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}*"MAGIC_UNFOLDING" {
329 PUSH_STATE(UserPragma);
330 RETURN(MAGIC_UNFOLDING_UPRAGMA);
332 <Code,GlaExt>"{-#"{WS}*"GENERATE_SPECS" {
333 /* these are handled by hscpp */
337 <Code,GlaExt>"{-#"{WS}*"OPTIONS" {
338 /* these are for the driver! */
342 <Code,GlaExt>"{-#"{WS}*"SOURCE"{WS}*"#"?"-}" {
343 /* these are used by `make depend' and the
344 compiler to indicate that a module should
345 be imported from source */
347 RETURN(SOURCE_UPRAGMA);
350 <Code,GlaExt>"{-#"{WS}*[A-Z_]+ {
351 fprintf(stderr, "%s:%d: Warning: Unrecognised pragma '",
352 input_filename, hsplineno);
353 format_string(stderr, (unsigned char *) yytext, yyleng);
354 fputs("'\n", stderr);
358 <UserPragma>"#-}" { POP_STATE; RETURN(END_UPRAGMA); }
362 * Haskell keywords. `scc' is actually a Glasgow extension, but it is
363 * intentionally accepted as a keyword even for normal <Code>.
367 <Code,GlaExt>"case" { RETURN(CASE); }
368 <Code,GlaExt>"class" { RETURN(CLASS); }
369 <Code,GlaExt,UserPragma>"data" { RETURN(DATA); }
370 <Code,GlaExt>"default" { RETURN(DEFAULT); }
371 <Code,GlaExt>"deriving" { RETURN(DERIVING); }
372 <Code,GlaExt>"do" { RETURN(DO); }
373 <Code,GlaExt>"else" { RETURN(ELSE); }
374 <Code,GlaExt>"if" { RETURN(IF); }
375 <Code,GlaExt>"import" { RETURN(IMPORT); }
376 <Code,GlaExt>"in" { RETURN(IN); }
377 <Code,GlaExt>"infix" { RETURN(INFIX); }
378 <Code,GlaExt>"infixl" { RETURN(INFIXL); }
379 <Code,GlaExt>"infixr" { RETURN(INFIXR); }
380 <Code,GlaExt,UserPragma>"instance" { RETURN(INSTANCE); }
381 <Code,GlaExt>"let" { RETURN(LET); }
382 <Code,GlaExt>"module" { RETURN(MODULE); }
383 <Code,GlaExt>"newtype" { RETURN(NEWTYPE); }
384 <Code,GlaExt>"of" { RETURN(OF); }
385 <Code,GlaExt>"then" { RETURN(THEN); }
386 <Code,GlaExt>"type" { RETURN(TYPE); }
387 <Code,GlaExt>"where" { RETURN(WHERE); }
389 <Code,GlaExt>"as" { RETURN(AS); }
390 <Code,GlaExt>"hiding" { RETURN(HIDING); }
391 <Code,GlaExt>"qualified" { RETURN(QUALIFIED); }
393 <Code,GlaExt>"_scc_" { RETURN(SCC); }
394 <GlaExt>"_ccall_" { RETURN(CCALL); }
395 <GlaExt>"_ccall_GC_" { RETURN(CCALL_GC); }
396 <GlaExt>"_casm_" { RETURN(CASM); }
397 <GlaExt>"_casm_GC_" { RETURN(CASM_GC); }
401 * Haskell operators: special, reservedops and useful varsyms
405 <Code,GlaExt,UserPragma>"(" { RETURN(OPAREN); }
406 <Code,GlaExt,UserPragma>")" { RETURN(CPAREN); }
407 <Code,GlaExt,UserPragma>"[" { RETURN(OBRACK); }
408 <Code,GlaExt,UserPragma>"]" { RETURN(CBRACK); }
409 <Code,GlaExt>"{" { RETURN(OCURLY); }
410 <Code,GlaExt>"}" { RETURN(CCURLY); }
411 <Code,GlaExt,UserPragma>"," { RETURN(COMMA); }
412 <Code,GlaExt>";" { RETURN(SEMI); }
413 <Code,GlaExt>"`" { RETURN(BQUOTE); }
414 <Code,GlaExt>"_" { RETURN(WILDCARD); }
416 <Code,GlaExt>".." { RETURN(DOTDOT); }
417 <Code,GlaExt,UserPragma>"::" { RETURN(DCOLON); }
418 <Code,GlaExt,UserPragma>"=" { RETURN(EQUAL); }
419 <Code,GlaExt>"\\" { RETURN(LAMBDA); }
420 <Code,GlaExt>"|" { RETURN(VBAR); }
421 <Code,GlaExt>"<-" { RETURN(LARROW); }
422 <Code,GlaExt,UserPragma>"->" { RETURN(RARROW); }
423 <Code,GlaExt>"-" { RETURN(MINUS); }
424 <Code,GlaExt>"+" { RETURN(PLUS); }
426 <Code,GlaExt,UserPragma>"=>" { RETURN(DARROW); }
427 <Code,GlaExt>"@" { RETURN(AT); }
428 <Code,GlaExt>"!" { RETURN(BANG); }
429 <Code,GlaExt>"~" { RETURN(LAZY); }
433 * Integers and (for Glasgow extensions) primitive integers. Note that
434 * we pass all of the text on to the parser, because flex/C can't handle
435 * arbitrary precision numbers.
