Fix a compiler warning in initStringInfo().
[pgsql.git] / src / tools / pg_bsd_indent / indent.c
blob2622cc6227af059c81ae2f9720769150984cac3c
1 /*-
2 * Copyright (c) 1985 Sun Microsystems, Inc.
3 * Copyright (c) 1976 Board of Trustees of the University of Illinois.
4 * Copyright (c) 1980, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
32 #if 0
33 #ifndef lint
34 static char sccsid[] = "@(#)indent.c 5.17 (Berkeley) 6/7/93";
35 #endif /* not lint */
36 #endif
38 #include "c.h"
40 #include <sys/param.h>
41 #include <err.h>
42 #include <fcntl.h>
43 #include <unistd.h>
44 #include <stdio.h>
45 #include <stdlib.h>
46 #include <string.h>
47 #include <ctype.h>
49 /* Tell indent_globs.h to define our global variables here */
50 #define DECLARE_INDENT_GLOBALS 1
52 #include "indent_globs.h"
53 #include "indent_codes.h"
54 #include "indent.h"
56 static void bakcopy(void);
57 static void indent_declaration(int cur_dec_ind, int tabs_to_var);
59 const char *in_name = "Standard Input"; /* will always point to name of input
60 * file */
61 const char *out_name = "Standard Output"; /* will always point to name
62 * of output file */
63 char bakfile[MAXPGPATH] = "";
65 int
66 main(int argc, char **argv)
68 int dec_ind; /* current indentation for declarations */
69 int di_stack[20]; /* a stack of structure indentation levels */
70 int force_nl; /* when true, code must be broken */
71 int hd_type = 0; /* used to store type of stmt for if (...),
72 * for (...), etc */
73 int i; /* local loop counter */
74 int scase; /* set to true when we see a case, so we will
75 * know what to do with the following colon */
76 int sp_sw; /* when true, we are in the expression of
77 * if(...), while(...), etc. */
78 int squest; /* when this is positive, we have seen a ?
79 * without the matching : in a <c>?<s>:<s>
80 * construct */
81 const char *t_ptr; /* used for copying tokens */
82 int tabs_to_var; /* true if using tabs to indent to var name */
83 int type_code; /* the type of token, returned by lexi */
85 int last_else = 0; /* true iff last keyword was an else */
86 const char *profile_name = NULL;
87 struct parser_state transient_state; /* a copy for lookup */
90 /*-----------------------------------------------*\
91 | INITIALIZATION |
92 \*-----------------------------------------------*/
94 found_err = 0;
96 ps.p_stack[0] = stmt; /* this is the parser's stack */
97 ps.last_nl = true; /* this is true if the last thing scanned was
98 * a newline */
99 ps.last_token = semicolon;
100 combuf = (char *) malloc(bufsize);
101 if (combuf == NULL)
102 err(1, NULL);
103 labbuf = (char *) malloc(bufsize);
104 if (labbuf == NULL)
105 err(1, NULL);
106 codebuf = (char *) malloc(bufsize);
107 if (codebuf == NULL)
108 err(1, NULL);
109 tokenbuf = (char *) malloc(bufsize);
110 if (tokenbuf == NULL)
111 err(1, NULL);
112 alloc_typenames();
113 l_com = combuf + bufsize - 5;
114 l_lab = labbuf + bufsize - 5;
115 l_code = codebuf + bufsize - 5;
116 l_token = tokenbuf + bufsize - 5;
117 combuf[0] = codebuf[0] = labbuf[0] = ' '; /* set up code, label, and
118 * comment buffers */
119 combuf[1] = codebuf[1] = labbuf[1] = '\0';
120 ps.else_if = 1; /* Default else-if special processing to on */
121 s_lab = e_lab = labbuf + 1;
122 s_code = e_code = codebuf + 1;
123 s_com = e_com = combuf + 1;
124 s_token = e_token = tokenbuf + 1;
126 in_buffer = (char *) malloc(10);
127 if (in_buffer == NULL)
128 err(1, NULL);
129 in_buffer_limit = in_buffer + 8;
130 buf_ptr = buf_end = in_buffer;
131 line_no = 1;
132 had_eof = ps.in_decl = ps.decl_on_line = break_comma = false;
133 sp_sw = force_nl = false;
134 ps.in_or_st = false;
135 ps.bl_line = true;
136 dec_ind = 0;
137 di_stack[ps.dec_nest = 0] = 0;
138 ps.want_blank = ps.in_stmt = ps.ind_stmt = false;
140 scase = ps.pcase = false;
141 squest = 0;
142 sc_end = NULL;
143 bp_save = NULL;
144 be_save = NULL;
146 output = NULL;
147 tabs_to_var = 0;
149 /*--------------------------------------------------*\
150 | COMMAND LINE SCAN |
151 \*--------------------------------------------------*/
153 #ifdef undef
154 max_col = 78; /* -l78 */
155 lineup_to_parens = 1; /* -lp */
156 lineup_to_parens_always = 0; /* -nlpl */
157 ps.ljust_decl = 0; /* -ndj */
158 ps.com_ind = 33; /* -c33 */
159 star_comment_cont = 1; /* -sc */
160 ps.ind_size = 8; /* -i8 */
161 verbose = 0;
162 ps.decl_indent = 16; /* -di16 */
163 ps.local_decl_indent = -1; /* if this is not set to some nonnegative value
164 * by an arg, we will set this equal to
165 * ps.decl_ind */
166 ps.indent_parameters = 1; /* -ip */
167 ps.decl_com_ind = 0; /* if this is not set to some positive value
168 * by an arg, we will set this equal to
169 * ps.com_ind */
170 btype_2 = 1; /* -br */
171 cuddle_else = 1; /* -ce */
172 ps.unindent_displace = 0; /* -d0 */
173 ps.case_indent = 0; /* -cli0 */
174 format_block_comments = 1; /* -fcb */
