndup(yytext, yyleng);
460 <GlaExt>("-")?"0"[Xx]{H}+"#" { /* hexadecimal */
461 yylval.uid = xstrndup(yytext, yyleng - 1);
464 <Code,GlaExt>"0"[Xx]{H}+ { /* hexadecimal */
465 yylval.uid = xstrndup(yytext, yyleng);
468 <GlaExt>("-")?{N}"#" {
469 yylval.uid = xstrndup(yytext, yyleng - 1);
472 <Code,GlaExt,UserPragma>{N} {
473 yylval.uid = xstrndup(yytext, yyleng);
479 * Floats and (for Glasgow extensions) primitive floats/doubles.
483 <GlaExt>("-")?{F}"##" {
484 yylval.uid = xstrndup(yytext, yyleng - 2);
487 <GlaExt>("-")?{F}"#" {
488 yylval.uid = xstrndup(yytext, yyleng - 1);
492 yylval.uid = xstrndup(yytext, yyleng);
498 * Funky ``foo'' style C literals for Glasgow extensions
502 <GlaExt>"``"[^']+"''" {
503 hsnewid(yytext + 2, yyleng - 4);
509 * Identifiers, both variables and operators. The trailing hash is allowed
510 * for Glasgow extensions.
516 /* These SHOULDNAE work in "Code" (sigh) */
518 <GlaExt,UserPragma>{Id}"#" {
519 if (! nonstandardFlag) {
520 char errbuf[ERR_BUF_SIZE];
521 sprintf(errbuf, "Non-standard identifier (trailing `#'): %s\n", yytext);
524 hsnewid(yytext, yyleng);
525 RETURN(isconstr(yytext) ? CONID : VARID);
527 <Code,GlaExt,UserPragma>{Id} {
528 hsnewid(yytext, yyleng);
529 RETURN(isconstr(yytext) ? CONID : VARID);
531 <Code,GlaExt,UserPragma>{SId} {
532 hsnewid(yytext, yyleng);
533 RETURN(isconstr(yytext) ? CONSYM : VARSYM);
535 <Code,GlaExt,UserPragma>{Mod}"."{Id}"#" {
537 if (! nonstandardFlag) {
538 char errbuf[ERR_BUF_SIZE];
539 sprintf(errbuf, "Non-standard identifier (trailing `#'): %s\n", yytext);
542 is_constr = hsnewqid(yytext, yyleng);
543 RETURN(is_constr ? QCONID : QVARID);
545 <Code,GlaExt,UserPragma>{Mod}"."{Id} {
546 BOOLEAN is_constr = hsnewqid(yytext, yyleng);
547 RETURN(is_constr ? QCONID : QVARID);
549 <Code,GlaExt,UserPragma>{Mod}"."{SId} {
550 BOOLEAN is_constr = hsnewqid(yytext, yyleng);
551 RETURN(is_constr ? QCONSYM : QVARSYM);
555 /* Why is `{Id}#` matched this way, and `{Id}` lexed as three tokens? --JSM */
557 /* Because we can make the former well-behaved (we defined them).
559 Sadly, the latter is defined by Haskell, which allows such
560 la-la land constructs as `{-a 900-line comment-} foo`. (WDP 94/12)
564 <GlaExt,UserPragma>"`"{Id}"#`" {
565 hsnewid(yytext + 1, yyleng - 2);
566 RETURN(isconstr(yytext+1) ? CONSYM : VARSYM);
571 * Character literals. The first form is the quick form, for character
572 * literals that don't contain backslashes. Literals with backslashes are
573 * lexed through multiple rules. First, we match the open ' and as many
574 * normal characters as possible. This puts us into the <Char> state, where
575 * a backslash is legal. Then, we match the backslash and move into the
576 * <CharEsc> state. When we drop out of <CharEsc>, we collect more normal
577 * characters and the close '. We may end up with too many characters, but
578 * this allows us to easily share the lex rules with strings. Excess characters
579 * are ignored with a warning.