439 <GlaExt>("-")?"0"[Oo]{O}+"#" { /* octal */
440 yylval.uid = xstrndup(yytext, yyleng - 1);
443 <Code,GlaExt>"0"[Oo]{O}+ { /* octal */
444 yylval.uid = xstrndup(yytext, yyleng);
447 <GlaExt>("-")?"0"[Xx]{H}+"#" { /* hexadecimal */
448 yylval.uid = xstrndup(yytext, yyleng - 1);
451 <Code,GlaExt>"0"[Xx]{H}+ { /* hexadecimal */
452 yylval.uid = xstrndup(yytext, yyleng);
455 <GlaExt>("-")?{N}"#" {
456 yylval.uid = xstrndup(yytext, yyleng - 1);
459 <Code,GlaExt,UserPragma>{N} {
460 yylval.uid = xstrndup(yytext, yyleng);
466 * Floats and (for Glasgow extensions) primitive floats/doubles.
470 <GlaExt>("-")?{F}"##" {
471 yylval.uid = xstrndup(yytext, yyleng - 2);
474 <GlaExt>("-")?{F}"#" {
475 yylval.uid = xstrndup(yytext, yyleng - 1);
479 yylval.uid = xstrndup(yytext, yyleng);
485 * Funky ``foo'' style C literals for Glasgow extensions
489 <GlaExt>"``"[^']+"''" {
490 hsnewid(yytext + 2, yyleng - 4);
496 * Identifiers, both variables and operators. The trailing hash is allowed
497 * for Glasgow extensions.
503 /* These SHOULDNAE work in "Code" (sigh) */
505 <GlaExt,UserPragma>{Id}"#" {
506 if (! nonstandardFlag) {
507 char errbuf[ERR_BUF_SIZE];
508 sprintf(errbuf, "Non-standard identifier (trailing `#'): %s\n", yytext);
511 hsnewid(yytext, yyleng);
512 RETURN(isconstr(yytext) ? CONID : VARID);
514 <Code,GlaExt,UserPragma>{Id} {
515 hsnewid(yytext, yyleng);
516 RETURN(isconstr(yytext) ? CONID : VARID);
518 <Code,GlaExt,UserPragma>{SId} {
519 hsnewid(yytext, yyleng);
520 RETURN(isconstr(yytext) ? CONSYM : VARSYM);
522 <Code,GlaExt,UserPragma>{Mod}"."{Id}"#" {
524 if (! nonstandardFlag) {
525 char errbuf[ERR_BUF_SIZE];
526 sprintf(errbuf, "Non-standard identifier (trailing `#'): %s\n", yytext);
529 is_constr = hsnewqid(yytext, yyleng);
530 RETURN(is_constr ? QCONID : QVARID);
532 <Code,GlaExt,UserPragma>{Mod}"."{Id} {
533 BOOLEAN is_constr = hsnewqid(yytext, yyleng);
534 RETURN(is_constr ? QCONID : QVARID);
536 <Code,GlaExt,UserPragma>{Mod}"."{SId} {
537 BOOLEAN is_constr = hsnewqid(yytext, yyleng);
538 RETURN(is_constr ? QCONSYM : QVARSYM);
542 /* Why is `{Id}#` matched this way, and `{Id}` lexed as three tokens? --JSM */
544 /* Because we can make the former well-behaved (we defined them).