175 format_col1_comments = 1; /* -fc1 */
176 procnames_start_line = 1; /* -psl */
177 proc_calls_space = 0; /* -npcs */
178 comment_delimiter_on_blankline = 1; /* -cdb */
179 ps.leave_comma = 1; /* -nbc */
180 #endif
182 for (i = 1; i < argc; ++i)
183 if (strcmp(argv[i], "-npro") == 0)
184 break;
185 else if (argv[i][0] == '-' && argv[i][1] == 'P' && argv[i][2] != '\0')
186 profile_name = argv[i]; /* non-empty -P (set profile) */
187 set_defaults();
188 if (i >= argc)
189 set_profile(profile_name);
191 for (i = 1; i < argc; ++i) {
194 * look thru args (if any) for changes to defaults
196 if (argv[i][0] != '-') {/* no flag on parameter */
197 if (input == NULL) { /* we must have the input file */
198 in_name = argv[i]; /* remember name of input file */
199 input = fopen(in_name, "r");
200 if (input == NULL) /* check for open error */
201 err(1, "%s", in_name);
202 continue;
204 else if (output == NULL) { /* we have the output file */
205 out_name = argv[i]; /* remember name of output file */
206 if (strcmp(in_name, out_name) == 0) { /* attempt to overwrite
207 * the file */
208 errx(1, "input and output files must be different");
210 output = fopen(out_name, "wb");
211 if (output == NULL) /* check for create error */
212 err(1, "%s", out_name);
213 continue;
215 errx(1, "unknown parameter: %s", argv[i]);
217 else
218 set_option(argv[i]);
219 } /* end of for */
220 if (input == NULL)
221 input = stdin;
222 if (output == NULL) {
223 if (input == stdin)
224 output = stdout;
225 else {
226 out_name = in_name;
227 bakcopy();
231 if (ps.com_ind <= 1)
232 ps.com_ind = 2; /* dont put normal comments before column 2 */
233 if (block_comment_max_col <= 0)
234 block_comment_max_col = max_col;
235 if (ps.local_decl_indent < 0) /* if not specified by user, set this */
236 ps.local_decl_indent = ps.decl_indent;
237 if (ps.decl_com_ind <= 0) /* if not specified by user, set this */
238 ps.decl_com_ind = ps.ljust_decl ? (ps.com_ind <= 10 ? 2 : ps.com_ind - 8) : ps.com_ind;
239 if (continuation_indent == 0)
240 continuation_indent = ps.ind_size;
241 fill_buffer(); /* get first batch of stuff into input buffer */
243 parse(semicolon);
245 char *p = buf_ptr;
246 int col = 1;
248 while (1) {
249 if (*p == ' ')
250 col++;
251 else if (*p == '\t')
252 col = tabsize * (1 + (col - 1) / tabsize) + 1;
253 else
254 break;
255 p++;
257 if (col > ps.ind_size)
258 ps.ind_level = ps.i_l_follow = col / ps.ind_size;
262 * START OF MAIN LOOP
265 while (1) { /* this is the main loop. it will go until we
266 * reach eof */
267 int comment_buffered = false;
269 type_code = lexi(&ps); /* lexi reads one token. The actual
270 * characters read are stored in "token". lexi
271 * returns a code indicating the type of token */
274 * The following code moves newlines and comments following an if (),
275 * while (), else, etc. up to the start of the following stmt to
276 * a buffer. This allows proper handling of both kinds of brace
277 * placement (-br, -bl) and cuddling "else" (-ce).
280 while (ps.search_brace) {
281 switch (type_code) {
282 case newline:
283 if (sc_end == NULL) {
284 save_com = sc_buf;
285 save_com[0] = save_com[1] = ' ';
286 sc_end = &save_com[2];
288 *sc_end++ = '\n';
290 * We may have inherited a force_nl == true from the previous
291 * token (like a semicolon). But once we know that a newline
292 * has been scanned in this loop, force_nl should be false.
294 * However, the force_nl == true must be preserved if newline
295 * is never scanned in this loop, so this assignment cannot be
296 * done earlier.
298 force_nl = false;
299 case form_feed:
300 break;
301 case comment:
302 if (sc_end == NULL) {
304 * Copy everything from the start of the line, because
305 * pr_comment() will use that to calculate original
306 * indentation of a boxed comment.
308 memcpy(sc_buf, in_buffer, buf_ptr - in_buffer - 4);
309 save_com = sc_buf + (buf_ptr - in_buffer - 4);
310 save_com[0] = save_com[1] = ' ';
311 sc_end = &save_com[2];
313 comment_buffered = true;
314 *sc_end++ = '/'; /* copy in start of comment */
315 *sc_end++ = '*';
316 for (;;) { /* loop until we get to the end of the comment */
317 *sc_end = *buf_ptr++;
318 if (buf_ptr >= buf_end)
319 fill_buffer();
320 if (*sc_end++ == '*' && *buf_ptr == '/')
321 break; /* we are at end of comment */
322 if (sc_end >= &save_com[sc_size]) { /* check for temp buffer
323 * overflow */
324 diag2(1, "Internal buffer overflow - Move big comment from right after if, while, or whatever");
325 fflush(output);
326 exit(1);
329 *sc_end++ = '/'; /* add ending slash */
330 if (++buf_ptr >= buf_end) /* get past / in buffer */
331 fill_buffer();
332 break;
333 case lbrace:
335 * Put KNF-style lbraces before the buffered up tokens and
336 * jump out of this loop in order to avoid copying the token
337 * again under the default case of the switch below.
