1 /*@ S-nail - a mail user agent derived from Berkeley Mail.
2 *@ Shell "word", file- and other name expansions, incl. file globbing.
3 *@ TODO v15: peek signal states while opendir/readdir/etc.
5 * Copyright (c) 2000-2004 Gunnar Ritter, Freiburg i. Br., Germany.
6 * Copyright (c) 2012 - 2017 Steffen (Daode) Nurpmeso <steffen@sdaoden.eu>.
9 * Copyright (c) 1980, 1993
10 * The Regents of the University of California. All rights reserved.
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39 #ifndef HAVE_AMALGAMATION
51 * Environment variable names used by the utilities in the Shell and
52 * Utilities volume of POSIX.1-2008 consist solely of uppercase
53 * letters, digits, and the <underscore> ('_') from the characters
54 * defined in Portable Character Set and do not begin with a digit.
55 * Other characters may be permitted by an implementation;
56 * applications shall tolerate the presence of such names.
57 * We do support the hyphen-minus "-" (except in last position for ${x[:]-y}).
58 * We support some special parameter names for one-letter(++) variable names;
59 * these have counterparts in the code that manages internal variables,
60 * and some more special treatment below! */
61 #define a_SHEXP_ISVARC(C) (alnumchar(C) || (C) == '_' || (C) == '-')
62 #define a_SHEXP_ISVARC_BAD1ST(C) (digitchar(C)) /* (Actually assumed below!) */
63 #define a_SHEXP_ISVARC_BADNST(C) ((C) == '-')
65 enum a_shexp_quote_flags
{
67 a_SHEXP_QUOTE_ROUNDTRIP
= 1u<<0, /* Result won't be consumed immediately */
69 a_SHEXP_QUOTE_T_REVSOL
= 1u<<8, /* Type: by reverse solidus */
70 a_SHEXP_QUOTE_T_SINGLE
= 1u<<9, /* Type: single-quotes */
71 a_SHEXP_QUOTE_T_DOUBLE
= 1u<<10, /* Type: double-quotes */
72 a_SHEXP_QUOTE_T_DOLLAR
= 1u<<11, /* Type: dollar-single-quotes */
73 a_SHEXP_QUOTE_T_MASK
= a_SHEXP_QUOTE_T_REVSOL
| a_SHEXP_QUOTE_T_SINGLE
|
74 a_SHEXP_QUOTE_T_DOUBLE
| a_SHEXP_QUOTE_T_DOLLAR
,
76 a_SHEXP_QUOTE__FREESHIFT
= 16u
80 struct a_shexp_glob_ctx
{
81 char const *sgc_patdat
; /* Remaining pattern (at and below level) */
83 struct n_string
*sgc_outer
; /* Resolved path up to this level */
89 struct a_shexp_quote_ctx
{
90 struct n_string
*sqc_store
; /* Result storage */
91 struct str sqc_input
; /* Input data, topmost level */
93 ui32_t sqc_cnt_single
;
94 ui32_t sqc_cnt_double
;
95 ui32_t sqc_cnt_dollar
;
96 enum a_shexp_quote_flags sqc_flags
;
100 struct a_shexp_quote_lvl
{
101 struct a_shexp_quote_lvl
*sql_link
; /* Outer level */
102 struct str sql_dat
; /* This level (has to) handle(d) */
103 enum a_shexp_quote_flags sql_flags
;
107 /* Locate the user's mailbox file (where new, unread mail is queued) */
108 static char *a_shexp_findmail(char const *user
, bool_t force
);
110 /* Expand ^~/? and ^~USER/? constructs.
111 * Returns the completely resolved (maybe empty or identical to input)
112 * salloc()ed string */
113 static char *a_shexp_tilde(char const *s
);
115 /* Perform fnmatch(3). May return NULL on error */
116 static char *a_shexp_globname(char const *name
, enum fexp_mode fexpm
);
118 static bool_t
a_shexp__glob(struct a_shexp_glob_ctx
*sgcp
,
119 struct n_strlist
**slpp
);
120 static int a_shexp__globsort(void const *cvpa
, void const *cvpb
);
123 /* Parse an input string and create a sh(1)ell-quoted result */
124 static void a_shexp__quote(struct a_shexp_quote_ctx
*sqcp
,
125 struct a_shexp_quote_lvl
*sqlp
);
128 a_shexp_findmail(char const *user
, bool_t force
){
134 if((cp
= ok_vlook(inbox
)) != NULL
&& *cp
!= '\0'){
135 /* _NFOLDER extra introduced to avoid % recursion loops */
136 if((rv
= fexpand(cp
, FEXP_NSPECIAL
| FEXP_NFOLDER
| FEXP_NSHELL
)
139 n_err(_("*inbox* expansion failed, using $MAIL / built-in: %s\n"), cp
);
141 /* Heirloom compatibility: an IMAP *folder* becomes "%" */
143 else if(cp
== NULL
&& !strcmp(user
, ok_vlook(LOGNAME
)) &&
144 which_protocol(cp
= n_folder_query(), FAL0
, FAL0
, NULL
)
146 /* TODO Compat handling of *folder* with IMAP! */
147 n_OBSOLETE("no more expansion of *folder* in \"%\": "
148 "please set *inbox*");
154 if((cp
= ok_vlook(MAIL
)) != NULL
){
163 ul
= strlen(user
) +1;
164 i
= sizeof(VAL_MAIL
) -1 + 1 + ul
;
167 memcpy(rv
, VAL_MAIL
, (i
= sizeof(VAL_MAIL
) -1));
169 memcpy(&rv
[++i
], user
, ul
);
177 a_shexp_tilde(char const *s
){
184 if(*(rp
= &s
[1]) == '/' || *rp
== '\0'){
188 if((rp
= strchr(np
= rp
, '/')) != NULL
){
189 nl
= PTR2SIZE(rp
- np
);
190 np
= savestrbuf(np
, nl
);
195 if((pwp
= getpwnam(np
)) == NULL
){
203 rv
= salloc(nl
+ 1 + rl
+1);
206 memcpy(rv
+ nl
, rp
, rl
);
216 a_shexp_globname(char const *name
, enum fexp_mode fexpm
){
218 struct a_shexp_glob_ctx sgc
;
219 struct n_string outer
;
220 struct n_strlist
*slp
;
224 memset(&sgc
, 0, sizeof sgc
);
225 sgc
.sgc_patlen
= strlen(name
);
226 sgc
.sgc_patdat
= savestrbuf(name
, sgc
.sgc_patlen
);
227 sgc
.sgc_outer
= n_string_reserve(n_string_creat(&outer
), sgc
.sgc_patlen
);
228 sgc
.sgc_flags
= ((fexpm
& FEXP_SILENT
) != 0);
230 if(a_shexp__glob(&sgc
, &slp
))
234 n_string_gut(&outer
);
240 cp
= n_UNCONST(N_("File pattern does not match"));
242 }else if(slp
->sl_next
== NULL
)
243 cp
= savestrbuf(slp
->sl_dat
, slp
->sl_len
);