546 Sadly, the latter is defined by Haskell, which allows such
547 la-la land constructs as `{-a 900-line comment-} foo`. (WDP 94/12)
551 <GlaExt,UserPragma>"`"{Id}"#`" {
552 hsnewid(yytext + 1, yyleng - 2);
553 RETURN(isconstr(yytext+1) ? CONSYM : VARSYM);
558 * Character literals. The first form is the quick form, for character
559 * literals that don't contain backslashes. Literals with backslashes are
560 * lexed through multiple rules. First, we match the open ' and as many
561 * normal characters as possible. This puts us into the <Char> state, where
562 * a backslash is legal. Then, we match the backslash and move into the
563 * <CharEsc> state. When we drop out of <CharEsc>, we collect more normal
564 * characters and the close '. We may end up with too many characters, but
565 * this allows us to easily share the lex rules with strings. Excess characters
566 * are ignored with a warning.
570 <GlaExt>'({CHAR}|"\"")"'#" {
571 yylval.uhstring = installHstring(1, yytext+1);
574 <Code,GlaExt>'({CHAR}|"\"")' {
575 yylval.uhstring = installHstring(1, yytext+1);
578 <Code,GlaExt>'' {char errbuf[ERR_BUF_SIZE];
579 sprintf(errbuf, "'' is not a valid character (or string) literal\n");
582 <Code,GlaExt>'({CHAR}|"\"")* {
583 hsmlcolno = hspcolno;
585 addtext(yytext+1, yyleng-1);
588 <Char>({CHAR}|"\"")*'# {
592 addtext(yytext, yyleng - 2);
593 text = fetchtext(&length);
595 if (! nonstandardFlag) {
596 char errbuf[ERR_BUF_SIZE];
597 sprintf(errbuf, "`Char-hash' literals are non-standard: %s\n", text);
602 fprintf(stderr, "%s:%d:%d: Unboxed character literal '",
603 input_filename, hsplineno, hspcolno + 1);
604 format_string(stderr, (unsigned char *) text, length);
605 fputs("' too long\n", stderr);
608 yylval.uhstring = installHstring(1, text);
609 hspcolno = hsmlcolno;
613 <Char>({CHAR}|"\"")*' {
617 addtext(yytext, yyleng - 1);
618 text = fetchtext(&length);
621 fprintf(stderr, "%s:%d:%d: Character literal '",
622 input_filename, hsplineno, hspcolno + 1);
623 format_string(stderr, (unsigned char *) text, length);
624 fputs("' too long\n", stderr);
627 yylval.uhstring = installHstring(1, text);
628 hspcolno = hsmlcolno;
632 <Char>({CHAR}|"\"")+ { addtext(yytext, yyleng); }
637 * String literals. The first form is the quick form, for string literals
638 * that don't contain backslashes. Literals with backslashes are lexed
639 * through multiple rules. First, we match the open " and as many normal
640 * characters as possible. This puts us into the <String> state, where
641 * a backslash is legal. Then, we match the backslash and move into the
642 * <StringEsc> state. When we drop out of <StringEsc>, we collect more normal
643 * characters, moving back and forth between <String> and <StringEsc> as more
644 * backslashes are encountered. (We may even digress into <Comment> mode if we
645 * find a comment in a gap between backslashes.) Finally, we read the last chunk
646 * of normal characters and the close ".
650 <GlaExt>"\""({CHAR}|"'")*"\""# {
651 yylval.uhstring = installHstring(yyleng-3, yytext+1);
652 /* the -3 accounts for the " on front, "# on the end */
655 <Code,GlaExt>"\""({CHAR}|"'")*"\"" {
656 yylval.uhstring = installHstring(yyleng-2, yytext+1);
659 <Code,GlaExt>"\""({CHAR}|"'")* {
660 hsmlcolno = hspcolno;
662 addtext(yytext+1, yyleng-1);
665 <String>({CHAR}|"'")*"\"#" {
669 addtext(yytext, yyleng-2);
670 text = fetchtext(&length);
672 if (! nonstandardFlag) {
673 char errbuf[ERR_BUF_SIZE];
674 sprintf(errbuf, "`String-hash' literals are non-standard: %s\n", text);
678 yylval.uhstring = installHstring(length, text);
679 hspcolno = hsmlcolno;
683 <String>({CHAR}|"'")*"\"" {
687 addtext(yytext, yyleng-1);
688 text = fetchtext(&length);
690 yylval.uhstring = installHstring(length, text);
691 hspcolno = hsmlcolno;
695 <String>({CHAR}|"'")+ { addtext(yytext, yyleng); }
699 * Character and string escapes are roughly the same, but strings have the
700 * extra `\&' sequence which is not allowed for characters. Also, comments
701 * are allowed in the <StringEsc> state. (See the comment section much
704 * NB: Backslashes and tabs are stored in strings as themselves.