583 <GlaExt>'({CHAR}|"\"")"'#" {
584 yylval.uhstring = installHstring(1, yytext+1);
587 <Code,GlaExt>'({CHAR}|"\"")' {
588 yylval.uhstring = installHstring(1, yytext+1);
591 <Code,GlaExt>'' {char errbuf[ERR_BUF_SIZE];
592 sprintf(errbuf, "'' is not a valid character (or string) literal\n");
595 <Code,GlaExt>'({CHAR}|"\"")* {
596 hsmlcolno = hspcolno;
598 addtext(yytext+1, yyleng-1);
601 <Char>({CHAR}|"\"")*'# {
605 addtext(yytext, yyleng - 2);
606 text = fetchtext(&length);
608 if (! nonstandardFlag) {
609 char errbuf[ERR_BUF_SIZE];
610 sprintf(errbuf, "`Char-hash' literals are non-standard: %s\n", text);
615 fprintf(stderr, "%s:%d:%d: Unboxed character literal '",
616 input_filename, hsplineno, hspcolno + 1);
617 format_string(stderr, (unsigned char *) text, length);
618 fputs("' too long\n", stderr);
621 yylval.uhstring = installHstring(1, text);
622 hspcolno = hsmlcolno;
626 <Char>({CHAR}|"\"")*' {
630 addtext(yytext, yyleng - 1);
631 text = fetchtext(&length);
634 fprintf(stderr, "%s:%d:%d: Character literal '",
635 input_filename, hsplineno, hspcolno + 1);
636 format_string(stderr, (unsigned char *) text, length);
637 fputs("' too long\n", stderr);
640 yylval.uhstring = installHstring(1, text);
641 hspcolno = hsmlcolno;
645 <Char>({CHAR}|"\"")+ { addtext(yytext, yyleng); }
650 * String literals. The first form is the quick form, for string literals
651 * that don't contain backslashes. Literals with backslashes are lexed
652 * through multiple rules. First, we match the open " and as many normal
653 * characters as possible. This puts us into the <String> state, where
654 * a backslash is legal. Then, we match the backslash and move into the
655 * <StringEsc> state. When we drop out of <StringEsc>, we collect more normal
656 * characters, moving back and forth between <String> and <StringEsc> as more
657 * backslashes are encountered. (We may even digress into <Comment> mode if we
658 * find a comment in a gap between backslashes.) Finally, we read the last chunk
659 * of normal characters and the close ".
663 <GlaExt>"\""({CHAR}|"'")*"\""# {
664 yylval.uhstring = installHstring(yyleng-3, yytext+1);
665 /* the -3 accounts for the " on front, "# on the end */
668 <Code,GlaExt>"\""({CHAR}|"'")*"\"" {
669 yylval.uhstring = installHstring(yyleng-2, yytext+1);
672 <Code,GlaExt>"\""({CHAR}|"'")* {
673 hsmlcolno = hspcolno;
675 addtext(yytext+1, yyleng-1);
678 <String>({CHAR}|"'")*"\"#" {
682 addtext(yytext, yyleng-2);
683 text = fetchtext(&length);
685 if (! nonstandardFlag) {
686 char errbuf[ERR_BUF_SIZE];
687 sprintf(errbuf, "`String-hash' literals are non-standard: %s\n", text);
691 yylval.uhstring = installHstring(length, text);
692 hspcolno = hsmlcolno;
696 <String>({CHAR}|"'")*"\"" {
700 addtext(yytext, yyleng-1);
701 text = fetchtext(&length);
703 yylval.uhstring = installHstring(length, text);
704 hspcolno = hsmlcolno;
708 <String>({CHAR}|"'")+ { addtext(yytext, yyleng); }
712 * Character and string escapes are roughly the same, but strings have the
713 * extra `\&' sequence which is not allowed for characters. Also, comments
714 * are allowed in the <StringEsc> state. (See the comment section much
717 * NB: Backslashes and tabs are stored in strings as themselves.
718 * But if we print them (in printtree.c), they must go out as
719 * "\\\\" and "\\t" respectively. (This is because of the bogus
720 * intermediate format that the parser produces. It uses '\t' fpr end of
721 * string, so it needs to be able to escape tabs, which means that it
722 * also needs to be able to escape the escape character ('\\'). Sigh.