244 else if(fexpm
& FEXP_MULTIOK
){
245 struct n_strlist
**sorta
, *xslp
;
249 for(xslp
= slp
; xslp
!= NULL
; xslp
= xslp
->sl_next
){
251 l
+= xslp
->sl_len
+ 1;
254 sorta
= smalloc(sizeof(*sorta
) * no
);
256 for(xslp
= slp
; xslp
!= NULL
; xslp
= xslp
->sl_next
)
258 qsort(sorta
, no
, sizeof *sorta
, &a_shexp__globsort
);
262 for(i
= 0; i
< no
; ++i
){
264 memcpy(&cp
[l
], xslp
->sl_dat
, xslp
->sl_len
);
271 n_pstate
|= n_PS_EXPAND_MULTIRESULT
;
273 cp
= n_UNCONST(N_("File pattern matches multiple results"));
279 struct n_strlist
*tmp
= slp
;
288 if(!(fexpm
& FEXP_SILENT
)){
289 name
= n_shexp_quote_cp(name
, FAL0
);
290 n_err("%s: %s\n", V_(cp
), name
);
295 #else /* HAVE_FNMATCH */
298 if(!(fexpm
& FEXP_SILENT
))
299 n_err(_("No filename pattern (fnmatch(3)) support compiled in\n"));
300 return savestr(name
);
306 a_shexp__glob(struct a_shexp_glob_ctx
*sgcp
, struct n_strlist
**slpp
){
307 enum{a_SILENT
= 1<<0, a_DEEP
=1<<1, a_SALLOC
=1<<2};
309 struct a_shexp_glob_ctx nsgc
;
313 char const *ccp
, *myp
;
316 /* We need some special treatment for the outermost level */
317 if(!(sgcp
->sgc_flags
& a_DEEP
)){
318 if(sgcp
->sgc_patlen
> 0 && sgcp
->sgc_patdat
[0] == '/'){
319 myp
= n_string_cp(n_string_push_c(sgcp
->sgc_outer
, '/'));
325 myp
= n_string_cp(sgcp
->sgc_outer
);
326 old_outerlen
= sgcp
->sgc_outer
->s_len
;
328 /* Separate current directory/pattern level from any possible remaining
329 * pattern in order to be able to use it for fnmatch(3) */
330 if((ccp
= memchr(sgcp
->sgc_patdat
, '/', sgcp
->sgc_patlen
)) == NULL
)
334 nsgc
.sgc_flags
|= a_DEEP
;
335 sgcp
->sgc_patlen
= PTR2SIZE((nsgc
.sgc_patdat
= &ccp
[1]) -
336 &sgcp
->sgc_patdat
[0]);
337 nsgc
.sgc_patlen
-= sgcp
->sgc_patlen
;
339 if(sgcp
->sgc_patlen
> 0){
340 assert(sgcp
->sgc_patdat
[sgcp
->sgc_patlen
-1] == '/');
341 ((char*)n_UNCONST(sgcp
->sgc_patdat
))[--sgcp
->sgc_patlen
] = '\0';
345 /* Our current directory level */
346 /* xxx Plenty of room for optimizations, like quickshot lstat(2) which may
347 * xxx be the (sole) result depending on pattern surroundings, etc. */
348 if((dp
= opendir(myp
)) == NULL
){
351 switch((err
= n_err_no
)){
353 ccp
= N_("cannot access paths under non-directory");
356 ccp
= N_("path component of (sub)pattern non-existent");
359 ccp
= N_("file permission for file (sub)pattern denied");
362 ccp
= N_("cannot handle file (sub)pattern");
367 /* As necessary, quote bytes in the current pattern */
373 for(need
= FAL0
, i
= 0, myp
= sgcp
->sgc_patdat
; *myp
!= '\0'; ++myp
)
375 case '\'': case '"': case '\\': case '$':
387 for(i
= 0, myp
= sgcp
->sgc_patdat
; *myp
!= '\0'; ++myp
)
389 case '\'': case '"': case '\\': case '$':
400 myp
= sgcp
->sgc_patdat
;
403 while((dep
= readdir(dp
)) != NULL
){
404 switch(fnmatch(myp
, dep
->d_name
, FNM_PATHNAME
| FNM_PERIOD
)){
406 /* A match expresses the desire to recurse if there is more pattern */
407 if(nsgc
.sgc_patlen
> 0){
410 n_string_push_cp((sgcp
->sgc_outer
->s_len
> 1
411 ? n_string_push_c(sgcp
->sgc_outer
, '/') : sgcp
->sgc_outer
),
415 #ifdef HAVE_DIRENT_TYPE
416 if(dep
->d_type
== DT_DIR
)
418 else if(dep
->d_type
== DT_LNK
|| dep
->d_type
== DT_UNKNOWN
)
423 if(stat(n_string_cp(sgcp
->sgc_outer
), &sb
)){
424 ccp
= N_("I/O error when querying file status");
426 }else if(S_ISDIR(sb
.st_mode
))
430 /* TODO We recurse with current dir FD open, which could E[MN]FILE!
431 * TODO Instead save away a list of such n_string's for later */
432 if(isdir
&& !a_shexp__glob(&nsgc
, slpp
)){
437 n_string_trunc(sgcp
->sgc_outer
, old_outerlen
);
439 struct n_strlist
*slp
;
442 i
= strlen(dep
->d_name
);
443 j
= (old_outerlen
> 0) ? old_outerlen
+ 1 + i
: i
;
444 slp
= n_STRLIST_ALLOC(j
);
446 slpp
= &slp
->sl_next
;
448 if((j
= old_outerlen
) > 0){
449 memcpy(&slp
->sl_dat
[0], sgcp
->sgc_outer
->s_dat
, j
);
450 if(slp
->sl_dat
[j
-1] != '/')
451 slp
->sl_dat
[j
++] = '/';
453 memcpy(&slp
->sl_dat
[j
], dep
->d_name
, i
);
454 slp
->sl_dat
[j
+= i
] = '\0';
461 ccp
= N_("fnmatch(3) cannot handle file (sub)pattern");
471 return (ccp
== NULL
);
474 if(!(sgcp
->sgc_flags
& a_SILENT
)){
477 if(sgcp
->sgc_outer
->s_len
> 0){
478 s2
= n_shexp_quote_cp(n_string_cp(sgcp
->sgc_outer
), FAL0
);
483 n_err("%s: %s%s%s\n", V_(ccp
), s2
, s3
,
484 n_shexp_quote_cp(sgcp
->sgc_patdat
, FAL0
));
490 a_shexp__globsort(void const *cvpa
, void const *cvpb
){
492 struct n_strlist
const * const *slpa
, * const *slpb
;
497 rv
= asccasecmp((*slpa
)->sl_dat
, (*slpb
)->sl_dat
);
501 #endif /* HAVE_FNMATCH */
504 a_shexp__quote(struct a_shexp_quote_ctx
*sqcp
, struct a_shexp_quote_lvl
*sqlp
){
505 /* XXX Because of the problems caused by ISO C multibyte interface we cannot
506 * XXX use the recursive implementation because of stateful encodings.
507 * XXX I.e., if a quoted substring cannot be self-contained - the data after
508 * XXX the quote relies on "the former state", then this doesn't make sense.
509 * XXX Therefore this is not fully programmed out but instead only detects
510 * XXX the "most fancy" quoting necessary, and directly does that.
511 * XXX As a result of this, T_REVSOL and T_DOUBLE are not even considered.