339 if (sc_end != NULL && btype_2) {
340 save_com[0] = '{';
342 * Originally the lbrace may have been alone on its own
343 * line, but it will be moved into "the else's line", so
344 * if there was a newline resulting from the "{" before,
345 * it must be scanned now and ignored.
347 while (isspace((unsigned char)*buf_ptr)) {
348 if (++buf_ptr >= buf_end)
349 fill_buffer();
350 if (*buf_ptr == '\n')
351 break;
353 goto sw_buffer;
355 /* FALLTHROUGH */
356 default: /* it is the start of a normal statement */
358 int remove_newlines;
360 remove_newlines =
361 /* "} else" */
362 (type_code == sp_nparen && *token == 'e' &&
363 e_code != s_code && e_code[-1] == '}')
364 /* "else if" */
365 || (type_code == sp_paren && *token == 'i' &&
366 last_else && ps.else_if);
367 if (remove_newlines)
368 force_nl = false;
369 if (sc_end == NULL) { /* ignore buffering if
370 * comment wasn't saved up */
371 ps.search_brace = false;
372 goto check_type;
374 while (sc_end > save_com && isblank((unsigned char)sc_end[-1])) {
375 sc_end--;
377 if (swallow_optional_blanklines ||
378 (!comment_buffered && remove_newlines)) {
379 force_nl = !remove_newlines;
380 while (sc_end > save_com && sc_end[-1] == '\n') {
381 sc_end--;
384 if (force_nl) { /* if we should insert a nl here, put
385 * it into the buffer */
386 force_nl = false;
387 --line_no; /* this will be re-increased when the
388 * newline is read from the buffer */
389 *sc_end++ = '\n';
390 *sc_end++ = ' ';
391 if (verbose) /* print error msg if the line was
392 * not already broken */
393 diag2(0, "Line broken");
395 for (t_ptr = token; *t_ptr; ++t_ptr)
396 *sc_end++ = *t_ptr;
398 sw_buffer:
399 ps.search_brace = false; /* stop looking for start of
400 * stmt */
401 bp_save = buf_ptr; /* save current input buffer */
402 be_save = buf_end;
403 buf_ptr = save_com; /* fix so that subsequent calls to
404 * lexi will take tokens out of
405 * save_com */
406 *sc_end++ = ' ';/* add trailing blank, just in case */
407 buf_end = sc_end;
408 sc_end = NULL;
409 break;
411 } /* end of switch */
413 * We must make this check, just in case there was an unexpected
414 * EOF.
416 if (type_code != 0) {
418 * The only intended purpose of calling lexi() below is to
419 * categorize the next token in order to decide whether to
420 * continue buffering forthcoming tokens. Once the buffering
421 * is over, lexi() will be called again elsewhere on all of
422 * the tokens - this time for normal processing.
424 * Calling it for this purpose is a bug, because lexi() also
425 * changes the parser state and discards leading whitespace,
426 * which is needed mostly for comment-related considerations.
428 * Work around the former problem by giving lexi() a copy of
429 * the current parser state and discard it if the call turned
430 * out to be just a look ahead.
432 * Work around the latter problem by copying all whitespace
433 * characters into the buffer so that the later lexi() call
434 * will read them.
436 if (sc_end != NULL) {
437 while (*buf_ptr == ' ' || *buf_ptr == '\t') {
438 *sc_end++ = *buf_ptr++;
439 if (sc_end >= &save_com[sc_size]) {
440 errx(1, "input too long");
443 if (buf_ptr >= buf_end) {
444 fill_buffer();
447 transient_state = ps;
448 type_code = lexi(&transient_state); /* read another token */
449 if (type_code != newline && type_code != form_feed &&
450 type_code != comment && !transient_state.search_brace) {
451 ps = transient_state;
454 } /* end of while (search_brace) */
455 last_else = 0;
456 check_type:
457 if (type_code == 0) { /* we got eof */
458 if (s_lab != e_lab || s_code != e_code
459 || s_com != e_com) /* must dump end of line */
460 dump_line();
461 if (ps.tos > 1) /* check for balanced braces */
462 diag2(1, "Stuff missing from end of file");
464 if (verbose) {
465 printf("There were %d output lines and %d comments\n",
466 ps.out_lines, ps.out_coms);
467 printf("(Lines with comments)/(Lines with code): %6.3f\n",
468 (1.0 * ps.com_lines) / code_lines);
470 fflush(output);
471 exit(found_err);
473 if (
474 (type_code != comment) &&
475 (type_code != newline) &&
476 (type_code != preesc) &&
477 (type_code != form_feed)) {
478 if (force_nl &&
479 (type_code != semicolon) &&
480 (type_code != lbrace || !btype_2)) {
481 /* we should force a broken line here */
482 if (verbose)
483 diag2(0, "Line broken");
484 dump_line();
485 ps.want_blank = false; /* dont insert blank at line start */
486 force_nl = false;