726 <Char>\\ { PUSH_STATE(CharEsc); }
727 <String>\\& /* Ignore */ ;
728 <String>\\ { PUSH_STATE(StringEsc); noGap = TRUE; }
730 <CharEsc>\\ { addchar(*yytext); POP_STATE; }
731 <StringEsc>\\ { if (noGap) { addchar(*yytext); } POP_STATE; }
733 <CharEsc,StringEsc>["'] { addchar(*yytext); POP_STATE; }
734 <CharEsc,StringEsc>NUL { addchar('\000'); POP_STATE; }
735 <CharEsc,StringEsc>SOH { addchar('\001'); POP_STATE; }
736 <CharEsc,StringEsc>STX { addchar('\002'); POP_STATE; }
737 <CharEsc,StringEsc>ETX { addchar('\003'); POP_STATE; }
738 <CharEsc,StringEsc>EOT { addchar('\004'); POP_STATE; }
739 <CharEsc,StringEsc>ENQ { addchar('\005'); POP_STATE; }
740 <CharEsc,StringEsc>ACK { addchar('\006'); POP_STATE; }
741 <CharEsc,StringEsc>BEL |
742 <CharEsc,StringEsc>a { addchar('\007'); POP_STATE; }
743 <CharEsc,StringEsc>BS |
744 <CharEsc,StringEsc>b { addchar('\010'); POP_STATE; }
745 <CharEsc,StringEsc>HT |
746 <CharEsc,StringEsc>t { addchar('\011'); POP_STATE; }
747 <CharEsc,StringEsc>LF |
748 <CharEsc,StringEsc>n { addchar('\012'); POP_STATE; }
749 <CharEsc,StringEsc>VT |
750 <CharEsc,StringEsc>v { addchar('\013'); POP_STATE; }
751 <CharEsc,StringEsc>FF |
752 <CharEsc,StringEsc>f { addchar('\014'); POP_STATE; }
753 <CharEsc,StringEsc>CR |
754 <CharEsc,StringEsc>r { addchar('\015'); POP_STATE; }
755 <CharEsc,StringEsc>SO { addchar('\016'); POP_STATE; }
756 <CharEsc,StringEsc>SI { addchar('\017'); POP_STATE; }
757 <CharEsc,StringEsc>DLE { addchar('\020'); POP_STATE; }
758 <CharEsc,StringEsc>DC1 { addchar('\021'); POP_STATE; }
759 <CharEsc,StringEsc>DC2 { addchar('\022'); POP_STATE; }
760 <CharEsc,StringEsc>DC3 { addchar('\023'); POP_STATE; }
761 <CharEsc,StringEsc>DC4 { addchar('\024'); POP_STATE; }
762 <CharEsc,StringEsc>NAK { addchar('\025'); POP_STATE; }
763 <CharEsc,StringEsc>SYN { addchar('\026'); POP_STATE; }
764 <CharEsc,StringEsc>ETB { addchar('\027'); POP_STATE; }
765 <CharEsc,StringEsc>CAN { addchar('\030'); POP_STATE; }
766 <CharEsc,StringEsc>EM { addchar('\031'); POP_STATE; }
767 <CharEsc,StringEsc>SUB { addchar('\032'); POP_STATE; }
768 <CharEsc,StringEsc>ESC { addchar('\033'); POP_STATE; }
769 <CharEsc,StringEsc>FS { addchar('\034'); POP_STATE; }
770 <CharEsc,StringEsc>GS { addchar('\035'); POP_STATE; }
771 <CharEsc,StringEsc>RS { addchar('\036'); POP_STATE; }
772 <CharEsc,StringEsc>US { addchar('\037'); POP_STATE; }
773 <CharEsc,StringEsc>SP { addchar('\040'); POP_STATE; }
774 <CharEsc,StringEsc>DEL { addchar('\177'); POP_STATE; }
775 <CharEsc,StringEsc>"^"{CNTRL} { char c = yytext[1] - '@'; addchar(c); POP_STATE; }
776 <CharEsc,StringEsc>{D}+ {
777 int i = strtol(yytext, NULL, 10);
781 char errbuf[ERR_BUF_SIZE];
782 sprintf(errbuf, "Numeric escape \"\\%s\" out of range\n",
788 <CharEsc,StringEsc>o{O}+ {
789 int i = strtol(yytext + 1, NULL, 8);
793 char errbuf[ERR_BUF_SIZE];
794 sprintf(errbuf, "Numeric escape \"\\%s\" out of range\n",
800 <CharEsc,StringEsc>x{H}+ {
801 int i = strtol(yytext + 1, NULL, 16);
805 char errbuf[ERR_BUF_SIZE];
806 sprintf(errbuf, "Numeric escape \"\\%s\" out of range\n",
816 * Nested comments. The major complication here is in trying to match the
817 * longest lexemes possible, for better performance. (See the flex document.)