512 * XXX Otherwise we rather have to convert to wide first and act on that,
513 * XXX e.g., call visual_info(n_VISUAL_INFO_WOUT_CREATE) on entire input */
514 #undef a_SHEXP_QUOTE_RECURSE /* XXX (Needs complete revisit, then) */
515 #ifdef a_SHEXP_QUOTE_RECURSE
516 # define jrecurse jrecurse
517 struct a_shexp_quote_lvl sql
;
519 # define jrecurse jstep
521 struct n_visual_info_ctx vic
;
522 union {struct a_shexp_quote_lvl
*head
; struct n_string
*store
;} u
;
525 char const *ib
, *ib_base
;
528 ib_base
= ib
= sqlp
->sql_dat
.s
;
529 il
= sqlp
->sql_dat
.l
;
530 flags
= sqlp
->sql_flags
;
532 /* Iterate over the entire input, classify characters and type of quotes
533 * along the way. Whenever a quote change has to be applied, adjust flags
534 * for the new situation -, setup sql.* and recurse- */
540 if(flags
& a_SHEXP_QUOTE_T_DOLLAR
)
542 if(c
== '\t' && (flags
& (a_SHEXP_QUOTE_T_REVSOL
|
543 a_SHEXP_QUOTE_T_SINGLE
| a_SHEXP_QUOTE_T_DOUBLE
)))
545 #ifdef a_SHEXP_QUOTE_RECURSE
546 ++sqcp
->sqc_cnt_dollar
;
548 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_DOLLAR
;
550 }else if(blankspacechar(c
) || c
== '|' || c
== '&' || c
== ';' ||
551 /* Whereas we don't support those, quote them for the sh(1)ell */
552 c
== '(' || c
== ')' || c
== '<' || c
== '>' ||
553 c
== '"' || c
== '$'){
554 if(flags
& a_SHEXP_QUOTE_T_MASK
)
556 #ifdef a_SHEXP_QUOTE_RECURSE
557 ++sqcp
->sqc_cnt_single
;
559 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_SINGLE
;
562 if(flags
& (a_SHEXP_QUOTE_T_MASK
& ~a_SHEXP_QUOTE_T_SINGLE
))
564 #ifdef a_SHEXP_QUOTE_RECURSE
565 ++sqcp
->sqc_cnt_dollar
;
567 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_DOLLAR
;
569 }else if(c
== '\\' || (c
== '#' && ib
== ib_base
)){
570 if(flags
& a_SHEXP_QUOTE_T_MASK
)
572 #ifdef a_SHEXP_QUOTE_RECURSE
573 ++sqcp
->sqc_cnt_single
;
575 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_SINGLE
;
577 }else if(!asciichar(c
)){
578 /* Need to keep together multibytes */
579 #ifdef a_SHEXP_QUOTE_RECURSE
580 memset(&vic
, 0, sizeof vic
);
584 n_VISUAL_INFO_ONE_CHAR
| n_VISUAL_INFO_SKIP_ERRORS
);
586 /* xxx check whether resulting \u would be ASCII */
587 if(!(flags
& a_SHEXP_QUOTE_ROUNDTRIP
) ||
588 (flags
& a_SHEXP_QUOTE_T_DOLLAR
)){
589 #ifdef a_SHEXP_QUOTE_RECURSE
597 #ifdef a_SHEXP_QUOTE_RECURSE
598 ++sqcp
->sqc_cnt_dollar
;
600 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_DOLLAR
;
606 sqlp
->sql_flags
= flags
;
608 /* Level made the great and completed processing input. Reverse the list of
609 * levels, detect the "most fancy" quote type needed along this way */
610 /* XXX Due to restriction as above very crude */
611 for(flags
= 0, il
= 0, u
.head
= NULL
; sqlp
!= NULL
;){
612 struct a_shexp_quote_lvl
*tmp
;
614 tmp
= sqlp
->sql_link
;
615 sqlp
->sql_link
= u
.head
;
617 il
+= sqlp
->sql_dat
.l
;
618 if(sqlp
->sql_flags
& a_SHEXP_QUOTE_T_MASK
)
619 il
+= (sqlp
->sql_dat
.l
>> 1);
620 flags
|= sqlp
->sql_flags
;
625 /* Finally work the substrings in the correct order, adjusting quotes along
626 * the way as necessary. Start off with the "most fancy" quote, so that
627 * the user sees an overall boundary she can orientate herself on.
628 * We do it like that to be able to give the user some "encapsulation
629 * experience", to address what strikes me is a problem of sh(1)ell quoting:
630 * different to, e.g., perl(1), where you see at a glance where a string
631 * starts and ends, sh(1) quoting occurs at the "top level", disrupting the
632 * visual appearance of "a string" as such */
633 u
.store
= n_string_reserve(sqcp
->sqc_store
, il
);
635 if(flags
& a_SHEXP_QUOTE_T_DOLLAR
){
636 u
.store
= n_string_push_buf(u
.store
, "$'", sizeof("$'") -1);
637 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_DOLLAR
;
638 }else if(flags
& a_SHEXP_QUOTE_T_DOUBLE
){
639 u
.store
= n_string_push_c(u
.store
, '"');
640 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_DOUBLE
;
641 }else if(flags
& a_SHEXP_QUOTE_T_SINGLE
){
642 u
.store
= n_string_push_c(u
.store
, '\'');
643 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_SINGLE
;
644 }else /*if(flags & a_SHEXP_QUOTE_T_REVSOL)*/
645 flags
&= ~a_SHEXP_QUOTE_T_MASK
;
647 /* Work all the levels */
648 for(; sqlp
!= NULL
; sqlp
= sqlp
->sql_link
){
649 /* As necessary update our mode of quoting */
650 #ifdef a_SHEXP_QUOTE_RECURSE
653 switch(sqlp
->sql_flags
& a_SHEXP_QUOTE_T_MASK
){
654 case a_SHEXP_QUOTE_T_DOLLAR
:
655 if(!(flags
& a_SHEXP_QUOTE_T_DOLLAR
))
656 il
= a_SHEXP_QUOTE_T_DOLLAR
;
658 case a_SHEXP_QUOTE_T_DOUBLE
:
659 if(!(flags
& (a_SHEXP_QUOTE_T_DOUBLE
| a_SHEXP_QUOTE_T_DOLLAR
)))
660 il
= a_SHEXP_QUOTE_T_DOLLAR
;
662 case a_SHEXP_QUOTE_T_SINGLE
:
663 if(!(flags
& (a_SHEXP_QUOTE_T_REVSOL
| a_SHEXP_QUOTE_T_SINGLE
|
664 a_SHEXP_QUOTE_T_DOUBLE
| a_SHEXP_QUOTE_T_DOLLAR
)))
665 il
= a_SHEXP_QUOTE_T_SINGLE
