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>"unsafe" { RETURN(UNSAFE); }
405 <GlaExt>"_stdcall" { RETURN(STDCALL); }
406 <GlaExt>"_ccall" { RETURN(C_CALL); }
407 <GlaExt>"_pascal" { RETURN(PASCAL); }
408 <GlaExt>"_fastcall" { RETURN(FASTCALL); }
409 <GlaExt>"dynamic" { RETURN(DYNAMIC); }
413 * Haskell operators: special, reservedops and useful varsyms
417 <Code,GlaExt,UserPragma>"(" { RETURN(OPAREN); }
418 <Code,GlaExt,UserPragma>")" { RETURN(CPAREN); }
419 <Code,GlaExt,UserPragma>"[" { RETURN(OBRACK); }
420 <Code,GlaExt,UserPragma>"]" { RETURN(CBRACK); }
421 <Code,GlaExt>"{" { RETURN(OCURLY); }
422 <Code,GlaExt>"}" { RETURN(CCURLY); }
423 <Code,GlaExt,UserPragma>"," { RETURN(COMMA); }
424 <Code,GlaExt>";" { RETURN(SEMI); }
425 <Code,GlaExt>"`" { RETURN(BQUOTE); }
426 <Code,GlaExt>"_" { RETURN(WILDCARD); }
428 <Code,GlaExt>".." { RETURN(DOTDOT); }
429 <Code,GlaExt,UserPragma>"::" { RETURN(DCOLON); }
430 <Code,GlaExt,UserPragma>"=" { RETURN(EQUAL); }
431 <Code,GlaExt>"\\" { RETURN(LAMBDA); }
432 <Code,GlaExt>"|" { RETURN(VBAR); }
433 <Code,GlaExt>"<-" { RETURN(LARROW); }
434 <Code,GlaExt,UserPragma>"->" { RETURN(RARROW); }
435 <Code,GlaExt>"-" { RETURN(MINUS); }
436 <Code,GlaExt>"+" { RETURN(PLUS); }
438 <Code,GlaExt,UserPragma>"=>" { RETURN(DARROW); }
439 <Code,GlaExt>"@" { RETURN(AT); }
440 <Code,GlaExt>"!" { RETURN(BANG); }
441 <Code,GlaExt>"~" { RETURN(LAZY); }
445 * Integers and (for Glasgow extensions) primitive integers. Note that
446 * we pass all of the text on to the parser, because flex/C can't handle
447 * arbitrary precision numbers.
451 <GlaExt>("-")?"0"[Oo]{O}+"#" { /* octal */
452 yylval.uid = xstrndup(yytext, yyleng - 1);
455 <Code,GlaExt>"0"[Oo]{O}+ { /* octal */
456 yylval.uid = xstrndup(yytext, yyleng);
459 <GlaExt>("-")?"0"[Xx]{H}+"#" { /* hexadecimal */
460 yylval.uid = xstrndup(yytext, yyleng - 1);
463 <Code,GlaExt>"0"[Xx]{H}+ { /* hexadecimal */
464 yylval.uid = xstrndup(yytext, yyleng);
467 <GlaExt>("-")?{N}"#" {
468 yylval.uid = xstrndup(yytext, yyleng - 1);
471 <Code,GlaExt,UserPragma>{N} {
472 yylval.uid = xstrndup(yytext, yyleng);
478 * Floats and (for Glasgow extensions) primitive floats/doubles.
482 <GlaExt>("-")?{F}"##" {
483 yylval.uid = xstrndup(yytext, yyleng - 2);
486 <GlaExt>("-")?{F}"#" {
487 yylval.uid = xstrndup(yytext, yyleng - 1);
491 yylval.uid = xstrndup(yytext, yyleng);
497 * Funky ``foo'' style C literals for Glasgow extensions
501 <GlaExt>"``"[^']+"''" {
502 hsnewid(yytext + 2, yyleng - 4);
508 * Identifiers, both variables and operators. The trailing hash is allowed
509 * for Glasgow extensions.
