* PATCH: fix for (#s) and (#e) in param substs
@ 2000-04-14 11:27 Peter Stephenson
2000-04-14 11:42 ` Andrej Borsenkow
2000-04-14 12:00 ` Zefram
0 siblings, 2 replies; 5+ messages in thread
From: Peter Stephenson @ 2000-04-14 11:27 UTC (permalink / raw)
To: Zsh hackers list
This wasn't as bad as it could have been. Things like
${foo//(#s)d*time(#e)/d fight/}
are all now supposed to work. One thing I like about this is it allows you
to forget about not only anchor characters in / and //, but even ## and %%,
and with the (S) parameter flag # and % too --- the /-forms together with
the start and end assertions present a single consistent interface for
pattern substitutions.
It's amazing what I can get excited about nowadays.
By the way, I inserted a hostage to fortune in the manual by claiming that
`(^(#s))' etc. also work, i.e. you can assert that you are not at the start
of the string. I believe this to be the case, but it is a good deal more
hairy as it relies on the exclusion code, which is a bit stomach-turning.
I've caught another potential bug that exclusions in groups pretend the
string ends at the end of the group, which could have made (#e) match
erroneously. I don't claim that's the end of the matter.
Index: Doc/Zsh/expn.yo
===================================================================
RCS file: /cvsroot/zsh/zsh/Doc/Zsh/expn.yo,v
retrieving revision 1.4
diff -u -r1.4 expn.yo
--- Doc/Zsh/expn.yo 2000/04/06 18:44:01 1.4
+++ Doc/Zsh/expn.yo 2000/04/14 11:26:05
@@ -1309,6 +1309,16 @@
`tt(*((#s)|/)test((#e)|/)*)' matches a path segment `tt(test)' in any of
the following strings: tt(test), tt(test/at/start), tt(at/end/test),
tt(in/test/middle).
+
+Another use is in parameter substitution; for example
+`tt(${array/(#s)A*Z(#e)})' will remove only elements of an array which
+match the complete pattern `tt(A*Z)'. There are other ways of performing
+many operations of this type, however the combination of the substitution
+operations `tt(/)' and `tt(//)' with the `tt((#s))' and `tt((#e))' flags
+provides a single simple and memorable method.
+
+Note that assertions of the form `tt((^(#s)))' also work, i.e. match
+anywhere except at the start of the string.
)
enditem()
Index: Src/glob.c
===================================================================
RCS file: /cvsroot/zsh/zsh/Src/glob.c,v
retrieving revision 1.3
diff -u -r1.3 glob.c
--- Src/glob.c 2000/04/07 02:27:44 1.3
+++ Src/glob.c 2000/04/14 11:26:05
@@ -2056,6 +2056,39 @@
}
/**/
+static void
+set_pat_start(Patprog p, int offs)
+{
+ /*
+ * If we are messing around with the test string by advancing up
+ * it from the start, we need to tell the pattern matcher that
+ * a start-of-string assertion, i.e. (#s), should fail. Hence
+ * we test whether the offset of the real start of string from
+ * the actual start, passed as offs, is zero.
+ */
+ if (offs)
+ p->flags |= PAT_NOTSTART;
+ else
+ p->flags &= ~PAT_NOTSTART;
+}
+
+/**/
+static void
+set_pat_end(Patprog p, char null_me)
+{
+ /*
+ * If we are messing around with the string by shortening it at the
+ * tail, we need to tell the pattern matcher that an end-of-string
+ * assertion, i.e. (#e), should fail. Hence we test whether
+ * the character null_me about to be zapped is or is not already a null.