;
668 case a_SHEXP_QUOTE_T_REVSOL
:
669 if(!(flags
& (a_SHEXP_QUOTE_T_REVSOL
| a_SHEXP_QUOTE_T_SINGLE
|
670 a_SHEXP_QUOTE_T_DOUBLE
| a_SHEXP_QUOTE_T_DOLLAR
)))
671 il
= a_SHEXP_QUOTE_T_REVSOL
;
676 if(flags
& (a_SHEXP_QUOTE_T_SINGLE
| a_SHEXP_QUOTE_T_DOLLAR
))
677 u
.store
= n_string_push_c(u
.store
, '\'');
678 else if(flags
& a_SHEXP_QUOTE_T_DOUBLE
)
679 u
.store
= n_string_push_c(u
.store
, '"');
680 flags
&= ~a_SHEXP_QUOTE_T_MASK
;
683 if(flags
& a_SHEXP_QUOTE_T_DOLLAR
)
684 u
.store
= n_string_push_buf(u
.store
, "$'", sizeof("$'") -1);
685 else if(flags
& a_SHEXP_QUOTE_T_DOUBLE
)
686 u
.store
= n_string_push_c(u
.store
, '"');
687 else if(flags
& a_SHEXP_QUOTE_T_SINGLE
)
688 u
.store
= n_string_push_c(u
.store
, '\'');
690 #endif /* a_SHEXP_QUOTE_RECURSE */
692 /* Work the level's substring */
693 ib
= sqlp
->sql_dat
.s
;
694 il
= sqlp
->sql_dat
.l
;
702 assert(c
== '\t' || (flags
& a_SHEXP_QUOTE_T_DOLLAR
));
703 assert((flags
& (a_SHEXP_QUOTE_T_REVSOL
| a_SHEXP_QUOTE_T_SINGLE
|
704 a_SHEXP_QUOTE_T_DOUBLE
| a_SHEXP_QUOTE_T_DOLLAR
)));
706 case 0x07: c
= 'a'; break;
707 case 0x08: c
= 'b'; break;
708 case 0x0A: c
= 'n'; break;
709 case 0x0B: c
= 'v'; break;
710 case 0x0C: c
= 'f'; break;
711 case 0x0D: c
= 'r'; break;
712 case 0x1B: c
= 'E'; break;
715 if(flags
& a_SHEXP_QUOTE_T_DOLLAR
){
719 if(flags
& a_SHEXP_QUOTE_T_REVSOL
)
720 u
.store
= n_string_push_c(u
.store
, '\\');
723 u
.store
= n_string_push_c(u
.store
, '\\');
725 u
.store
= n_string_push_c(u
.store
, 'c');
729 }else if(blankspacechar(c
) || c
== '|' || c
== '&' || c
== ';' ||
730 /* Whereas we don't support those, quote them for the sh(1)ell */
731 c
== '(' || c
== ')' || c
== '<' || c
== '>' ||
732 c
== '"' || c
== '$'){
733 if(flags
& (a_SHEXP_QUOTE_T_SINGLE
| a_SHEXP_QUOTE_T_DOLLAR
))
735 assert(flags
& (a_SHEXP_QUOTE_T_REVSOL
| a_SHEXP_QUOTE_T_DOUBLE
));
736 u
.store
= n_string_push_c(u
.store
, '\\');
739 if(flags
& a_SHEXP_QUOTE_T_DOUBLE
)
741 assert(!(flags
& a_SHEXP_QUOTE_T_SINGLE
));
742 u
.store
= n_string_push_c(u
.store
, '\\');
744 }else if(c
== '\\' || (c
== '#' && ib
== ib_base
)){
745 if(flags
& a_SHEXP_QUOTE_T_SINGLE
)
747 assert(flags
& (a_SHEXP_QUOTE_T_REVSOL
| a_SHEXP_QUOTE_T_DOUBLE
|
748 a_SHEXP_QUOTE_T_DOLLAR
));
749 u
.store
= n_string_push_c(u
.store
, '\\');
751 }else if(asciichar(c
)){
752 /* Shorthand: we can simply push that thing out */
754 u
.store
= n_string_push_c(u
.store
, c
);
757 /* Not an ASCII character, take care not to split up multibyte
758 * sequences etc. For the sake of compile testing, don't enwrap in
759 * HAVE_ALWAYS_UNICODE_LOCALE || HAVE_NATCH_CHAR */
760 if(n_psonce
& n_PSO_UNICODE
){
767 if((uc
= n_utf8_to_utf32(&ib2
, &il2
)) != UI32_MAX
){
771 il2
= PTR2SIZE(&ib2
[0] - &ib
[0]);
772 if((flags
& a_SHEXP_QUOTE_ROUNDTRIP
) || uc
== 0xFFFDu
){
773 /* Use padding to make ambiguities impossible */
774 il3
= snprintf(itoa
, sizeof itoa
, "\\%c%0*X",
775 (uc
> 0xFFFFu
? 'U' : 'u'),
776 (int)(uc
> 0xFFFFu
? 8 : 4), uc
);
782 u
.store
= n_string_push_buf(u
.store
, cp
, il3
);
783 ib
+= il2
, il
-= il2
;
788 memset(&vic
, 0, sizeof vic
);
792 n_VISUAL_INFO_ONE_CHAR
| n_VISUAL_INFO_SKIP_ERRORS
);
794 /* Work this substring as sensitive as possible */
796 if(!(flags
& a_SHEXP_QUOTE_ROUNDTRIP
))
797 u
.store
= n_string_push_buf(u
.store
, ib
, il
);
799 else if((vic
.vic_indat
= n_iconv_onetime_cp(n_ICONV_NONE
,
800 "utf-8", ok_vlook(ttycharset
), savestrbuf(ib
, il
))) != NULL
){
805 il2
= strlen(ib2
= vic
.vic_indat
);
806 if((uc
= n_utf8_to_utf32(&ib2
, &il2
)) != UI32_MAX
){
809 il2
= PTR2SIZE(&ib2
[0] - &vic
.vic_indat
[0]);
810 /* Use padding to make ambiguities impossible */
811 il3
= snprintf(itoa
, sizeof itoa
, "\\%c%0*X",
812 (uc
> 0xFFFFu
? 'U' : 'u'),
813 (int)(uc
> 0xFFFFu
? 8 : 4), uc
);
814 u
.store
= n_string_push_buf(u
.store
, itoa
, il3
);
824 u
.store
= n_string_push_buf(u
.store
, "\\xFF",
826 n_c_to_hex_base16(&u
.store
->s_dat
[u
.store
->s_len
- 2], *ib
++);
835 /* Close an open quote */
836 if(flags
& (a_SHEXP_QUOTE_T_SINGLE
| a_SHEXP_QUOTE_T_DOLLAR
))
837 u
.store
= n_string_push_c(u
.store
, '\'');
838 else if(flags
& a_SHEXP_QUOTE_T_DOUBLE
)
839 u
.store
= n_string_push_c(u
.store
, '"');
840 #ifdef a_SHEXP_QUOTE_RECURSE
846 #ifdef a_SHEXP_QUOTE_RECURSE
848 sqlp
->sql_dat
.l
-= il
;
851 sql
.sql_dat
.s
= n_UNCONST(ib
);
853 sql
.sql_flags
= flags
;
854 a_shexp__quote(sqcp
, &sql
);
859 #undef a_SHEXP_QUOTE_RECURSE
863 fexpand(char const *name
, enum fexp_mode fexpm
) /* TODO in parts: -> URL::!! */
866 char const *res
, *cp
;
867 bool_t dyn
, haveproto
;
870 n_pstate
&= ~n_PS_EXPAND_MULTIRESULT
;
873 /* The order of evaluation is "%" and "#" expand into constants.
874 * "&" can expand into "+". "+" can expand into shell meta characters.
875 * Shell meta characters expand into constants.