705 * But if we print them (in printtree.c), they must go out as
706 * "\\\\" and "\\t" respectively. (This is because of the bogus
707 * intermediate format that the parser produces. It uses '\t' fpr end of
708 * string, so it needs to be able to escape tabs, which means that it
709 * also needs to be able to escape the escape character ('\\'). Sigh.
713 <Char>\\ { PUSH_STATE(CharEsc); }
714 <String>\\& /* Ignore */ ;
715 <String>\\ { PUSH_STATE(StringEsc); noGap = TRUE; }
717 <CharEsc>\\ { addchar(*yytext); POP_STATE; }
718 <StringEsc>\\ { if (noGap) { addchar(*yytext); } POP_STATE; }
720 <CharEsc,StringEsc>["'] { addchar(*yytext); POP_STATE; }
721 <CharEsc,StringEsc>NUL { addchar('\000'); POP_STATE; }
722 <CharEsc,StringEsc>SOH { addchar('\001'); POP_STATE; }
723 <CharEsc,StringEsc>STX { addchar('\002'); POP_STATE; }
724 <CharEsc,StringEsc>ETX { addchar('\003'); POP_STATE; }
725 <CharEsc,StringEsc>EOT { addchar('\004'); POP_STATE; }
726 <CharEsc,StringEsc>ENQ { addchar('\005'); POP_STATE; }
727 <CharEsc,StringEsc>ACK { addchar('\006'); POP_STATE; }
728 <CharEsc,StringEsc>BEL |
729 <CharEsc,StringEsc>a { addchar('\007'); POP_STATE; }
730 <CharEsc,StringEsc>BS |
731 <CharEsc,StringEsc>b { addchar('\010'); POP_STATE; }
732 <CharEsc,StringEsc>HT |
733 <CharEsc,StringEsc>t { addchar('\011'); POP_STATE; }
734 <CharEsc,StringEsc>LF |
735 <CharEsc,StringEsc>n { addchar('\012'); POP_STATE; }
736 <CharEsc,StringEsc>VT |
737 <CharEsc,StringEsc>v { addchar('\013'); POP_STATE; }
738 <CharEsc,StringEsc>FF |
739 <CharEsc,StringEsc>f { addchar('\014'); POP_STATE; }
740 <CharEsc,StringEsc>CR |
741 <CharEsc,StringEsc>r { addchar('\015'); POP_STATE; }
742 <CharEsc,StringEsc>SO { addchar('\016'); POP_STATE; }
743 <CharEsc,StringEsc>SI { addchar('\017'); POP_STATE; }
744 <CharEsc,StringEsc>DLE { addchar('\020'); POP_STATE; }
745 <CharEsc,StringEsc>DC1 { addchar('\021'); POP_STATE; }
746 <CharEsc,StringEsc>DC2 { addchar('\022'); POP_STATE; }
747 <CharEsc,StringEsc>DC3 { addchar('\023'); POP_STATE; }
748 <CharEsc,StringEsc>DC4 { addchar('\024'); POP_STATE; }
749 <CharEsc,StringEsc>NAK { addchar('\025'); POP_STATE; }
750 <CharEsc,StringEsc>SYN { addchar('\026'); POP_STATE; }
751 <CharEsc,StringEsc>ETB { addchar('\027'); POP_STATE; }
752 <CharEsc,StringEsc>CAN { addchar('\030'); POP_STATE; }
753 <CharEsc,StringEsc>EM { addchar('\031'); POP_STATE; }
754 <CharEsc,StringEsc>SUB { addchar('\032'); POP_STATE; }
755 <CharEsc,StringEsc>ESC { addchar('\033'); POP_STATE; }
756 <CharEsc,StringEsc>FS { addchar('\034'); POP_STATE; }
757 <CharEsc,StringEsc>GS { addchar('\035'); POP_STATE; }
758 <CharEsc,StringEsc>RS { addchar('\036'); POP_STATE; }
759 <CharEsc,StringEsc>US { addchar('\037'); POP_STATE; }
760 <CharEsc,StringEsc>SP { addchar('\040'); POP_STATE; }
761 <CharEsc,StringEsc>DEL { addchar('\177'); POP_STATE; }
762 <CharEsc,StringEsc>"^"{CNTRL} { char c = yytext[1] - '@'; addchar(c); POP_STATE; }
763 <CharEsc,StringEsc>{D}+ {
764 int i = strtol(yytext, NULL, 10);
768 char errbuf[ERR_BUF_SIZE];
769 sprintf(errbuf, "Numeric escape \"\\%s\" out of range\n",
775 <CharEsc,StringEsc>o{O}+ {
776 int i = strtol(yytext + 1, NULL, 8);
780 char errbuf[ERR_BUF_SIZE];
781 sprintf(errbuf, "Numeric escape \"\\%s\" out of range\n",
787 <CharEsc,StringEsc>x{H}+ {
788 int i = strtol(yytext + 1, NULL, 16);
792 char errbuf[ERR_BUF_SIZE];
793 sprintf(errbuf, "Numeric escape \"\\%s\" out of range\n",
803 * Nested comments. The major complication here is in trying to match the
804 * longest lexemes possible, for better performance. (See the flex document.)