515 /* These SHOULDNAE work in "Code" (sigh) */
517 <GlaExt,UserPragma>{Id}"#" {
518 if (! nonstandardFlag) {
519 char errbuf[ERR_BUF_SIZE];
520 sprintf(errbuf, "Non-standard identifier (trailing `#'): %s\n", yytext);
523 hsnewid(yytext, yyleng);
524 RETURN(isconstr(yytext) ? CONID : VARID);
526 <Code,GlaExt,UserPragma>{Id} {
527 hsnewid(yytext, yyleng);
528 RETURN(isconstr(yytext) ? CONID : VARID);
530 <Code,GlaExt,UserPragma>{SId} {
531 hsnewid(yytext, yyleng);
532 RETURN(isconstr(yytext) ? CONSYM : VARSYM);
534 <Code,GlaExt,UserPragma>{Mod}"."{Id}"#" {
536 if (! nonstandardFlag) {
537 char errbuf[ERR_BUF_SIZE];
538 sprintf(errbuf, "Non-standard identifier (trailing `#'): %s\n", yytext);
541 is_constr = hsnewqid(yytext, yyleng);
542 RETURN(is_constr ? QCONID : QVARID);
544 <Code,GlaExt,UserPragma>{Mod}"."{Id} {
545 BOOLEAN is_constr = hsnewqid(yytext, yyleng);
546 RETURN(is_constr ? QCONID : QVARID);
548 <Code,GlaExt,UserPragma>{Mod}"."{SId} {
549 BOOLEAN is_constr = hsnewqid(yytext, yyleng);
550 RETURN(is_constr ? QCONSYM : QVARSYM);
554 /* Why is `{Id}#` matched this way, and `{Id}` lexed as three tokens? --JSM */
556 /* Because we can make the former well-behaved (we defined them).
558 Sadly, the latter is defined by Haskell, which allows such
559 la-la land constructs as `{-a 900-line comment-} foo`. (WDP 94/12)
563 <GlaExt,UserPragma>"`"{Id}"#`" {
564 hsnewid(yytext + 1, yyleng - 2);
565 RETURN(isconstr(yytext+1) ? CONSYM : VARSYM);
570 * Character literals. The first form is the quick form, for character
571 * literals that don't contain backslashes. Literals with backslashes are
572 * lexed through multiple rules. First, we match the open ' and as many
573 * normal characters as possible. This puts us into the <Char> state, where
574 * a backslash is legal. Then, we match the backslash and move into the
575 * <CharEsc> state. When we drop out of <CharEsc>, we collect more normal
576 * characters and the close '. We may end up with too many characters, but
577 * this allows us to easily share the lex rules with strings. Excess characters
578 * are ignored with a warning.
582 <GlaExt>'({CHAR}|"\"")"'#" {
583 yylval.uhstring = installHstring(1, yytext+1);
586 <Code,GlaExt>'({CHAR}|"\"")' {
587 yylval.uhstring = installHstring(1, yytext+1);
590 <Code,GlaExt>'' {char errbuf[ERR_BUF_SIZE];
591 sprintf(errbuf, "'' is not a valid character (or string) literal\n");
594 <Code,GlaExt>'({CHAR}|"\"")* {
595 hsmlcolno = hspcolno;
597 addtext(yytext+1, yyleng-1);
600 <Char>({CHAR}|"\"")*'# {
604 addtext(yytext, yyleng - 2);
605 text = fetchtext(&length);
607 if (! nonstandardFlag) {
608 char errbuf[ERR_BUF_SIZE];
609 sprintf(errbuf, "`Char-hash' literals are non-standard: %s\n", text);
614 fprintf(stderr, "%s:%d:%d: Unboxed character literal '",
615 input_filename, hsplineno, hspcolno + 1);