488 ps.in_stmt = true; /* turn on flag which causes an extra level of
489 * indentation. this is turned off by a ; or
490 * '}' */
491 if (s_com != e_com) { /* the turkey has embedded a comment
492 * in a line. fix it */
493 int len = e_com - s_com;
495 CHECK_SIZE_CODE(len + 3);
496 *e_code++ = ' ';
497 memcpy(e_code, s_com, len);
498 e_code += len;
499 *e_code++ = ' ';
500 *e_code = '\0'; /* null terminate code sect */
501 ps.want_blank = false;
502 e_com = s_com;
505 else if (type_code != comment) /* preserve force_nl thru a comment */
506 force_nl = false; /* cancel forced newline after newline, form
507 * feed, etc */
511 /*-----------------------------------------------------*\
512 | do switch on type of token scanned |
513 \*-----------------------------------------------------*/
514 CHECK_SIZE_CODE(3); /* maximum number of increments of e_code
515 * before the next CHECK_SIZE_CODE or
516 * dump_line() is 2. After that there's the
517 * final increment for the null character. */
518 switch (type_code) { /* now, decide what to do with the token */
520 case form_feed: /* found a form feed in line */
521 ps.use_ff = true; /* a form feed is treated much like a newline */
522 dump_line();
523 ps.want_blank = false;
524 break;
526 case newline:
527 if (ps.last_token != comma || ps.p_l_follow > 0
528 || !ps.leave_comma || ps.block_init || !break_comma || s_com != e_com) {
529 dump_line();
530 ps.want_blank = false;
532 ++line_no; /* keep track of input line number */
533 break;
535 case lparen: /* got a '(' or '[' */
536 /* count parens to make Healy happy */
537 if (++ps.p_l_follow == nitems(ps.paren_indents)) {
538 diag3(0, "Reached internal limit of %d unclosed parens",
539 nitems(ps.paren_indents));
540 ps.p_l_follow--;
542 if (*token == '[')
543 /* not a function pointer declaration or a function call */;
544 else if (ps.in_decl && !ps.block_init && !ps.dumped_decl_indent &&
545 ps.procname[0] == '\0' && ps.paren_level == 0) {
546 /* function pointer declarations */
547 indent_declaration(dec_ind, tabs_to_var);
548 ps.dumped_decl_indent = true;
550 else if (ps.want_blank &&
551 ((ps.last_token != ident && ps.last_token != funcname) ||
552 /* offsetof (1) is never allowed a space; sizeof (2) gets
553 * one iff -bs; all other keywords (>2) always get a space
554 * before lparen */
555 ps.keyword + Bill_Shannon > 2))
556 *e_code++ = ' ';
557 ps.want_blank = false;
558 *e_code++ = token[0];
559 ps.paren_indents[ps.p_l_follow - 1] = count_spaces_until(1, s_code, e_code) - 1;
560 if (sp_sw && ps.p_l_follow == 1 && extra_expression_indent
561 && ps.paren_indents[0] < 2 * ps.ind_size)
562 ps.paren_indents[0] = 2 * ps.ind_size;
563 if (ps.in_or_st && *token == '(' && ps.tos <= 2) {
565 * this is a kluge to make sure that declarations will be
566 * aligned right if proc decl has an explicit type on it, i.e.
567 * "int a(x) {..."
569 parse(semicolon); /* I said this was a kluge... */
570 ps.in_or_st = false; /* turn off flag for structure decl or
571 * initialization */
574 * parenthesized type following sizeof or offsetof is not a cast,
575 * and we assume the same for any other non-keyword identifier,
576 * to support macros that take types
578 if (ps.last_token == ident &&
579 (ps.keyword == 0 || ps.keyword == 1 || ps.keyword == 2))
580 ps.not_cast_mask |= 1 << ps.p_l_follow;
581 break;
583 case rparen: /* got a ')' or ']' */
584 if (ps.cast_mask & (1 << ps.p_l_follow) & ~ps.not_cast_mask) {
585 ps.last_u_d = true;
586 ps.cast_mask &= (1 << ps.p_l_follow) - 1;
587 ps.want_blank = space_after_cast;
588 } else
589 ps.want_blank = true;
590 ps.not_cast_mask &= (1 << ps.p_l_follow) - 1;
591 if (--ps.p_l_follow < 0) {
592 ps.p_l_follow = 0;
593 diag3(0, "Extra %c", *token);
595 if (e_code == s_code) /* if the paren starts the line */
596 ps.paren_level = ps.p_l_follow; /* then indent it */
598 *e_code++ = token[0];
600 if (sp_sw && (ps.p_l_follow == 0)) { /* check for end of if
601 * (...), or some such */
602 sp_sw = false;
603 force_nl = true;/* must force newline after if */
604 ps.last_u_d = true; /* inform lexi that a following
605 * operator is unary */
606 ps.in_stmt = false; /* dont use stmt continuation
607 * indentation */
609 parse(hd_type); /* let parser worry about if, or whatever */
611 ps.search_brace = btype_2; /* this should ensure that constructs
612 * such as main(){...} and int[]{...}
613 * have their braces put in the right
614 * place */