805 * That's why the rules look so bizarre.
809 <Code,GlaExt,UserPragma,StringEsc>"{-" {
810 noGap = FALSE; nested_comments = 1; PUSH_STATE(Comment);
814 <Comment>"-"+[^-{}]+ |
815 <Comment>"{"+[^-{}]+ ;
816 <Comment>"{-" { nested_comments++; }
817 <Comment>"-}" { if (--nested_comments == 0) POP_STATE; }
822 * Illegal characters. This used to be a single rule, but we might as well
823 * pass on as much information as we have, so now we indicate our state in
828 <INITIAL,Code,GlaExt,UserPragma>(.|\n) {
829 fprintf(stderr, "%s:%d:%d: Illegal character: `",
830 input_filename, hsplineno, hspcolno + 1);
831 format_string(stderr, (unsigned char *) yytext, 1);
832 fputs("'\n", stderr);
836 fprintf(stderr, "%s:%d:%d: Illegal character: `",
837 input_filename, hsplineno, hspcolno + 1);
838 format_string(stderr, (unsigned char *) yytext, 1);
839 fputs("' in a character literal\n", stderr);
843 fprintf(stderr, "%s:%d:%d: Illegal character escape: `\\",
844 input_filename, hsplineno, hspcolno + 1);
845 format_string(stderr, (unsigned char *) yytext, 1);
846 fputs("'\n", stderr);
849 <String>(.|\n) { if (nonstandardFlag) {
850 addtext(yytext, yyleng);
852 fprintf(stderr, "%s:%d:%d: Illegal character: `",
853 input_filename, hsplineno, hspcolno + 1);
854 format_string(stderr, (unsigned char *) yytext, 1);
855 fputs("' in a string literal\n", stderr);
861 fprintf(stderr, "%s:%d:%d: Illegal string escape: `\\",
862 input_filename, hsplineno, hspcolno + 1);
863 format_string(stderr, (unsigned char *) yytext, 1);
864 fputs("'\n", stderr);
867 fprintf(stderr, "%s:%d:%d: Illegal character: `",
868 input_filename, hsplineno, hspcolno + 1);
869 format_string(stderr, (unsigned char *) yytext, 1);
870 fputs("' in a string gap\n", stderr);
877 * End of file. In any sub-state, this is an error. However, for the primary
878 * <Code> and <GlaExt> states, this is perfectly normal. We just return an EOF
879 * and let the yylex() wrapper deal with whatever has to be done next (e.g.
880 * adding virtual close curlies, or closing an interface and returning to the
881 * primary source file.
883 * Note that flex does not call YY_USER_ACTION for <<EOF>> rules. Hence the
884 * line/column advancement has to be done by hand.