818 * That's why the rules look so bizarre.
822 <Code,GlaExt,UserPragma,StringEsc>"{-" {
823 noGap = FALSE; nested_comments = 1; PUSH_STATE(Comment);
827 <Comment>"-"+[^-{}]+ |
828 <Comment>"{"+[^-{}]+ ;
829 <Comment>"{-" { nested_comments++; }
830 <Comment>"-}" { if (--nested_comments == 0) POP_STATE; }
835 * Illegal characters. This used to be a single rule, but we might as well
836 * pass on as much information as we have, so now we indicate our state in
841 <INITIAL,Code,GlaExt,UserPragma>(.|\n) {
842 fprintf(stderr, "%s:%d:%d: Illegal character: `",
843 input_filename, hsplineno, hspcolno + 1);
844 format_string(stderr, (unsigned char *) yytext, 1);
845 fputs("'\n", stderr);
849 fprintf(stderr, "%s:%d:%d: Illegal character: `",
850 input_filename, hsplineno, hspcolno + 1);
851 format_string(stderr, (unsigned char *) yytext, 1);
852 fputs("' in a character literal\n", stderr);
856 fprintf(stderr, "%s:%d:%d: Illegal character escape: `\\",
857 input_filename, hsplineno, hspcolno + 1);
858 format_string(stderr, (unsigned char *) yytext, 1);
859 fputs("'\n", stderr);
862 <String>(.|\n) { if (nonstandardFlag) {
863 addtext(yytext, yyleng);
865 fprintf(stderr, "%s:%d:%d: Illegal character: `",
866 input_filename, hsplineno, hspcolno + 1);
867 format_string(stderr, (unsigned char *) yytext, 1);
868 fputs("' in a string literal\n", stderr);
874 fprintf(stderr, "%s:%d:%d: Illegal string escape: `\\",
875 input_filename, hsplineno, hspcolno + 1);
876 format_string(stderr, (unsigned char *) yytext, 1);
877 fputs("'\n", stderr);
880 fprintf(stderr, "%s:%d:%d: Illegal character: `",
881 input_filename, hsplineno, hspcolno + 1);
882 format_string(stderr, (unsigned char *) yytext, 1);
883 fputs("' in a string gap\n", stderr);
890 * End of file. In any sub-state, this is an error. However, for the primary
891 * <Code> and <GlaExt> states, this is perfectly normal. We just return an EOF
892 * and let the yylex() wrapper deal with whatever has to be done next (e.g.
893 * adding virtual close curlies, or closing an interface and returning to the
894 * primary source file.
896 * Note that flex does not call YY_USER_ACTION for <<EOF>> rules. Hence the
897 * line/column advancement has to be done by hand.
901 <Char,CharEsc><<EOF>> {
902 hsplineno = hslineno; hspcolno = hscolno;
903 hsperror("unterminated character literal");
906 hsplineno = hslineno; hspcolno = hscolno;
907 hsperror("unterminated comment");
909 <String,StringEsc><<EOF>> {
910 hsplineno = hslineno; hspcolno = hscolno;
911 hsperror("unterminated string literal");
913 <UserPragma><<EOF>> {
914 hsplineno = hslineno; hspcolno = hscolno;
915 hsperror("unterminated user-specified pragma");
917 <Code,GlaExt><<EOF>> { hsplineno = hslineno; hspcolno = hscolno; return(EOF); }