876 * This way, we make no recursive expansion */
877 if ((fexpm
& FEXP_NSHORTCUT
) || (res
= shortcut_expand(name
)) == NULL
)
878 res
= n_UNCONST(name
);
881 n_UNINIT(proto
.s
, NULL
), n_UNINIT(proto
.l
, 0);
883 for(cp
= res
; *cp
&& *cp
!= ':'; ++cp
)
886 if(cp
[0] == ':' && cp
[1] == '/' && cp
[2] == '/'){
888 proto
.s
= n_UNCONST(res
);
890 proto
.l
= PTR2SIZE(cp
- res
);
895 if(!(fexpm
& FEXP_NSPECIAL
)){
900 if(res
[1] == ':' && res
[2] != '\0'){
906 force
= (res
[1] != '\0');
907 res
= a_shexp_findmail((force
? &res
[1] : ok_vlook(LOGNAME
)),
916 if (prevfile
[0] == '\0') {
917 n_err(_("No previous file\n"));
925 res
= ok_vlook(MBOX
);
931 if (res
[0] == '@' && which_protocol(mailname
, FAL0
, FAL0
, NULL
)
933 res
= str_concat_csvl(&s
, protbase(mailname
), "/", &res
[1], NULL
)->s
;
938 /* POSIX: if *folder* unset or null, "+" shall be retained */
939 if (!(fexpm
& FEXP_NFOLDER
) && *res
== '+' &&
940 *(cp
= n_folder_query()) != '\0') {
941 res
= str_concat_csvl(&s
, cp
, &res
[1], NULL
)->s
;
945 /* Do some meta expansions */
946 if((fexpm
& (FEXP_NSHELL
| FEXP_NVAR
)) != FEXP_NVAR
&&
947 ((fexpm
& FEXP_NSHELL
) ? (strchr(res
, '$') != NULL
)
948 : anyof(res
, "{}[]*?$"))){
951 if(fexpm
& FEXP_NOPROTO
)
954 cp
= haveproto
? savecat(savestrbuf(proto
.s
, proto
.l
), res
) : res
;
956 switch(which_protocol(cp
, TRU1
, FAL0
, NULL
)){
969 struct n_string shou
, *shoup
;
971 shin
.s
= n_UNCONST(res
);
973 shoup
= n_string_creat_auto(&shou
);
975 enum n_shexp_state shs
;
977 /* TODO shexp: take care to not include backtick eval once avail! */
978 shs
= n_shexp_parse_token((n_SHEXP_PARSE_LOG_D_V
|
979 n_SHEXP_PARSE_QUOTE_AUTO_FIXED
| n_SHEXP_PARSE_QUOTE_AUTO_DQ
|
980 n_SHEXP_PARSE_QUOTE_AUTO_CLOSE
), shoup
, &shin
, NULL
);
981 if(shs
& n_SHEXP_STATE_STOP
)
984 res
= n_string_cp(shoup
);
985 /*shoup = n_string_drop_ownership(shoup);*/
989 res
= a_shexp_tilde(res
);
991 if(!(fexpm
& FEXP_NSHELL
) &&
992 (res
= a_shexp_globname(res
, fexpm
)) == NULL
)
996 }else if(res
[0] == '~'){
997 res
= a_shexp_tilde(res
);
1002 if(res
!= NULL
&& haveproto
){
1003 res
= savecat(savestrbuf(proto
.s
, proto
.l
), res
);
1007 if(fexpm
& FEXP_LOCAL
){
1008 switch (which_protocol(res
, FAL0
, FAL0
, NULL
)) {
1010 case PROTO_MAILDIR
: /* Cannot happen since we don't stat(2), but.. */
1013 n_err(_("Not a local file or directory: %s\n"),
1014 n_shexp_quote_cp(name
, FAL0
));
1021 if(res
!= NULL
&& !dyn
)
1024 return n_UNCONST(res
);
1027 FL
enum n_shexp_state
1028 n_shexp_parse_token(enum n_shexp_parse_flags flags
, struct n_string
*store
,
1029 struct str
*input
, void const **cookie
){
1030 /* TODO shexp_parse_token: WCHAR */
1031 ui32_t last_known_meta_trim_len
;
1032 char c2
, c
, quotec
, utf
[8];
1033 enum n_shexp_state rv
;
1035 char const *ifs
, *ifs_ws
, *ib_save
, *ib
;
1038 a_SKIPQ
= 1u<<0, /* Skip rest of this quote (\u0 ..) */
1039 a_SKIPT
= 1u<<1, /* Skip entire token (\c@) */
1040 a_SKIPMASK
= a_SKIPQ
| a_SKIPT
,
1041 a_SURPLUS
= 1u<<2, /* Extended sequence interpretation */
1042 a_NTOKEN
= 1u<<3, /* "New token": e.g., comments are possible */
1043 a_BRACE
= 1u<<4, /* Variable substitution: brace enclosed */
1044 a_DIGIT1
= 1u<<5, /* ..first character was digit */
1045 a_NONDIGIT
= 1u<<6, /* ..has seen any non-digits */
1046 a_VARSUBST_MASK
= n_BITENUM_MASK(4, 6),
1048 a_ROUND_MASK
= a_SKIPT
| (int)~n_BITENUM_MASK(0, 7),
1051 a_CONSUME
= 1u<<10, /* When done, "consume" remaining input */
1056 assert((flags
& n_SHEXP_PARSE_DRYRUN
) || store
!= NULL
);
1057 assert(input
!= NULL
);
1058 assert(input
->l
== 0 || input
->s
!= NULL
);
1059 assert(!(flags
& n_SHEXP_PARSE_LOG
) || !(flags
& n_SHEXP_PARSE_LOG_D_V
));
1060 assert(!(flags
& n_SHEXP_PARSE_IFS_ADD_COMMA
) ||
1061 !(flags
& n_SHEXP_PARSE_IFS_IS_COMMA
));
1062 assert(!(flags
& n_SHEXP_PARSE_QUOTE_AUTO_FIXED
) ||
1063 (flags
& n__SHEXP_PARSE_QUOTE_AUTO_MASK
));
1065 if((flags
& n_SHEXP_PARSE_LOG_D_V
) && (n_poption
& n_PO_D_V
))
1066 flags
|= n_SHEXP_PARSE_LOG
;
1067 if(flags
& n_SHEXP_PARSE_QUOTE_AUTO_FIXED
)
1068 flags
|= n_SHEXP_PARSE_QUOTE_AUTO_CLOSE
;
1070 if((flags
& n_SHEXP_PARSE_TRUNC
) && store
!= NULL
)
1071 store
= n_string_trunc(store
, 0);
1073 if(flags
& (n_SHEXP_PARSE_IFS_VAR
| n_SHEXP_PARSE_TRIM_IFSSPACE
)){
1074 ifs
= ok_vlook(ifs
);
1075 ifs_ws
= ok_vlook(ifs_ws
);
1080 if((il
= input
->l
) == UIZ_MAX
)
1081 input
->l
= il
= strlen(ib
);
1084 if(cookie
!= NULL
&& *cookie
!= NULL
){
1085 assert(!(flags
& n_SHEXP_PARSE_DRYRUN
));
1089 rv
= n_SHEXP_STATE_NONE
;
1091 rv
&= n_SHEXP_STATE_WS_LEAD
;
1092 state
&= a_ROUND_MASK
;
1094 /* In cookie mode, the next ARGV entry is the token already, unchanged,
1095 * since it has already been expanded before! */
1096 if(state
& a_COOKIE
){
1097 char const * const *xcookie
, *cp
;
1101 if((store
= n_string_push_cp(store
, *xcookie
))->s_len
> 0)
1102 rv
|= n_SHEXP_STATE_OUTPUT
;
1103 if(*++xcookie
== NULL
){
1106 flags
|= n_SHEXP_PARSE_QUOTE_AUTO_DQ
; /* ..why we are here! */
1108 *cookie
= n_UNCONST(xcookie
);
1110 for(cp
= &n_string_cp(store
)[i
]; (c
= *cp
++) != '\0';)
1112 rv
|= n_SHEXP_STATE_CONTROL
;
1116 /* The last exploded cookie will join with the yielded input token, so
1117 * simply fall through in this case */
1118 if(state
& a_COOKIE
)
1122 if(flags
& n_SHEXP_PARSE_TRIM_SPACE
){
1123 for(; il
> 0; ++ib
, --il
){
1124 if(!blankspacechar(*ib
))
1126 rv
|= n_SHEXP_STATE_WS_LEAD
;
1130 if(flags
& n_SHEXP_PARSE_TRIM_IFSSPACE
){
1131 for(; il
> 0; ++ib
, --il
){
1132 if(strchr(ifs_ws
, *ib
) == NULL
)
1134 rv
|= n_SHEXP_STATE_WS_LEAD
;
1138 input
->s
= n_UNCONST(ib
);
1143 rv
|= n_SHEXP_STATE_STOP
;
1148 store
= n_string_reserve(store
, n_MIN(il
, 32)); /* XXX */
1150 switch(flags
& n__SHEXP_PARSE_QUOTE_AUTO_MASK
){
1151 case n_SHEXP_PARSE_QUOTE_AUTO_SQ
:
1153 rv
|= n_SHEXP_STATE_QUOTE
;
1155 case n_SHEXP_PARSE_QUOTE_AUTO_DQ
:
1158 case n_SHEXP_PARSE_QUOTE_AUTO_DSQ
:
1161 rv
|= n_SHEXP_STATE_QUOTE
;
1170 /* TODO n_SHEXP_PARSE_META_SEMICOLON++, well, hack: we are not the shell,
1171 * TODO we are not a language, and therefore the general *ifs-ws* and normal
1172 * TODO whitespace trimming that input lines undergo (in a_go_evaluate())
1173 * TODO has already happened, our result will be used *as is*, and therefore
1174 * TODO we need to be aware of and remove trailing unquoted WS that would
1175 * TODO otherwise remain, after we have seen a semicolon sequencer.