616 format_string(stderr, (unsigned char *) text, length);
617 fputs("' too long\n", stderr);
620 yylval.uhstring = installHstring(1, text);
621 hspcolno = hsmlcolno;
625 <Char>({CHAR}|"\"")*' {
629 addtext(yytext, yyleng - 1);
630 text = fetchtext(&length);
633 fprintf(stderr, "%s:%d:%d: Character literal '",
634 input_filename, hsplineno, hspcolno + 1);
635 format_string(stderr, (unsigned char *) text, length);
636 fputs("' too long\n", stderr);
639 yylval.uhstring = installHstring(1, text);
640 hspcolno = hsmlcolno;
644 <Char>({CHAR}|"\"")+ { addtext(yytext, yyleng); }
649 * String literals. The first form is the quick form, for string literals
650 * that don't contain backslashes. Literals with backslashes are lexed
651 * through multiple rules. First, we match the open " and as many normal
652 * characters as possible. This puts us into the <String> state, where
653 * a backslash is legal. Then, we match the backslash and move into the
654 * <StringEsc> state. When we drop out of <StringEsc>, we collect more normal
655 * characters, moving back and forth between <String> and <StringEsc> as more
656 * backslashes are encountered. (We may even digress into <Comment> mode if we
657 * find a comment in a gap between backslashes.) Finally, we read the last chunk
658 * of normal characters and the close ".
662 <GlaExt>"\""({CHAR}|"'")*"\""# {
663 yylval.uhstring = installHstring(yyleng-3, yytext+1);
664 /* the -3 accounts for the " on front, "# on the end */
667 <Code,GlaExt>"\""({CHAR}|"'")*"\"" {
668 yylval.uhstring = installHstring(yyleng-2, yytext+1);
671 <Code,GlaExt>"\""({CHAR}|"'")* {
672 hsmlcolno = hspcolno;
674 addtext(yytext+1, yyleng-1);
677 <String>({CHAR}|"'")*"\"#" {
681 addtext(yytext, yyleng-2);
682 text = fetchtext(&length);
684 if (! nonstandardFlag) {
685 char errbuf[ERR_BUF_SIZE];
686 sprintf(errbuf, "`String-hash' literals are non-standard: %s\n", text);
690 yylval.uhstring = installHstring(length, text);
691 hspcolno = hsmlcolno;
695 <String>({CHAR}|"'")*"\"" {
699 addtext(yytext, yyleng-1);
700 text = fetchtext(&length);
702 yylval.uhstring = installHstring(length, text);
703 hspcolno = hsmlcolno;
707 <String>({CHAR}|"'")+ { addtext(yytext, yyleng); }
711 * Character and string escapes are roughly the same, but strings have the
712 * extra `\&' sequence which is not allowed for characters. Also, comments
713 * are allowed in the <StringEsc> state. (See the comment section much
716 * NB: Backslashes and tabs are stored in strings as themselves.
717 * But if we print them (in printtree.c), they must go out as
718 * "\\\\" and "\\t" respectively. (This is because of the bogus
719 * intermediate format that the parser produces. It uses '\t' fpr end of
720 * string, so it needs to be able to escape tabs, which means that it
721 * also needs to be able to escape the escape character ('\\'). Sigh.