888 <Char,CharEsc><<EOF>> {
889 hsplineno = hslineno; hspcolno = hscolno;
890 hsperror("unterminated character literal");
893 hsplineno = hslineno; hspcolno = hscolno;
894 hsperror("unterminated comment");
896 <String,StringEsc><<EOF>> {
897 hsplineno = hslineno; hspcolno = hscolno;
898 hsperror("unterminated string literal");
900 <UserPragma><<EOF>> {
901 hsplineno = hslineno; hspcolno = hscolno;
902 hsperror("unterminated user-specified pragma");
904 <Code,GlaExt><<EOF>> { hsplineno = hslineno; hspcolno = hscolno; return(EOF); }
908 /**********************************************************************
911 * YACC/LEX Initialisation etc. *
914 **********************************************************************/
917 We initialise input_filename to "<stdin>".
918 This allows unnamed sources to be piped into the parser.
924 input_filename = xstrdup("<stdin>");
926 /* We must initialize the input buffer _now_, because we call
927 setyyin _before_ calling yylex for the first time! */
928 yy_switch_to_buffer(yy_create_buffer(stdin, YY_BUF_SIZE));
937 new_filename(char *f) /* This looks pretty dodgy to me (WDP) */
939 if (input_filename != NULL)
940 free(input_filename);
941 input_filename = xstrdup(f);
944 /**********************************************************************
947 * Layout Processing *
950 **********************************************************************/
953 The following section deals with Haskell Layout conventions
954 forcing insertion of ; or } as appropriate
960 return (!forgetindent && INDENTON);
964 /* Enter new context and set new indentation level */
969 fprintf(stderr, "hssetindent:hscolno=%d,hspcolno=%d,INDENTPT[%d]=%d\n", hscolno, hspcolno, icontexts, INDENTPT);
973 * partain: first chk that new indent won't be less than current one; this code
974 * doesn't make sense to me; hscolno tells the position of the _end_ of the
975 * current token; what that has to do with indenting, I don't know.
979 if (hscolno - 1 <= INDENTPT) {
981 return; /* Empty input OK for Haskell 1.1 */
983 char errbuf[ERR_BUF_SIZE];
985 sprintf(errbuf, "Layout error -- indentation should be > %d cols", INDENTPT);
989 hsentercontext((hspcolno << 1) | 1);
993 /* Enter a new context without changing the indentation level */
998 fprintf(stderr, "hsincindent:hscolno=%d,hspcolno=%d,INDENTPT[%d]=%d\n", hscolno, hspcolno, icontexts, INDENTPT);
1000 hsentercontext(indenttab[icontexts] & ~1);
1004 /* Turn off indentation processing, usually because an explicit "{" has been seen */
1008 forgetindent = TRUE;
1012 /* Enter a new layout context. */
1014 hsentercontext(int indent)
1016 /* Enter new context and set indentation as specified */
1017 if (++icontexts >= MAX_CONTEXTS) {
1018 char errbuf[ERR_BUF_SIZE];
1020 sprintf(errbuf, "`wheres' and `cases' nested too deeply (>%d)", MAX_CONTEXTS - 1);
1023 forgetindent = FALSE;
1024 indenttab[icontexts] = indent;
1026 fprintf(stderr, "hsentercontext:indent=%d,hscolno=%d,hspcolno=%d,INDENTPT[%d]=%d\n", indent, hscolno, hspcolno, icontexts, INDENTPT);
1031 /* Exit a layout context */
1037 fprintf(stderr, "hsendindent:hscolno=%d,hspcolno=%d,INDENTPT[%d]=%d\n", hscolno, hspcolno, icontexts, INDENTPT);
1042 * Return checks the indentation level and returns ;, } or the specified token.