1176 * By sheer luck we only need to track this in non-quote-mode */
1177 last_known_meta_trim_len
= UI32_MAX
;
1182 /* If no quote-mode active.. */
1184 if(c
== '"' || c
== '\''){
1189 state
&= ~a_SURPLUS
;
1191 last_known_meta_trim_len
= UI32_MAX
;
1192 rv
|= n_SHEXP_STATE_QUOTE
;
1197 last_known_meta_trim_len
= UI32_MAX
;
1202 rv
|= n_SHEXP_STATE_QUOTE
;
1207 }else if(c
== '\\'){
1208 /* Outside of quotes this just escapes any next character, but a sole
1209 * <reverse solidus> at EOS is left unchanged */
1213 last_known_meta_trim_len
= UI32_MAX
;
1215 /* A comment may it be if no token has yet started */
1216 else if(c
== '#' && (state
& a_NTOKEN
)){
1217 rv
|= n_SHEXP_STATE_STOP
;
1218 /*last_known_meta_trim_len = UI32_MAX;*/
1221 /* Metacharacters which separate tokens must be turned on explicitly */
1223 rv
|= n_SHEXP_STATE_META_VERTBAR
;
1224 /* The parsed sequence may be _the_ output, so ensure we don't
1225 * include the metacharacter, then. */
1226 /*if(flags & n_SHEXP_PARSE_META_VERTBAR) TODO trail ws strip then */
1227 if(flags
& (n_SHEXP_PARSE_DRYRUN
| n_SHEXP_PARSE_META_KEEP
))
1229 /*last_known_meta_trim_len = UI32_MAX;*/
1232 rv
|= n_SHEXP_STATE_META_AMPERSAND
;
1233 /* The parsed sequence may be _the_ output, so ensure we don't
1234 * include the metacharacter, then. */
1235 /*if(flags & n_SHEXP_PARSE_META_AMPERSAND)TODO trail ws strip then*/
1236 if(flags
& (n_SHEXP_PARSE_DRYRUN
| n_SHEXP_PARSE_META_KEEP
))
1238 /*last_known_meta_trim_len = UI32_MAX;*/
1241 rv
|= n_SHEXP_STATE_META_SEMICOLON
;
1242 if(flags
& n_SHEXP_PARSE_META_SEMICOLON
){
1244 n_go_input_inject(n_GO_INPUT_INJECT_COMMIT
, ib
, il
);
1246 rv
|= n_SHEXP_STATE_STOP
;
1247 if(!(flags
& n_SHEXP_PARSE_DRYRUN
) &&
1248 (rv
& n_SHEXP_STATE_OUTPUT
) &&
1249 last_known_meta_trim_len
!= UI32_MAX
)
1250 store
= n_string_trunc(store
, last_known_meta_trim_len
);
1252 /* The parsed sequence may be _the_ output, so ensure we don't
1253 * include the metacharacter, then. */
1254 if(flags
& (n_SHEXP_PARSE_DRYRUN
| n_SHEXP_PARSE_META_KEEP
))
1256 /*last_known_meta_trim_len = UI32_MAX;*/
1258 }else if(c
== ',' && (flags
&
1259 (n_SHEXP_PARSE_IFS_ADD_COMMA
| n_SHEXP_PARSE_IFS_IS_COMMA
))){
1260 /* The parsed sequence may be _the_ output, so ensure we don't
1261 * include the metacharacter, then. */
1262 if(flags
& (n_SHEXP_PARSE_DRYRUN
| n_SHEXP_PARSE_META_KEEP
))
1264 /*last_known_meta_trim_len = UI32_MAX;*/
1269 blnk
= blankchar(c
) ? 1 : 0;
1270 blnk
|= ((flags
& (n_SHEXP_PARSE_IFS_VAR
|
1271 n_SHEXP_PARSE_TRIM_IFSSPACE
)) &&
1272 strchr(ifs_ws
, c
) != NULL
) ? 2 : 0;
1274 if((!(flags
& n_SHEXP_PARSE_IFS_VAR
) && (blnk
& 1)) ||
1275 ((flags
& n_SHEXP_PARSE_IFS_VAR
) &&
1276 ((blnk
& 2) || strchr(ifs
, c
) != NULL
))){
1277 if(!(flags
& n_SHEXP_PARSE_IFS_IS_COMMA
)){
1278 /* The parsed sequence may be _the_ output, so ensure we don't
1279 * include the metacharacter, then. */
1280 if(flags
& (n_SHEXP_PARSE_DRYRUN
| n_SHEXP_PARSE_META_KEEP
))
1282 /*last_known_meta_trim_len = UI32_MAX;*/
1289 if(blnk
&& store
!= NULL
){
1290 if(last_known_meta_trim_len
== UI32_MAX
)
1291 last_known_meta_trim_len
= store
->s_len
;
1293 last_known_meta_trim_len
= UI32_MAX
;
1297 assert(!(state
& a_NTOKEN
));
1298 if(c
== quotec
&& !(flags
& n_SHEXP_PARSE_QUOTE_AUTO_FIXED
)){
1299 state
&= a_ROUND_MASK
;
1301 /* Users may need to recognize the presence of empty quotes */
1302 rv
|= n_SHEXP_STATE_OUTPUT
;
1304 }else if(c
== '\\' && (state
& a_SURPLUS
)){
1306 /* A sole <reverse solidus> at EOS is treated as-is! This is ok
1307 * since the "closing quote" error will occur next, anyway */
1310 else if((c2
= *ib
) == quotec
){
1313 }else if(quotec
== '"'){
1314 /* Double quotes, POSIX says:
1315 * The <backslash> shall retain its special meaning as an
1316 * escape character (see Section 2.2.1) only when followed
1317 * by one of the following characters when considered
1318 * special: $ ` " \ <newline> */
1322 /* case '"': already handled via c2 == quotec */
1331 /* Dollar-single-quote */
1335 /* case '\'': already handled via c2 == quotec */
1340 case 'b': c
= '\b'; break;
1341 case 'f': c
= '\f'; break;
1342 case 'n': c
= '\n'; break;
1343 case 'r': c
= '\r'; break;
1344 case 't': c
= '\t'; break;
1345 case 'v': c
= '\v'; break;
1348 case 'e': c
= '\033'; break;
1350 /* Control character */
1353 goto j_dollar_ungetc
;
1355 if(state
& a_SKIPMASK
)
1357 c
= upperconv(c2
) ^ 0x40;
1358 if((ui8_t
)c
> 0x1F && c
!= 0x7F){ /* ASCII C0: 0..1F, 7F */
1359 if(flags
& n_SHEXP_PARSE_LOG
)
1360 n_err(_("Invalid \\c notation: %.*s\n"),
1361 (int)input
->l
, input
->s
);
1362 rv
|= n_SHEXP_STATE_ERR_CONTROL
;
1364 /* As an implementation-defined extension, support \c@
1365 * EQ printf(1) alike \c */
1372 /* Octal sequence: 1 to 3 octal bytes */
1374 /* As an extension (dependent on where you look, echo(1), or
1375 * awk(1)/tr(1) etc.), allow leading "0" octal indicator */