+ */
+ if (null_me)
+ p->flags |= PAT_NOTEND;
+ else
+ p->flags &= ~PAT_NOTEND;
+}
+
+/**/
static int
igetmatch(char **sp, Patprog p, int fl, int n, char *replstr)
{
@@ -2068,6 +2101,9 @@
if (p->mustoff && !strstr((char *)s, (char *)p + p->mustoff))
matched = 0;
+ /* in case we used the prog before... */
+ p->flags &= ~(PAT_NOTSTART|PAT_NOTEND);
+
if (fl & SUB_ALL) {
i = matched && pattry(p, s);
*sp = get_match_ret(*sp, 0, i ? l : 0, fl, i ? replstr : 0);
@@ -2092,6 +2128,7 @@
*/
for (t = s; t < mpos; METAINC(t)) {
sav = *t;
+ set_pat_end(p, sav);
*t = '\0';
if (pattry(p, s)) {
mpos = patinput;
@@ -2112,6 +2149,7 @@
* There's no optimization here. */
patoffset = ml;
for (t = s + l; t >= s; t--, patoffset--) {
+ set_pat_start(p, t-s);
if (pattry(p, t)) {
*sp = get_match_ret(*sp, t - s, l, fl, replstr);
patoffset = 0;
@@ -2128,6 +2166,7 @@
* move forward along string until we get a match. *
* Again there's no optimisation. */
for (i = 0, t = s; i < l; i++, t++, patoffset++) {
+ set_pat_start(p, t-s);
if (pattry(p, t)) {
*sp = get_match_ret(*sp, i, l, fl, replstr);
patoffset = 0;
@@ -2141,6 +2180,7 @@
case SUB_SUBSTR:
/* Smallest at start, but matching substrings. */
+ set_pat_start(p, l);
if (!(fl & SUB_GLOBAL) && pattry(p, s + l) && !--n) {
*sp = get_match_ret(*sp, 0, 0, fl, replstr);
return 1;
@@ -2155,12 +2195,14 @@
matched = 0;
for (; t < s + l; t++, patoffset++) {
/* Find the longest match from this position. */
+ set_pat_start(p, t-s);
if (pattry(p, t) && patinput > t) {
char *mpos = patinput;
if (!(fl & SUB_LONG) && !(p->flags & PAT_PURES)) {
char *ptr;
for (ptr = t; ptr < mpos; METAINC(ptr)) {
sav = *ptr;
+ set_pat_end(p, sav);
*ptr = '\0';
if (pattry(p, t)) {
mpos = patinput;
@@ -2209,6 +2251,7 @@
* at the start. Goodness knows if this is a good idea
* with global substitution, so it doesn't happen.
*/
+ set_pat_start(p, l);
if ((fl & (SUB_LONG|SUB_GLOBAL)) == SUB_LONG &&
pattry(p, s + l) && !--n) {
*sp = get_match_ret(*sp, 0, 0, fl, replstr);
@@ -2219,6 +2262,7 @@
case (SUB_END|SUB_SUBSTR):
/* Shortest at end with substrings */
patoffset = ml;
+ set_pat_start(p, l);
if (pattry(p, s + l) && !--n) {
*sp = get_match_ret(*sp, l, l, fl, replstr);
patoffset = 0;
@@ -2230,6 +2274,7 @@
for (t = s + l - 1; t >= s; t--, patoffset--) {
if (t > s && t[-1] == Meta)
t--;
+ set_pat_start(p, t-s);
if (pattry(p, t) && patinput > t && !--n) {
/* Found the longest match */
char *mpos = patinput;
@@ -2237,6 +2282,7 @@
char *ptr;
for (ptr = t; ptr < mpos; METAINC(ptr)) {
sav = *ptr;
+ set_pat_end(p, sav);
*ptr = '\0';
if (pattry(p, t)) {
mpos = patinput;
@@ -2252,6 +2298,7 @@
}
}
patoffset = ml;
+ set_pat_start(p, l);
if ((fl & SUB_LONG) && pattry(p, s + l) && !--n) {
*sp = get_match_ret(*sp, l, l, fl, replstr);
patoffset = 0;
Index: Src/pattern.c
===================================================================
RCS file: /cvsroot/zsh/zsh/Src/pattern.c,v
retrieving revision 1.4
diff -u -r1.4 pattern.c
--- Src/pattern.c 2000/04/06 18:44:01 1.4
+++ Src/pattern.c 2000/04/14 11:26:05
@@ -1757,6 +1757,9 @@
* over, that doesn't matter: we should fail anyway.