921 /**********************************************************************
924 * YACC/LEX Initialisation etc. *
927 **********************************************************************/
930 We initialise input_filename to "<stdin>".
931 This allows unnamed sources to be piped into the parser.
937 input_filename = xstrdup("<stdin>");
939 /* We must initialize the input buffer _now_, because we call
940 setyyin _before_ calling yylex for the first time! */
941 yy_switch_to_buffer(yy_create_buffer(stdin, YY_BUF_SIZE));
950 new_filename(char *f) /* This looks pretty dodgy to me (WDP) */
952 if (input_filename != NULL)
953 free(input_filename);
954 input_filename = xstrdup(f);
957 /**********************************************************************
960 * Layout Processing *
963 **********************************************************************/
966 The following section deals with Haskell Layout conventions
967 forcing insertion of ; or } as appropriate
973 return (!forgetindent && INDENTON);
977 /* Enter new context and set new indentation level */
982 fprintf(stderr, "hssetindent:hscolno=%d,hspcolno=%d,INDENTPT[%d]=%d\n", hscolno, hspcolno, icontexts, INDENTPT);
986 * partain: first chk that new indent won't be less than current one; this code
987 * doesn't make sense to me; hscolno tells the position of the _end_ of the
988 * current token; what that has to do with indenting, I don't know.
992 if (hscolno - 1 <= INDENTPT) {
994 return; /* Empty input OK for Haskell 1.1 */
996 char errbuf[ERR_BUF_SIZE];
998 sprintf(errbuf, "Layout error -- indentation should be > %d cols", INDENTPT);
1002 hsentercontext((hspcolno << 1) | 1);
1006 /* Enter a new context without changing the indentation level */
1011 fprintf(stderr, "hsincindent:hscolno=%d,hspcolno=%d,INDENTPT[%d]=%d\n", hscolno, hspcolno, icontexts, INDENTPT);
1013 hsentercontext(indenttab[icontexts] & ~1);
1017 /* Turn off indentation processing, usually because an explicit "{" has been seen */
1021 forgetindent = TRUE;
1025 /* Enter a new layout context. */
1027 hsentercontext(int indent)
1029 /* Enter new context and set indentation as specified */
1030 if (++icontexts >= MAX_CONTEXTS) {
1031 char errbuf[ERR_BUF_SIZE];
1033 sprintf(errbuf, "`wheres' and `cases' nested too deeply (>%d)", MAX_CONTEXTS - 1);
1036 forgetindent = FALSE;
1037 indenttab[icontexts] = indent;
1039 fprintf(stderr, "hsentercontext:indent=%d,hscolno=%d,hspcolno=%d,INDENTPT[%d]=%d\n", indent, hscolno, hspcolno, icontexts, INDENTPT);
1044 /* Exit a layout context */
1050 fprintf(stderr, "hsendindent:hscolno=%d,hspcolno=%d,INDENTPT[%d]=%d\n", hscolno, hspcolno, icontexts, INDENTPT);
1055 * Return checks the indentation level and returns ;, } or the specified token.
1065 if (hsshouldindent()) {
1066 if (hspcolno < INDENTPT) {
1068 fprintf(stderr, "inserted '}' before %d (%d:%d:%d:%d)\n", tok, hspcolno, hscolno, yyleng, INDENTPT);
1072 } else if (hspcolno == INDENTPT) {
1074 fprintf(stderr, "inserted ';' before %d (%d:%d)\n", tok, hspcolno, INDENTPT);
1082 fprintf(stderr, "returning %d (%d:%d)\n", tok, hspcolno, INDENTPT);
1089 * Redefine yylex to check for stacked tokens, yylex1() is the original yylex()
1095 static BOOLEAN eof = FALSE;
1098 if (hssttok != -1) {
1106 endlineno = hslineno;
1107 if ((tok = yylex1()) != EOF)
1113 if (icontexts > icontexts_save) {
1116 indenttab[icontexts] = 0;
1119 hsperror("missing '}' at end of file");
1120 } else if (hsbuf_save != NULL) {
1122 yy_delete_buffer(YY_CURRENT_BUFFER);
1123 yy_switch_to_buffer(hsbuf_save);
1125 new_filename(filename_save);
1126 free(filename_save);
1127 hslineno = hslineno_save;
1128 hsplineno = hsplineno_save;