1376 if(il
> 0 && (c
= *ib
) >= '0' && c
<= '7'){
1381 case '1': case '2': case '3':
1382 case '4': case '5': case '6': case '7':
1384 if(il
> 0 && (c
= *ib
) >= '0' && c
<= '7'){
1385 c2
= (c2
<< 3) | (c
- '0');
1388 if(il
> 0 && (c
= *ib
) >= '0' && c
<= '7'){
1389 if(!(state
& a_SKIPMASK
) && (ui8_t
)c2
> 0x1F){
1390 if(flags
& n_SHEXP_PARSE_LOG
)
1391 n_err(_("\\0 argument exceeds a byte: %.*s\n"),
1392 (int)input
->l
, input
->s
);
1393 rv
|= n_SHEXP_STATE_ERR_NUMBER
;
1395 /* Write unchanged */
1397 rv
|= n_SHEXP_STATE_OUTPUT
;
1398 if(!(flags
& n_SHEXP_PARSE_DRYRUN
))
1399 store
= n_string_push_buf(store
, ib_save
,
1400 PTR2SIZE(ib
- ib_save
));
1403 c2
= (c2
<< 3) | (c
-= '0');
1406 if(state
& a_SKIPMASK
)
1408 if((c
= c2
) == '\0'){
1414 /* ISO 10646 / Unicode sequence, 8 or 4 hexadecimal bytes */
1423 goto j_dollar_ungetc
;
1427 /* Hexadecimal sequence, 1 or 2 hexadecimal bytes */
1431 goto j_dollar_ungetc
;
1435 static ui8_t
const hexatoi
[] = { /* XXX uses ASCII */
1436 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15
1441 for(no
= j
= 0; i
-- > 0; --il
, ++ib
, ++j
){
1445 no
+= hexatoi
[(ui8_t
)((c
) - ((c
) <= '9' ? 48
1446 : ((c
) <= 'F' ? 55 : 87)))];
1448 if(state
& a_SKIPMASK
)
1450 c2
= (c2
== 'U' || c2
== 'u') ? 'u' : 'x';
1451 if(flags
& n_SHEXP_PARSE_LOG
)
1452 n_err(_("Invalid \\%c notation: %.*s\n"),
1453 c2
, (int)input
->l
, input
->s
);
1454 rv
|= n_SHEXP_STATE_ERR_NUMBER
;
1460 /* Unicode massage */
1461 if((c2
!= 'U' && c2
!= 'u') || n_uasciichar(no
)){
1462 if((c
= (char)no
) == '\0')
1466 else if(!(state
& a_SKIPMASK
)){
1467 if(!(flags
& n_SHEXP_PARSE_DRYRUN
))
1468 store
= n_string_reserve(store
, n_MAX(j
, 4));
1470 if(no
> 0x10FFFF){ /* XXX magic; CText */
1471 if(flags
& n_SHEXP_PARSE_LOG
)
1472 n_err(_("\\U argument exceeds 0x10FFFF: %.*s\n"),
1473 (int)input
->l
, input
->s
);
1474 rv
|= n_SHEXP_STATE_ERR_NUMBER
;
1475 /* But normalize the output anyway */
1479 j
= n_utf32_to_utf8(no
, utf
);
1481 if(n_psonce
& n_PSO_UNICODE
){
1482 rv
|= n_SHEXP_STATE_OUTPUT
| n_SHEXP_STATE_UNICODE
;
1483 if(!(flags
& n_SHEXP_PARSE_DRYRUN
))
1484 store
= n_string_push_buf(store
, utf
, j
);
1491 icp
= n_iconv_onetime_cp(n_ICONV_NONE
,
1494 rv
|= n_SHEXP_STATE_OUTPUT
;
1495 if(!(flags
& n_SHEXP_PARSE_DRYRUN
))
1496 store
= n_string_push_cp(store
, icp
);
1501 if(!(flags
& n_SHEXP_PARSE_DRYRUN
)) Je_uni_norm
:{
1504 rv
|= n_SHEXP_STATE_OUTPUT
|
1505 n_SHEXP_STATE_ERR_UNICODE
;
1506 i
= snprintf(itoa
, sizeof itoa
, "\\%c%0*X",
1507 (no
> 0xFFFFu
? 'U' : 'u'),
1508 (int)(no
> 0xFFFFu
? 8 : 4), (ui32_t
)no
);
1509 store
= n_string_push_buf(store
, itoa
, i
);
1513 if(state
& a_SKIPMASK
)
1518 /* Extension: \$ can be used to expand a variable.
1519 * Bug|ad effect: if conversion fails, not written "as-is" */
1522 goto j_dollar_ungetc
;
1527 /* Follow bash(1) behaviour, print sequence unchanged */
1532 }else if(c
== '$' && quotec
== '"' && il
> 0) J_var_expand
:{
1533 state
&= ~a_VARSUBST_MASK
;
1537 if(!(state
& a_BRACE
) || il
> 1){
1538 char const *cp
, *vp
;
1544 state
&= ~a_EXPLODE
;
1546 for(i
= 0; il
> 0; --il
, ++ib
, ++i
){
1547 /* We have some special cases regarding special parameters,
1548 * so ensure these don't cause failure. This code has
1549 * counterparts in code that manages internal variables! */
1551 if(!a_SHEXP_ISVARC(c
)){
1552 if(i
== 0 && (c
== '*' || c
== '@' || c
== '#' ||
1553 c
== '?' || c
== '!' || c
== '^')){
1554 /* Skip over multiplexer */
1565 }else if(a_SHEXP_ISVARC_BAD1ST(c
)){
1569 state
|= a_NONDIGIT
;
1572 if(state
& a_SKIPMASK
){
1573 if((state
& a_BRACE
) && il
> 0 && *ib
== '}')
1578 if((state
& (a_DIGIT1
| a_NONDIGIT
)) == (a_DIGIT1
| a_NONDIGIT
)){
1579 if(state
& a_BRACE
){
1580 if(il
> 0 && *ib
== '}')
1583 rv
|= n_SHEXP_STATE_ERR_BRACE
;
1585 if(flags
& n_SHEXP_PARSE_LOG
)
1586 n_err(_("Invalid identifier for ${}: %.*s\n"),
1587 (int)input
->l
, input
->s
);
1588 rv
|= n_SHEXP_STATE_ERR_IDENTIFIER
;
1591 if(state
& a_BRACE
){
1592 if(flags
& n_SHEXP_PARSE_LOG
)
1593 n_err(_("Bad substitution for ${}: %.*s\n"),
1594 (int)input
->l
, input
->s
);
1595 rv
|= n_SHEXP_STATE_ERR_BADSUB
;
1596 if(il
> 0 && *ib
== '}')
1599 rv
|= n_SHEXP_STATE_ERR_BRACE
;
1604 if(state
& a_BRACE
){
1605 if(il
== 0 || *ib
!= '}'){
1606 if(flags
& n_SHEXP_PARSE_LOG
)
1607 n_err(_("No closing brace for ${}: %.*s\n"),
1608 (int)input
->l
, input
->s
);
1609 rv
|= n_SHEXP_STATE_ERR_QUOTEOPEN
|
1610 n_SHEXP_STATE_ERR_BRACE
;
1616 if(flags
& n_SHEXP_PARSE_DRYRUN
)
1619 /* We may shall explode "${@}" to a series of successive,
1620 * properly quoted tokens (instead). The first exploded
1621 * cookie will join with the current token */
1622 if((state
& a_EXPLODE
) && !(flags
& n_SHEXP_PARSE_DRYRUN
) &&
1623 cookie
!= NULL
&& n_var_vexplode(cookie
)){
1625 input
->s
= n_UNCONST(ib
);
1627 goto jrestart_empty
;
1630 /* Check getenv(3) shall no internal variable exist!