615 break;
617 case unary_op: /* this could be any unary operation */
618 if (!ps.dumped_decl_indent && ps.in_decl && !ps.block_init &&
619 ps.procname[0] == '\0' && ps.paren_level == 0) {
620 /* pointer declarations */
623 * if this is a unary op in a declaration, we should indent
624 * this token
626 for (i = 0; token[i]; ++i)
627 /* find length of token */;
628 indent_declaration(dec_ind - i, tabs_to_var);
629 ps.dumped_decl_indent = true;
631 else if (ps.want_blank)
632 *e_code++ = ' ';
635 int len = e_token - s_token;
637 CHECK_SIZE_CODE(len);
638 memcpy(e_code, token, len);
639 e_code += len;
641 ps.want_blank = false;
642 break;
644 case binary_op: /* any binary operation */
646 int len = e_token - s_token;
648 CHECK_SIZE_CODE(len + 1);
649 if (ps.want_blank)
650 *e_code++ = ' ';
651 memcpy(e_code, token, len);
652 e_code += len;
654 ps.want_blank = true;
655 break;
657 case postop: /* got a trailing ++ or -- */
658 *e_code++ = token[0];
659 *e_code++ = token[1];
660 ps.want_blank = true;
661 break;
663 case question: /* got a ? */
664 squest++; /* this will be used when a later colon
665 * appears so we can distinguish the
666 * <c>?<n>:<n> construct */
667 if (ps.want_blank)
668 *e_code++ = ' ';
669 *e_code++ = '?';
670 ps.want_blank = true;
671 break;
673 case casestmt: /* got word 'case' or 'default' */
674 scase = true; /* so we can process the later colon properly */
675 goto copy_id;
677 case colon: /* got a ':' */
678 if (squest > 0) { /* it is part of the <c>?<n>: <n> construct */
679 --squest;
680 if (ps.want_blank)
681 *e_code++ = ' ';
682 *e_code++ = ':';
683 ps.want_blank = true;
684 break;
686 if (ps.in_or_st) {
687 *e_code++ = ':';
688 ps.want_blank = false;
689 break;
691 ps.in_stmt = false; /* seeing a label does not imply we are in a
692 * stmt */
694 * turn everything so far into a label
697 int len = e_code - s_code;
699 CHECK_SIZE_LAB(len + 3);
700 memcpy(e_lab, s_code, len);
701 e_lab += len;
702 *e_lab++ = ':';
703 *e_lab = '\0';
704 e_code = s_code;
706 force_nl = ps.pcase = scase; /* ps.pcase will be used by
707 * dump_line to decide how to
708 * indent the label. force_nl
709 * will force a case n: to be
710 * on a line by itself */
711 scase = false;
712 ps.want_blank = false;
713 break;
715 case semicolon: /* got a ';' */
716 if (ps.dec_nest == 0)
717 ps.in_or_st = false;/* we are not in an initialization or
718 * structure declaration */
719 scase = false; /* these will only need resetting in an error */
720 squest = 0;
721 if (ps.last_token == rparen)
722 ps.in_parameter_declaration = 0;
723 ps.cast_mask = 0;
724 ps.not_cast_mask = 0;
725 ps.block_init = 0;
726 ps.block_init_level = 0;
727 ps.just_saw_decl--;
729 if (ps.in_decl && s_code == e_code && !ps.block_init &&
730 !ps.dumped_decl_indent && ps.paren_level == 0) {
731 /* indent stray semicolons in declarations */
732 indent_declaration(dec_ind - 1, tabs_to_var);
733 ps.dumped_decl_indent = true;
736 ps.in_decl = (ps.dec_nest > 0); /* if we were in a first level
737 * structure declaration, we
738 * arent any more */
740 if ((!sp_sw || hd_type != forstmt) && ps.p_l_follow > 0) {
743 * This should be true iff there were unbalanced parens in the
744 * stmt. It is a bit complicated, because the semicolon might
745 * be in a for stmt
747 diag2(1, "Unbalanced parens");
748 ps.p_l_follow = 0;
749 if (sp_sw) { /* this is a check for an if, while, etc. with
750 * unbalanced parens */
751 sp_sw = false;
752 parse(hd_type); /* dont lose the if, or whatever */
755 *e_code++ = ';';
756 ps.want_blank = true;
757 ps.in_stmt = (ps.p_l_follow > 0); /* we are no longer in the
758 * middle of a stmt */
760 if (!sp_sw) { /* if not if for (;;) */
761 parse(semicolon); /* let parser know about end of stmt */
762 force_nl = true;/* force newline after an end of stmt */
764 break;
766 case lbrace: /* got a '{' */
767 ps.in_stmt = false; /* dont indent the {} */
768 if (!ps.block_init)
769 force_nl = true;/* force other stuff on same line as '{' onto
770 * new line */
771 else if (ps.block_init_level <= 0)
772 ps.block_init_level = 1;
773 else
774 ps.block_init_level++;
776 if (s_code != e_code && !ps.block_init) {
777 if (!btype_2) {
778 dump_line();
779 ps.want_blank = false;
781 else if (ps.in_parameter_declaration && !ps.in_or_st) {
782 ps.i_l_follow = 0;
783 if (function_brace_split) { /* dump the line prior to the
784 * brace ... */
785 dump_line();
786 ps.want_blank = false;