725 <Char>\\ { PUSH_STATE(CharEsc); }
726 <String>\\& /* Ignore */ ;
727 <String>\\ { PUSH_STATE(StringEsc); noGap = TRUE; }
729 <CharEsc>\\ { addchar(*yytext); POP_STATE; }
730 <StringEsc>\\ { if (noGap) { addchar(*yytext); } POP_STATE; }
732 <CharEsc,StringEsc>["'] { addchar(*yytext); POP_STATE; }
733 <CharEsc,StringEsc>NUL { addchar('\000'); POP_STATE; }
734 <CharEsc,StringEsc>SOH { addchar('\001'); POP_STATE; }
735 <CharEsc,StringEsc>STX { addchar('\002'); POP_STATE; }
736 <CharEsc,StringEsc>ETX { addchar('\003'); POP_STATE; }
737 <CharEsc,StringEsc>EOT { addchar('\004'); POP_STATE; }
738 <CharEsc,StringEsc>ENQ { addchar('\005'); POP_STATE; }
739 <CharEsc,StringEsc>ACK { addchar('\006'); POP_STATE; }
740 <CharEsc,StringEsc>BEL |
741 <CharEsc,StringEsc>a { addchar('\007'); POP_STATE; }
742 <CharEsc,StringEsc>BS |
743 <CharEsc,StringEsc>b { addchar('\010'); POP_STATE; }
744 <CharEsc,StringEsc>HT |
745 <CharEsc,StringEsc>t { addchar('\011'); POP_STATE; }
746 <CharEsc,StringEsc>LF |
747 <CharEsc,StringEsc>n { addchar('\012'); POP_STATE; }
748 <CharEsc,StringEsc>VT |
749 <CharEsc,StringEsc>v { addchar('\013'); POP_STATE; }
750 <CharEsc,StringEsc>FF |
751 <CharEsc,StringEsc>f { addchar('\014'); POP_STATE; }
752 <CharEsc,StringEsc>CR |
753 <CharEsc,StringEsc>r { addchar('\015'); POP_STATE; }
754 <CharEsc,StringEsc>SO { addchar('\016'); POP_STATE; }
755 <CharEsc,StringEsc>SI { addchar('\017'); POP_STATE; }
756 <CharEsc,StringEsc>DLE { addchar('\020'); POP_STATE; }
757 <CharEsc,StringEsc>DC1 { addchar('\021'); POP_STATE; }
758 <CharEsc,StringEsc>DC2 { addchar('\022'); POP_STATE; }
759 <CharEsc,StringEsc>DC3 { addchar('\023'); POP_STATE; }
760 <CharEsc,StringEsc>DC4 { addchar('\024'); POP_STATE; }
761 <CharEsc,StringEsc>NAK { addchar('\025'); POP_STATE; }
762 <CharEsc,StringEsc>SYN { addchar('\026'); POP_STATE; }
763 <CharEsc,StringEsc>ETB { addchar('\027'); POP_STATE; }
764 <CharEsc,StringEsc>CAN { addchar('\030'); POP_STATE; }
765 <CharEsc,StringEsc>EM { addchar('\031'); POP_STATE; }
766 <CharEsc,StringEsc>SUB { addchar('\032'); POP_STATE; }
767 <CharEsc,StringEsc>ESC { addchar('\033'); POP_STATE; }
768 <CharEsc,StringEsc>FS { addchar('\034'); POP_STATE; }
769 <CharEsc,StringEsc>GS { addchar('\035'); POP_STATE; }
770 <CharEsc,StringEsc>RS { addchar('\036'); POP_STATE; }
771 <CharEsc,StringEsc>US { addchar('\037'); POP_STATE; }
772 <CharEsc,StringEsc>SP { addchar('\040'); POP_STATE; }
773 <CharEsc,StringEsc>DEL { addchar('\177'); POP_STATE; }
774 <CharEsc,StringEsc>"^"{CNTRL} { char c = yytext[1] - '@'; addchar(c); POP_STATE; }
775 <CharEsc,StringEsc>{D}+ {
776 int i = strtol(yytext, NULL, 10);
780 char errbuf[ERR_BUF_SIZE];
781 sprintf(errbuf, "Numeric escape \"\\%s\" out of range\n",
787 <CharEsc,StringEsc>o{O}+ {
788 int i = strtol(yytext + 1, NULL, 8);
792 char errbuf[ERR_BUF_SIZE];
793 sprintf(errbuf, "Numeric escape \"\\%s\" out of range\n",
799 <CharEsc,StringEsc>x{H}+ {
800 int i = strtol(yytext + 1, NULL, 16);
804 char errbuf[ERR_BUF_SIZE];
805 sprintf(errbuf, "Numeric escape \"\\%s\" out of range\n",
815 * Nested comments. The major complication here is in trying to match the
816 * longest lexemes possible, for better performance. (See the flex document.)