1052 if (hsshouldindent()) {
1053 if (hspcolno < INDENTPT) {
1055 fprintf(stderr, "inserted '}' before %d (%d:%d:%d:%d)\n", tok, hspcolno, hscolno, yyleng, INDENTPT);
1059 } else if (hspcolno == INDENTPT) {
1061 fprintf(stderr, "inserted ';' before %d (%d:%d)\n", tok, hspcolno, INDENTPT);
1069 fprintf(stderr, "returning %d (%d:%d)\n", tok, hspcolno, INDENTPT);
1076 * Redefine yylex to check for stacked tokens, yylex1() is the original yylex()
1082 static BOOLEAN eof = FALSE;
1085 if (hssttok != -1) {
1093 endlineno = hslineno;
1094 if ((tok = yylex1()) != EOF)
1100 if (icontexts > icontexts_save) {
1103 indenttab[icontexts] = 0;
1106 hsperror("missing '}' at end of file");
1107 } else if (hsbuf_save != NULL) {
1109 yy_delete_buffer(YY_CURRENT_BUFFER);
1110 yy_switch_to_buffer(hsbuf_save);
1112 new_filename(filename_save);
1113 free(filename_save);
1114 hslineno = hslineno_save;
1115 hsplineno = hsplineno_save;
1116 hscolno = hscolno_save;
1117 hspcolno = hspcolno_save;
1119 icontexts = icontexts_save - 1;
1122 fprintf(stderr, "finished reading interface (%d:%d:%d)\n", hscolno, hspcolno, INDENTPT);
1127 hsperror("No longer using yacc to parse interface files");
1132 abort(); /* should never get here! */
1136 /**********************************************************************
1139 * Input Processing for Interfaces -- Not currently used !!! *
1142 **********************************************************************/
1144 /* setyyin(file) open file as new lex input buffer */
1150 hsbuf_save = YY_CURRENT_BUFFER;
1151 if ((yyin = fopen(file, "r")) == NULL) {
1152 char errbuf[ERR_BUF_SIZE];
1154 sprintf(errbuf, "can't read \"%-.50s\"", file);
1157 yy_switch_to_buffer(yy_create_buffer(yyin, YY_BUF_SIZE));
1159 hslineno_save = hslineno;
1160 hsplineno_save = hsplineno;
1161 hslineno = hsplineno = 1;
1163 filename_save = input_filename;
1164 input_filename = NULL;
1166 hscolno_save = hscolno;
1167 hspcolno_save = hspcolno;
1168 hscolno = hspcolno = 0;
1169 etags_save = etags; /* do not do "etags" stuff in interfaces */
1170 etags = 0; /* We remember whether we are doing it in
1171 the module, so we can restore it later [WDP 94/09] */
1172 hsentercontext(-1); /* partain: changed this from 0 */
1173 icontexts_save = icontexts;
1175 fprintf(stderr, "reading %s (%d:%d:%d)\n", input_filename, hscolno_save, hspcolno_save, INDENTPT);
1180 layout_input(char *text, int len)
1183 fprintf(stderr, "Scanning \"%s\"\n", text);
1186 hsplineno = hslineno;
1198 hscolno += 8 - (hscolno % 8); /* Tabs stops are 8 columns apart */
1210 setstartlineno(void)
1212 startlineno = hsplineno;
1214 if (modulelineno == 0) {
1215 modulelineno = startlineno;
1221 fprintf(stderr,"%u\tsetstartlineno (col %u)\n",startlineno,hscolno);
1225 /**********************************************************************
1231 **********************************************************************/
1233 #define CACHE_SIZE YY_BUF_SIZE
1239 } textcache = { 0, 0, NULL };
1244 /* fprintf(stderr, "cleartext\n"); */
1246 if (textcache.allocated == 0) {
1247 textcache.allocated = CACHE_SIZE;
1248 textcache.text = xmalloc(CACHE_SIZE);
1253 addtext(char *text, unsigned length)
1255 /* fprintf(stderr, "addtext: %d %s\n", length, text); */
1260 if (textcache.next + length + 1 >= textcache.allocated) {
1261 textcache.allocated += length + CACHE_SIZE;
1262 textcache.text = xrealloc(textcache.text, textcache.allocated);
1264 bcopy(text, textcache.text + textcache.next, length);
1265 textcache.next += length;
1271 /* fprintf(stderr, "addchar: %c\n", c); */
1273 if (textcache.next + 2 >= textcache.allocated) {
1274 textcache.allocated += CACHE_SIZE;
1275 textcache.text = xrealloc(textcache.text, textcache.allocated);
1277 textcache.text[textcache.next++] = c;
1281 fetchtext(unsigned *length)
1283 /* fprintf(stderr, "fetchtext: %d\n", textcache.next); */
1285 *length = textcache.next;
1286 textcache.text[textcache.next] = '\0';
1287 return textcache.text;
1290 /**********************************************************************
1293 * Identifier Processing *
1296 **********************************************************************/
1299 hsnewid Enters an id of length n into the symbol table.
1303 hsnewid(char *name, int length)
1305 char save = name[length];
1307 name[length] = '\0';
1308 yylval.uid = installid(name);
1309 name[length] = save;
1313 hsnewqid(char *name, int length)
1316 char save = name[length];
1317 name[length] = '\0';
1319 dot = strchr(name, '.');
1321 yylval.uqid = mkaqual(installid(name),installid(dot+1));
1323 name[length] = save;
1325 return isconstr(dot+1);