787 } else /* add a space between the decl and brace */
788 ps.want_blank = true;
791 if (ps.in_parameter_declaration)
792 prefix_blankline_requested = 0;
794 if (ps.p_l_follow > 0) { /* check for preceding unbalanced
795 * parens */
796 diag2(1, "Unbalanced parens");
797 ps.p_l_follow = 0;
798 if (sp_sw) { /* check for unclosed if, for, etc. */
799 sp_sw = false;
800 parse(hd_type);
801 ps.ind_level = ps.i_l_follow;
804 if (s_code == e_code)
805 ps.ind_stmt = false; /* dont put extra indentation on line
806 * with '{' */
807 if (ps.in_decl && ps.in_or_st) { /* this is either a structure
808 * declaration or an init */
809 di_stack[ps.dec_nest] = dec_ind;
810 if (++ps.dec_nest == nitems(di_stack)) {
811 diag3(0, "Reached internal limit of %d struct levels",
812 nitems(di_stack));
813 ps.dec_nest--;
815 /* ? dec_ind = 0; */
817 else {
818 ps.decl_on_line = false; /* we can't be in the middle of
819 * a declaration, so don't do
820 * special indentation of
821 * comments */
822 if (blanklines_after_declarations_at_proctop
823 && ps.in_parameter_declaration)
824 postfix_blankline_requested = 1;
825 ps.in_parameter_declaration = 0;
826 ps.in_decl = false;
828 dec_ind = 0;
829 parse(lbrace); /* let parser know about this */
830 if (ps.want_blank) /* put a blank before '{' if '{' is not at
831 * start of line */
832 *e_code++ = ' ';
833 ps.want_blank = false;
834 *e_code++ = '{';
835 ps.just_saw_decl = 0;
836 break;
838 case rbrace: /* got a '}' */
839 if (ps.p_stack[ps.tos] == decl && !ps.block_init) /* semicolons can be
840 * omitted in
841 * declarations */
842 parse(semicolon);
843 if (ps.p_l_follow) {/* check for unclosed if, for, else. */
844 diag2(1, "Unbalanced parens");
845 ps.p_l_follow = 0;
846 sp_sw = false;
848 ps.just_saw_decl = 0;
849 ps.block_init_level--;
850 if (s_code != e_code && !ps.block_init) { /* '}' must be first on
851 * line */
852 if (verbose)
853 diag2(0, "Line broken");
854 dump_line();
856 *e_code++ = '}';
857 ps.want_blank = true;
858 ps.in_stmt = ps.ind_stmt = false;
859 if (ps.dec_nest > 0) { /* we are in multi-level structure
860 * declaration */
861 dec_ind = di_stack[--ps.dec_nest];
862 if (ps.dec_nest == 0 && !ps.in_parameter_declaration)
863 ps.just_saw_decl = 2;
864 ps.in_decl = true;
866 prefix_blankline_requested = 0;
867 parse(rbrace); /* let parser know about this */
868 ps.search_brace = cuddle_else && ps.p_stack[ps.tos] == ifhead
869 && ps.il[ps.tos] >= ps.ind_level;
870 if (ps.tos <= 1 && blanklines_after_procs && ps.dec_nest <= 0)
871 postfix_blankline_requested = 1;
872 break;
874 case swstmt: /* got keyword "switch" */
875 sp_sw = true;
876 hd_type = swstmt; /* keep this for when we have seen the
877 * expression */
878 goto copy_id; /* go move the token into buffer */
880 case sp_paren: /* token is if, while, for */
881 sp_sw = true; /* the interesting stuff is done after the
882 * expression is scanned */
883 hd_type = (*token == 'i' ? ifstmt :
884 (*token == 'w' ? whilestmt : forstmt));
887 * remember the type of header for later use by parser
889 goto copy_id; /* copy the token into line */
891 case sp_nparen: /* got else, do */
892 ps.in_stmt = false;
893 if (*token == 'e') {
894 if (e_code != s_code && (!cuddle_else || e_code[-1] != '}')) {
895 if (verbose)
896 diag2(0, "Line broken");
897 dump_line();/* make sure this starts a line */
898 ps.want_blank = false;
900 force_nl = true;/* also, following stuff must go onto new line */
901 last_else = 1;
902 parse(elselit);
904 else {
905 if (e_code != s_code) { /* make sure this starts a line */
906 if (verbose)
907 diag2(0, "Line broken");
908 dump_line();
909 ps.want_blank = false;
911 force_nl = true;/* also, following stuff must go onto new line */
912 last_else = 0;
913 parse(dolit);
915 goto copy_id; /* move the token into line */
917 case type_def:
918 case storage:
919 prefix_blankline_requested = 0;
920 goto copy_id;
922 case structure:
923 if (ps.p_l_follow > 0)
924 goto copy_id;
925 /* FALLTHROUGH */
926 case decl: /* we have a declaration type (int, etc.) */
927 parse(decl); /* let parser worry about indentation */
928 if (ps.last_token == rparen && ps.tos <= 1) {
929 if (s_code != e_code) {
930 dump_line();
931 ps.want_blank = 0;
934 if (ps.in_parameter_declaration && ps.indent_parameters && ps.dec_nest == 0) {
935 ps.ind_level = ps.i_l_follow = 1;
936 ps.ind_stmt = 0;
938 ps.in_or_st = true; /* this might be a structure or initialization