817 * That's why the rules look so bizarre.
821 <Code,GlaExt,UserPragma,StringEsc>"{-" {
822 noGap = FALSE; nested_comments = 1; PUSH_STATE(Comment);
826 <Comment>"-"+[^-{}]+ |
827 <Comment>"{"+[^-{}]+ ;
828 <Comment>"{-" { nested_comments++; }
829 <Comment>"-}" { if (--nested_comments == 0) POP_STATE; }
834 * Illegal characters. This used to be a single rule, but we might as well
835 * pass on as much information as we have, so now we indicate our state in
840 <INITIAL,Code,GlaExt,UserPragma>(.|\n) {
841 fprintf(stderr, "%s:%d:%d: Illegal character: `",
842 input_filename, hsplineno, hspcolno + 1);
843 format_string(stderr, (unsigned char *) yytext, 1);
844 fputs("'\n", stderr);
848 fprintf(stderr, "%s:%d:%d: Illegal character: `",
849 input_filename, hsplineno, hspcolno + 1);
850 format_string(stderr, (unsigned char *) yytext, 1);
851 fputs("' in a character literal\n", stderr);
855 fprintf(stderr, "%s:%d:%d: Illegal character escape: `\\",
856 input_filename, hsplineno, hspcolno + 1);
857 format_string(stderr, (unsigned char *) yytext, 1);
858 fputs("'\n", stderr);
861 <String>(.|\n) { if (nonstandardFlag) {
862 addtext(yytext, yyleng);
864 fprintf(stderr, "%s:%d:%d: Illegal character: `",
865 input_filename, hsplineno, hspcolno + 1);
866 format_string(stderr, (unsigned char *) yytext, 1);
867 fputs("' in a string literal\n", stderr);
873 fprintf(stderr, "%s:%d:%d: Illegal string escape: `\\",
874 input_filename, hsplineno, hspcolno + 1);
875 format_string(stderr, (unsigned char *) yytext, 1);
876 fputs("'\n", stderr);
879 fprintf(stderr, "%s:%d:%d: Illegal character: `",
880 input_filename, hsplineno, hspcolno + 1);
881 format_string(stderr, (unsigned char *) yytext, 1);
882 fputs("' in a string gap\n", stderr);
889 * End of file. In any sub-state, this is an error. However, for the primary
890 * <Code> and <GlaExt> states, this is perfectly normal. We just return an EOF
891 * and let the yylex() wrapper deal with whatever has to be done next (e.g.
892 * adding virtual close curlies, or closing an interface and returning to the
893 * primary source file.
895 * Note that flex does not call YY_USER_ACTION for <<EOF>> rules. Hence the
896 * line/column advancement has to be done by hand.
900 <Char,CharEsc><<EOF>> {
901 hsplineno = hslineno; hspcolno = hscolno;
902 hsperror("unterminated character literal");
905 hsplineno = hslineno; hspcolno = hscolno;
906 hsperror("unterminated comment");
908 <String,StringEsc><<EOF>> {
909 hsplineno = hslineno; hspcolno = hscolno;
910 hsperror("unterminated string literal");
912 <UserPragma><<EOF>> {
913 hsplineno = hslineno; hspcolno = hscolno;
914 hsperror("unterminated user-specified pragma");
916 <Code,GlaExt><<EOF>> { hsplineno = hslineno; hspcolno = hscolno; return(EOF); }