1129 hscolno = hscolno_save;
1130 hspcolno = hspcolno_save;
1132 icontexts = icontexts_save - 1;
1135 fprintf(stderr, "finished reading interface (%d:%d:%d)\n", hscolno, hspcolno, INDENTPT);
1140 hsperror("No longer using yacc to parse interface files");
1145 abort(); /* should never get here! */
1149 /**********************************************************************
1152 * Input Processing for Interfaces -- Not currently used !!! *
1155 **********************************************************************/
1157 /* setyyin(file) open file as new lex input buffer */
1163 hsbuf_save = YY_CURRENT_BUFFER;
1164 if ((yyin = fopen(file, "r")) == NULL) {
1165 char errbuf[ERR_BUF_SIZE];
1167 sprintf(errbuf, "can't read \"%-.50s\"", file);
1170 yy_switch_to_buffer(yy_create_buffer(yyin, YY_BUF_SIZE));
1172 hslineno_save = hslineno;
1173 hsplineno_save = hsplineno;
1174 hslineno = hsplineno = 1;
1176 filename_save = input_filename;
1177 input_filename = NULL;
1179 hscolno_save = hscolno;
1180 hspcolno_save = hspcolno;
1181 hscolno = hspcolno = 0;
1182 etags_save = etags; /* do not do "etags" stuff in interfaces */
1183 etags = 0; /* We remember whether we are doing it in
1184 the module, so we can restore it later [WDP 94/09] */
1185 hsentercontext(-1); /* partain: changed this from 0 */
1186 icontexts_save = icontexts;
1188 fprintf(stderr, "reading %s (%d:%d:%d)\n", input_filename, hscolno_save, hspcolno_save, INDENTPT);
1193 layout_input(char *text, int len)
1196 fprintf(stderr, "Scanning \"%s\"\n", text);
1199 hsplineno = hslineno;
1211 hscolno += 8 - (hscolno % 8); /* Tabs stops are 8 columns apart */
1223 setstartlineno(void)
1225 startlineno = hsplineno;
1227 if (modulelineno == 0) {
1228 modulelineno = startlineno;
1234 fprintf(stderr,"%u\tsetstartlineno (col %u)\n",startlineno,hscolno);
1238 /**********************************************************************
1244 **********************************************************************/
1246 #define CACHE_SIZE YY_BUF_SIZE
1252 } textcache = { 0, 0, NULL };
1257 /* fprintf(stderr, "cleartext\n"); */
1259 if (textcache.allocated == 0) {
1260 textcache.allocated = CACHE_SIZE;
1261 textcache.text = xmalloc(CACHE_SIZE);
1266 addtext(char *text, unsigned length)
1268 /* fprintf(stderr, "addtext: %d %s\n", length, text); */
1273 if (textcache.next + length + 1 >= textcache.allocated) {
1274 textcache.allocated += length + CACHE_SIZE;
1275 textcache.text = xrealloc(textcache.text, textcache.allocated);
1277 bcopy(text, textcache.text + textcache.next, length);
1278 textcache.next += length;
1284 /* fprintf(stderr, "addchar: %c\n", c); */
1286 if (textcache.next + 2 >= textcache.allocated) {
1287 textcache.allocated += CACHE_SIZE;
1288 textcache.text = xrealloc(textcache.text, textcache.allocated);
1290 textcache.text[textcache.next++] = c;
1294 fetchtext(unsigned *length)
1296 /* fprintf(stderr, "fetchtext: %d\n", textcache.next); */
1298 *length = textcache.next;
1299 textcache.text[textcache.next] = '\0';
1300 return textcache.text;
1303 /**********************************************************************
1306 * Identifier Processing *
1309 **********************************************************************/
1312 hsnewid Enters an id of length n into the symbol table.
1316 hsnewid(char *name, int length)
1318 char save = name[length];
1320 name[length] = '\0';
1321 yylval.uid = installid(name);
1322 name[length] = save;
1326 hsnewqid(char *name, int length)
1329 char save = name[length];
1330 name[length] = '\0';
1332 dot = strchr(name, '.');
1334 yylval.uqid = mkaqual(installid(name),installid(dot+1));
1336 name[length] = save;
1338 return isconstr(dot+1);