939 * declaration */
940 ps.in_decl = ps.decl_on_line = true;
941 if ( /* !ps.in_or_st && */ ps.dec_nest <= 0)
942 ps.just_saw_decl = 2;
943 prefix_blankline_requested = 0;
944 for (i = 0; token[i++];); /* get length of token */
946 if (ps.ind_level == 0 || ps.dec_nest > 0) {
947 /* global variable or struct member in local variable */
948 dec_ind = ps.decl_indent > 0 ? ps.decl_indent : i;
949 tabs_to_var = (use_tabs ? ps.decl_indent > 0 : 0);
950 } else {
951 /* local variable */
952 dec_ind = ps.local_decl_indent > 0 ? ps.local_decl_indent : i;
953 tabs_to_var = (use_tabs ? ps.local_decl_indent > 0 : 0);
955 goto copy_id;
957 case funcname:
958 case ident: /* got an identifier or constant */
959 if (ps.in_decl) {
960 if (type_code == funcname) {
961 ps.in_decl = false;
962 if (procnames_start_line && s_code != e_code) {
963 *e_code = '\0';
964 dump_line();
966 else if (ps.want_blank) {
967 *e_code++ = ' ';
969 ps.want_blank = false;
971 else if (!ps.block_init && !ps.dumped_decl_indent &&
972 ps.paren_level == 0) { /* if we are in a declaration, we
973 * must indent identifier */
974 indent_declaration(dec_ind, tabs_to_var);
975 ps.dumped_decl_indent = true;
976 ps.want_blank = false;
979 else if (sp_sw && ps.p_l_follow == 0) {
980 sp_sw = false;
981 force_nl = true;
982 ps.last_u_d = true;
983 ps.in_stmt = false;
984 parse(hd_type);
986 copy_id:
988 int len = e_token - s_token;
990 CHECK_SIZE_CODE(len + 1);
991 if (ps.want_blank)
992 *e_code++ = ' ';
993 memcpy(e_code, s_token, len);
994 e_code += len;
996 if (type_code != funcname)
997 ps.want_blank = true;
998 break;
1000 case strpfx:
1002 int len = e_token - s_token;
1004 CHECK_SIZE_CODE(len + 1);
1005 if (ps.want_blank)
1006 *e_code++ = ' ';
1007 memcpy(e_code, token, len);
1008 e_code += len;
1010 ps.want_blank = false;
1011 break;
1013 case period: /* treat a period kind of like a binary
1014 * operation */
1015 *e_code++ = '.'; /* move the period into line */
1016 ps.want_blank = false; /* dont put a blank after a period */
1017 break;
1019 case comma:
1020 ps.want_blank = (s_code != e_code); /* only put blank after comma
1021 * if comma does not start the
1022 * line */
1023 if (ps.in_decl && ps.procname[0] == '\0' && !ps.block_init &&
1024 !ps.dumped_decl_indent && ps.paren_level == 0) {
1025 /* indent leading commas and not the actual identifiers */
1026 indent_declaration(dec_ind - 1, tabs_to_var);
1027 ps.dumped_decl_indent = true;
1029 *e_code++ = ',';
1030 if (ps.p_l_follow == 0) {
1031 if (ps.block_init_level <= 0)
1032 ps.block_init = 0;
1033 if (break_comma && (!ps.leave_comma ||
1034 count_spaces_until(compute_code_target(), s_code, e_code) >
1035 max_col - tabsize))
1036 force_nl = true;
1038 break;
1040 case preesc: /* got the character '#' */
1041 if ((s_com != e_com) ||
1042 (s_lab != e_lab) ||
1043 (s_code != e_code))
1044 dump_line();
1045 CHECK_SIZE_LAB(1);
1046 *e_lab++ = '#'; /* move whole line to 'label' buffer */
1048 int in_comment = 0;
1049 int com_start = 0;
1050 char quote = 0;
1051 int com_end = 0;
1053 while (*buf_ptr == ' ' || *buf_ptr == '\t') {
1054 buf_ptr++;
1055 if (buf_ptr >= buf_end)
1056 fill_buffer();
1058 while (*buf_ptr != '\n' || (in_comment && !had_eof)) {
1059 CHECK_SIZE_LAB(2);
1060 *e_lab = *buf_ptr++;
1061 if (buf_ptr >= buf_end)
1062 fill_buffer();
1063 switch (*e_lab++) {
1064 case BACKSLASH:
1065 if (!in_comment) {
1066 *e_lab++ = *buf_ptr++;
1067 if (buf_ptr >= buf_end)
1068 fill_buffer();
1070 break;
1071 case '/':
1072 if (*buf_ptr == '*' && !in_comment && !quote) {
1073 in_comment = 1;
1074 *e_lab++ = *buf_ptr++;
1075 com_start = e_lab - s_lab - 2;
1077 break;
1078 case '"':
1079 if (quote == '"')
1080 quote = 0;
1081 break;
1082 case '\'':
1083 if (quote == '\'')
1084 quote = 0;
1085 break;
1086 case '*':
1087 if (*buf_ptr == '/' && in_comment) {
1088 in_comment = 0;
1089 *e_lab++ = *buf_ptr++;
1090 com_end = e_lab - s_lab;
1092 break;
1096 while (e_lab > s_lab && (e_lab[-1] == ' ' || e_lab[-1] == '\t'))
1097 e_lab--;
1098 if (e_lab - s_lab == com_end && bp_save == NULL) {
1099 /* comment on preprocessor line */
1100 if (sc_end == NULL) { /* if this is the first comment,
1101 * we must set up the buffer */
1102 save_com = sc_buf;
1103 sc_end = &save_com[0];
1105 else {
1106 *sc_end++ = '\n'; /* add newline between
1107 * comments */
1108 *sc_end++ = ' ';
1109 --line_no;
1111 if (sc_end - save_com + com_end - com_start > sc_size)