1631 * XXX We have some common idioms, avoid memory for them
1632 * XXX Even better would be var_vlook_buf()! */
1635 case '?': vp
= n_qm
; break;
1636 case '!': vp
= n_em
; break;
1637 case '*': vp
= n_star
; break;
1638 case '@': vp
= n_at
; break;
1639 case '#': vp
= n_ns
; break;
1640 default: goto j_var_look_buf
;
1644 vp
= savestrbuf(vp
, i
);
1646 if((cp
= n_var_vlook(vp
, TRU1
)) != NULL
){
1647 rv
|= n_SHEXP_STATE_OUTPUT
;
1648 store
= n_string_push_cp(store
, cp
);
1649 for(; (c
= *cp
) != '\0'; ++cp
)
1651 rv
|= n_SHEXP_STATE_CONTROL
;
1658 }else if(c
== '`' && quotec
== '"' && il
> 0){ /* TODO shell command */
1663 if(!(state
& a_SKIPMASK
)){
1664 rv
|= n_SHEXP_STATE_OUTPUT
;
1666 rv
|= n_SHEXP_STATE_CONTROL
;
1667 if(!(flags
& n_SHEXP_PARSE_DRYRUN
))
1668 store
= n_string_push_c(store
, c
);
1672 if(quotec
!= '\0' && !(flags
& n_SHEXP_PARSE_QUOTE_AUTO_CLOSE
)){
1673 if(flags
& n_SHEXP_PARSE_LOG
)
1674 n_err(_("No closing quote: %.*s\n"), (int)input
->l
, input
->s
);
1675 rv
|= n_SHEXP_STATE_ERR_QUOTEOPEN
;
1679 assert(!(state
& a_COOKIE
));
1680 if((flags
& n_SHEXP_PARSE_DRYRUN
) && store
!= NULL
){
1681 store
= n_string_push_buf(store
, input
->s
, PTR2SIZE(ib
- input
->s
));
1682 rv
|= n_SHEXP_STATE_OUTPUT
;
1685 if(state
& a_CONSUME
){
1686 input
->s
= n_UNCONST(&ib
[il
]);
1689 if(flags
& n_SHEXP_PARSE_TRIM_SPACE
){
1690 for(; il
> 0; ++ib
, --il
){
1691 if(!blankspacechar(*ib
))
1693 rv
|= n_SHEXP_STATE_WS_TRAIL
;
1697 if(flags
& n_SHEXP_PARSE_TRIM_IFSSPACE
){
1698 for(; il
> 0; ++ib
, --il
){
1699 if(strchr(ifs_ws
, *ib
) == NULL
)
1701 rv
|= n_SHEXP_STATE_WS_TRAIL
;
1706 input
->s
= n_UNCONST(ib
);
1709 if(!(rv
& n_SHEXP_STATE_STOP
)){
1710 if(!(rv
& n_SHEXP_STATE_OUTPUT
) && (flags
& n_SHEXP_PARSE_IGNORE_EMPTY
) &&
1712 goto jrestart_empty
;
1713 if(/*!(rv & n_SHEXP_STATE_OUTPUT) &&*/ il
== 0)
1714 rv
|= n_SHEXP_STATE_STOP
;
1717 if((state
& a_SKIPT
) && !(rv
& n_SHEXP_STATE_STOP
) &&
1718 (flags
& n__SHEXP_PARSE_META_MASK
))
1721 assert((rv
& n_SHEXP_STATE_OUTPUT
) || !(rv
& n_SHEXP_STATE_UNICODE
));
1722 assert((rv
& n_SHEXP_STATE_OUTPUT
) || !(rv
& n_SHEXP_STATE_CONTROL
));
1728 n_shexp_parse_token_cp(enum n_shexp_parse_flags flags
, char const **cp
){
1730 struct n_string sou
, *soup
;
1732 enum n_shexp_state shs
;
1737 input
.s
= n_UNCONST(*cp
);
1739 soup
= n_string_creat_auto(&sou
);
1741 shs
= n_shexp_parse_token(flags
, soup
, &input
, NULL
);
1742 if(shs
& n_SHEXP_STATE_ERR_MASK
){
1743 soup
= n_string_assign_cp(soup
, *cp
);
1748 rv
= n_string_cp(soup
);
1749 /*n_string_gut(n_string_drop_ownership(soup));*/
1754 FL
struct n_string
*
1755 n_shexp_quote(struct n_string
*store
, struct str
const *input
, bool_t rndtrip
){
1756 struct a_shexp_quote_lvl sql
;
1757 struct a_shexp_quote_ctx sqc
;
1760 assert(store
!= NULL
);
1761 assert(input
!= NULL
);
1762 assert(input
->l
== 0 || input
->s
!= NULL
);
1764 memset(&sqc
, 0, sizeof sqc
);
1765 sqc
.sqc_store
= store
;
1766 sqc
.sqc_input
.s
= input
->s
;
1767 if((sqc
.sqc_input
.l
= input
->l
) == UIZ_MAX
)
1768 sqc
.sqc_input
.l
= strlen(input
->s
);
1769 sqc
.sqc_flags
= rndtrip
? a_SHEXP_QUOTE_ROUNDTRIP
: a_SHEXP_QUOTE_NONE
;
1771 if(sqc
.sqc_input
.l
== 0)
1772 store
= n_string_push_buf(store
, "''", sizeof("''") -1);
1774 memset(&sql
, 0, sizeof sql
);
1775 sql
.sql_dat
= sqc
.sqc_input
;
1776 sql
.sql_flags
= sqc
.sqc_flags
;
1777 a_shexp__quote(&sqc
, &sql
);
1784 n_shexp_quote_cp(char const *cp
, bool_t rndtrip
){
1785 struct n_string store
;
1792 input
.s
= n_UNCONST(cp
);
1794 rv
= n_string_cp(n_shexp_quote(n_string_creat_auto(&store
), &input
,
1796 n_string_gut(n_string_drop_ownership(&store
));
1802 n_shexp_is_valid_varname(char const *name
){
1809 for(lc
= '\0'; (c
= *name
++) != '\0'; lc
= c
)
1810 if(!a_SHEXP_ISVARC(c
))
1812 else if(lc
== '\0' && a_SHEXP_ISVARC_BAD1ST(c
))
1814 if(a_SHEXP_ISVARC_BADNST(lc
))
1824 c_shcodec(void *vp
){
1826 struct n_string sou_b
, *soup
;
1830 char const **argv
, *varname
, *act
, *cp
;
1832 soup
= n_string_creat_auto(&sou_b
);
1834 varname
= (n_pstate
& n_PS_ARGMOD_VPUT
) ? *argv
++ : NULL
;
1837 for(cp
= act
; *cp
!= '\0' && !blankspacechar(*cp
); ++cp
)
1839 if((norndtrip
= (*act
== '+')))
1843 alen
= PTR2SIZE(cp
- act
);
1847 in
.l
= strlen(in
.s
= n_UNCONST(cp
));
1850 if(is_ascncaseprefix(act
, "encode", alen
))
1851 soup
= n_shexp_quote(soup
, &in
, !norndtrip
);
1852 else if(!norndtrip
&& is_ascncaseprefix(act
, "decode", alen
)){
1854 enum n_shexp_state shs
;
1856 shs
= n_shexp_parse_token((n_SHEXP_PARSE_LOG
|
1857 n_SHEXP_PARSE_IGNORE_EMPTY
), soup
, &in
, NULL
);
1858 if(shs
& n_SHEXP_STATE_ERR_MASK
){
1859 soup
= n_string_assign_cp(soup
, cp
);
1860 nerrn
= n_ERR_CANCELED
;
1864 if(shs
& n_SHEXP_STATE_STOP
)
1870 if(varname
!= NULL
){
1871 cp
= n_string_cp(soup
);
1872 if(!n_var_vset(varname
, (uintptr_t)cp
)){
1873 nerrn
= n_ERR_NOTSUP
;
1879 in
.s
= n_string_cp(soup
);
1881 makeprint(&in
, &out
);
1882 if(fprintf(n_stdout
, "%s\n", out
.s
) < 0){
1890 n_pstate_err_no
= nerrn
;
1892 return (vp
!= NULL
? 0 : 1);
1894 n_err(_("Synopsis: shcodec: <[+]e[ncode]|d[ecode]> <rest-of-line>\n"));
1895 nerrn
= n_ERR_INVAL
;