920 /**********************************************************************
923 * YACC/LEX Initialisation etc. *
926 **********************************************************************/
929 We initialise input_filename to "<stdin>".
930 This allows unnamed sources to be piped into the parser.
936 input_filename = xstrdup("<stdin>");
938 /* We must initialize the input buffer _now_, because we call
939 setyyin _before_ calling yylex for the first time! */
940 yy_switch_to_buffer(yy_create_buffer(stdin, YY_BUF_SIZE));
949 new_filename(char *f) /* This looks pretty dodgy to me (WDP) */
951 if (input_filename != NULL)
952 free(input_filename);
953 input_filename = xstrdup(f);
956 /**********************************************************************
959 * Layout Processing *
962 **********************************************************************/
965 The following section deals with Haskell Layout conventions
966 forcing insertion of ; or } as appropriate
972 return (!forgetindent && INDENTON);
976 /* Enter new context and set new indentation level */
981 fprintf(stderr, "hssetindent:hscolno=%d,hspcolno=%d,INDENTPT[%d]=%d\n", hscolno, hspcolno, icontexts, INDENTPT);
985 * partain: first chk that new indent won't be less than current one; this code
986 * doesn't make sense to me; hscolno tells the position of the _end_ of the
987 * current token; what that has to do with indenting, I don't know.
991 if (hscolno - 1 <= INDENTPT) {
993 return; /* Empty input OK for Haskell 1.1 */
995 char errbuf[ERR_BUF_SIZE];
997 sprintf(errbuf, "Layout error -- indentation should be > %d cols", INDENTPT);
1001 hsentercontext((hspcolno << 1) | 1);
1005 /* Enter a new context without changing the indentation level */
1010 fprintf(stderr, "hsincindent:hscolno=%d,hspcolno=%d,INDENTPT[%d]=%d\n", hscolno, hspcolno, icontexts, INDENTPT);
1012 hsentercontext(indenttab[icontexts] & ~1);
1016 /* Turn off indentation processing, usually because an explicit "{" has been seen */
1020 forgetindent = TRUE;
1024 /* Enter a new layout context. */
1026 hsentercontext(int indent)
1028 /* Enter new context and set indentation as specified */
1029 if (++icontexts >= MAX_CONTEXTS) {
1030 char errbuf[ERR_BUF_SIZE];
1032 sprintf(errbuf, "`wheres' and `cases' nested too deeply (>%d)", MAX_CONTEXTS - 1);
1035 forgetindent = FALSE;
1036 indenttab[icontexts] = indent;
1038 fprintf(stderr, "hsentercontext:indent=%d,hscolno=%d,hspcolno=%d,INDENTPT[%d]=%d\n", indent, hscolno, hspcolno, icontexts, INDENTPT);
1043 /* Exit a layout context */
1049 fprintf(stderr, "hsendindent:hscolno=%d,hspcolno=%d,INDENTPT[%d]=%d\n", hscolno, hspcolno, icontexts, INDENTPT);
1054 * Return checks the indentation level and returns ;, } or the specified token.
1064 if (hsshouldindent()) {
1065 if (hspcolno < INDENTPT) {
1067 fprintf(stderr, "inserted '}' before %d (%d:%d:%d:%d)\n", tok, hspcolno, hscolno, yyleng, INDENTPT);
1071 } else if (hspcolno == INDENTPT) {
1073 fprintf(stderr, "inserted ';' before %d (%d:%d)\n", tok, hspcolno, INDENTPT);
1081 fprintf(stderr, "returning %d (%d:%d)\n", tok, hspcolno, INDENTPT);
1088 * Redefine yylex to check for stacked tokens, yylex1() is the original yylex()
1094 static BOOLEAN eof = FALSE;
1097 if (hssttok != -1) {
1105 endlineno = hslineno;
1106 if ((tok = yylex1()) != EOF)
1112 if (icontexts > icontexts_save) {
1115 indenttab[icontexts] = 0;
1118 hsperror("missing '}' at end of file");
1119 } else if (hsbuf_save != NULL) {
1121 yy_delete_buffer(YY_CURRENT_BUFFER);
1122 yy_switch_to_buffer(hsbuf_save);
1124 new_filename(filename_save);
1125 free(filename_save);
1126 hslineno = hslineno_save;
1127 hsplineno = hsplineno_save;