1112 errx(1, "input too long");
1113 memmove(sc_end, s_lab + com_start, com_end - com_start);
1114 sc_end += com_end - com_start;
1115 e_lab = s_lab + com_start;
1116 while (e_lab > s_lab && (e_lab[-1] == ' ' || e_lab[-1] == '\t'))
1117 e_lab--;
1118 bp_save = buf_ptr; /* save current input buffer */
1119 be_save = buf_end;
1120 buf_ptr = save_com; /* fix so that subsequent calls to
1121 * lexi will take tokens out of
1122 * save_com */
1123 *sc_end++ = ' '; /* add trailing blank, just in case */
1124 buf_end = sc_end;
1125 sc_end = NULL;
1127 CHECK_SIZE_LAB(1);
1128 *e_lab = '\0'; /* null terminate line */
1129 ps.pcase = false;
1132 if (strncmp(s_lab, "#if", 3) == 0) { /* also ifdef, ifndef */
1133 if ((size_t)ifdef_level < nitems(state_stack)) {
1134 match_state[ifdef_level].tos = -1;
1135 state_stack[ifdef_level++] = ps;
1137 else
1138 diag2(1, "#if stack overflow");
1140 else if (strncmp(s_lab, "#el", 3) == 0) { /* else, elif */
1141 if (ifdef_level <= 0)
1142 diag2(1, s_lab[3] == 'i' ? "Unmatched #elif" : "Unmatched #else");
1143 else {
1144 match_state[ifdef_level - 1] = ps;
1145 ps = state_stack[ifdef_level - 1];
1148 else if (strncmp(s_lab, "#endif", 6) == 0) {
1149 if (ifdef_level <= 0)
1150 diag2(1, "Unmatched #endif");
1151 else
1152 ifdef_level--;
1153 } else {
1154 struct directives {
1155 int size;
1156 const char *string;
1158 recognized[] = {
1159 {7, "include"},
1160 {6, "define"},
1161 {5, "undef"},
1162 {4, "line"},
1163 {5, "error"},
1164 {6, "pragma"}
1166 int d = nitems(recognized);
1167 while (--d >= 0)
1168 if (strncmp(s_lab + 1, recognized[d].string, recognized[d].size) == 0)
1169 break;
1170 if (d < 0) {
1171 diag2(1, "Unrecognized cpp directive");
1172 break;
1175 if (blanklines_around_conditional_compilation) {
1176 postfix_blankline_requested++;
1177 n_real_blanklines = 0;
1179 else {
1180 postfix_blankline_requested = 0;
1181 prefix_blankline_requested = 0;
1183 break; /* subsequent processing of the newline
1184 * character will cause the line to be printed */
1186 case comment: /* we have gotten a / followed by * this is a biggie */
1187 pr_comment();
1188 break;
1189 } /* end of big switch stmt */
1191 *e_code = '\0'; /* make sure code section is null terminated */
1192 if (type_code != comment && type_code != newline && type_code != preesc)
1193 ps.last_token = type_code;
1194 } /* end of main while (1) loop */
1198 * copy input file to backup file if in_name is /blah/blah/blah/file, then
1199 * backup file will be ".Bfile" then make the backup file the input and
1200 * original input file the output
1202 static void
1203 bakcopy(void)
1205 int n,
1206 bakchn;
1207 char buff[8 * 1024];
1208 const char *p;
1210 /* construct file name .Bfile */
1211 for (p = in_name; *p; p++); /* skip to end of string */
1212 while (p > in_name && *p != '/') /* find last '/' */
1213 p--;
1214 if (*p == '/')
1215 p++;
1216 sprintf(bakfile, "%s.BAK", p);
1218 /* copy in_name to backup file */
1219 bakchn = creat(bakfile, 0600);
1220 if (bakchn < 0)
1221 err(1, "%s", bakfile);
1222 while ((n = read(fileno(input), buff, sizeof(buff))) > 0)
1223 if (write(bakchn, buff, n) != n)
1224 err(1, "%s", bakfile);
1225 if (n < 0)
1226 err(1, "%s", in_name);
1227 close(bakchn);
1228 fclose(input);
1230 /* re-open backup file as the input file */
1231 input = fopen(bakfile, "r");
1232 if (input == NULL)
1233 err(1, "%s", bakfile);
1234 /* now the original input file will be the output */
1235 output = fopen(in_name, "wb");
1236 if (output == NULL) {
1237 unlink(bakfile);
1238 err(1, "%s", in_name);
1242 static void
1243 indent_declaration(int cur_dec_ind, int tabs_to_var)
1245 int pos = e_code - s_code;
1246 char *startpos = e_code;
1249 * get the tab math right for indentations that are not multiples of tabsize
1251 if ((ps.ind_level * ps.ind_size) % tabsize != 0) {
1252 pos += (ps.ind_level * ps.ind_size) % tabsize;
1253 cur_dec_ind += (ps.ind_level * ps.ind_size) % tabsize;
1255 if (tabs_to_var) {
1256 int tpos;
1258 CHECK_SIZE_CODE(cur_dec_ind / tabsize);
1259 while ((tpos = tabsize * (1 + pos / tabsize)) <= cur_dec_ind) {
1260 *e_code++ = (!postgres_tab_rules ||
1261 tpos != pos + 1 ||
1262 cur_dec_ind >= tpos + tabsize) ? '\t' : ' ';
1263 pos = tpos;
1266 CHECK_SIZE_CODE(cur_dec_ind - pos + 1);
1267 while (pos < cur_dec_ind) {
1268 *e_code++ = ' ';
1269 pos++;
1271 if (e_code == startpos && ps.want_blank) {
1272 *e_code++ = ' ';
1273 ps.want_blank = false;