1128 hscolno = hscolno_save;
1129 hspcolno = hspcolno_save;
1131 icontexts = icontexts_save - 1;
1134 fprintf(stderr, "finished reading interface (%d:%d:%d)\n", hscolno, hspcolno, INDENTPT);
1139 hsperror("No longer using yacc to parse interface files");
1144 abort(); /* should never get here! */
1148 /**********************************************************************
1151 * Input Processing for Interfaces -- Not currently used !!! *
1154 **********************************************************************/
1156 /* setyyin(file) open file as new lex input buffer */
1162 hsbuf_save = YY_CURRENT_BUFFER;
1163 if ((yyin = fopen(file, "r")) == NULL) {
1164 char errbuf[ERR_BUF_SIZE];
1166 sprintf(errbuf, "can't read \"%-.50s\"", file);
1169 yy_switch_to_buffer(yy_create_buffer(yyin, YY_BUF_SIZE));
1171 hslineno_save = hslineno;
1172 hsplineno_save = hsplineno;
1173 hslineno = hsplineno = 1;
1175 filename_save = input_filename;
1176 input_filename = NULL;
1178 hscolno_save = hscolno;
1179 hspcolno_save = hspcolno;
1180 hscolno = hspcolno = 0;
1181 etags_save = etags; /* do not do "etags" stuff in interfaces */
1182 etags = 0; /* We remember whether we are doing it in
1183 the module, so we can restore it later [WDP 94/09] */
1184 hsentercontext(-1); /* partain: changed this from 0 */
1185 icontexts_save = icontexts;
1187 fprintf(stderr, "reading %s (%d:%d:%d)\n", input_filename, hscolno_save, hspcolno_save, INDENTPT);
1192 layout_input(char *text, int len)
1195 fprintf(stderr, "Scanning \"%s\"\n", text);
1198 hsplineno = hslineno;
1210 hscolno += 8 - (hscolno % 8); /* Tabs stops are 8 columns apart */
1222 setstartlineno(void)
1224 startlineno = hsplineno;
1226 if (modulelineno == 0) {
1227 modulelineno = startlineno;
1233 fprintf(stderr,"%u\tsetstartlineno (col %u)\n",startlineno,hscolno);
1237 /**********************************************************************
1243 **********************************************************************/
1245 #define CACHE_SIZE YY_BUF_SIZE
1251 } textcache = { 0, 0, NULL };
1256 /* fprintf(stderr, "cleartext\n"); */
1258 if (textcache.allocated == 0) {
1259 textcache.allocated = CACHE_SIZE;
1260 textcache.text = xmalloc(CACHE_SIZE);
1265 addtext(char *text, unsigned length)
1267 /* fprintf(stderr, "addtext: %d %s\n", length, text); */
1272 if (textcache.next + length + 1 >= textcache.allocated) {
1273 textcache.allocated += length + CACHE_SIZE;
1274 textcache.text = xrealloc(textcache.text, textcache.allocated);
1276 bcopy(text, textcache.text + textcache.next, length);
1277 textcache.next += length;
1283 /* fprintf(stderr, "addchar: %c\n", c); */
1285 if (textcache.next + 2 >= textcache.allocated) {
1286 textcache.allocated += CACHE_SIZE;
1287 textcache.text = xrealloc(textcache.text, textcache.allocated);
1289 textcache.text[textcache.next++] = c;
1293 fetchtext(unsigned *length)
1295 /* fprintf(stderr, "fetchtext: %d\n", textcache.next); */
1297 *length = textcache.next;
1298 textcache.text[textcache.next] = '\0';
1299 return textcache.text;
1302 /**********************************************************************
1305 * Identifier Processing *
1308 **********************************************************************/
1311 hsnewid Enters an id of length n into the symbol table.
1315 hsnewid(char *name, int length)
1317 char save = name[length];
1319 name[length] = '\0';
1320 yylval.uid = installid(name);
1321 name[length] = save;
1325 hsnewqid(char *name, int length)
1328 char save = name[length];
1329 name[length] = '\0';
1331 dot = strchr(name, '.');
1333 yylval.uqid = mkaqual(installid(name),installid(dot+1));
1335 name[length] = save;
1337 return isconstr(dot+1);