* The pointer also tells us where the asserted
* pattern matched for use by the exclusion.
+ *
+ * P.S. in case you were wondering, this code
+ * is horrible.
*/
Upat syncstrp;
unsigned char *oldsyncstr;
@@ -1782,6 +1785,7 @@
char savchar, *testptr;
char *savpatinstart = patinstart;
int savforce = forceerrs, savpatinlen = patinlen;
+ int savpatflags = patflags;
forceerrs = -1;
savglobdots = globdots;
matchederrs = errsfound;
@@ -1800,6 +1804,12 @@
testptr = patinstart + (syncpt - syncstrp->p);
DPUTS(testptr > matchpt, "BUG: EXCSYNC failed");
savchar = *testptr;
+ /*
+ * If this isn't really the end of the string,
+ * remember this for the (#e) assertion.
+ */
+ if (savchar)
+ patflags |= PAT_NOTEND;
*testptr = '\0';
next = PATNEXT(scan);
while (next && P_ISEXCLUDE(next)) {
@@ -1848,6 +1858,7 @@
next = PATNEXT(next);
}
*testptr = savchar;
+ patflags = savpatflags;
globdots = savglobdots;
forceerrs = savforce;
if (ret)
@@ -2015,11 +2026,11 @@
*/
return 0;
case P_ISSTART:
- if (patinput != patinstart)
+ if (patinput != patinstart || (patflags & PAT_NOTSTART))
fail = 1;
break;
case P_ISEND:
- if (*patinput)
+ if (*patinput || (patflags & PAT_NOTEND))
fail = 1;
break;
case P_END:
Index: Src/zsh.h
===================================================================
RCS file: /cvsroot/zsh/zsh/Src/zsh.h,v
retrieving revision 1.5
diff -u -r1.5 zsh.h
--- Src/zsh.h 2000/04/13 22:25:04 1.5
+++ Src/zsh.h 2000/04/14 11:26:05
@@ -1024,6 +1024,8 @@
#define PAT_STATIC 0x0040 /* Don't copy pattern to heap as per default */
#define PAT_SCAN 0x0080 /* Scanning, so don't try must-match test */
#define PAT_ZDUP 0x0100 /* Copy pattern in real memory */
+#define PAT_NOTSTART 0x0200 /* Start of string is not real start */
+#define PAT_NOTEND 0x0400 /* End of string is not real end */
/* Globbing flags: lower 8 bits gives approx count */
#define GF_LCMATCHUC 0x0100
--
Peter Stephenson <pws@cambridgesiliconradio.com>
Cambridge Silicon Radio, Unit 300, Science Park, Milton Road,
Cambridge, CB4 0XL, UK Tel: +44 (0)1223 392070
^ permalink raw reply [flat|nested] 5+ messages in thread
* RE: PATCH: fix for (#s) and (#e) in param substs
2000-04-14 11:27 PATCH: fix for (#s) and (#e) in param substs Peter Stephenson
@ 2000-04-14 11:42 ` Andrej Borsenkow
2000-04-14 12:15 ` Peter Stephenson
2000-04-14 12:00 ` Zefram
1 sibling, 1 reply; 5+ messages in thread
From: Andrej Borsenkow @ 2000-04-14 11:42 UTC (permalink / raw)
To: Peter Stephenson, Zsh hackers list
> This wasn't as bad as it could have been. Things like
>
> ${foo//(#s)d*time(#e)/d fight/}
>
> are all now supposed to work. One thing I like about this is
> it allows you
> to forget about not only anchor characters in / and //, but
> even ## and %%,
> and with the (S) parameter flag # and % too --- the /-forms
> together with
> the start and end assertions present a single consistent interface for
> pattern substitutions.
>
Just a question - is there any speed penalty in using / instead of # or
%?
-andrej
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: PATCH: fix for (#s) and (#e) in param substs
2000-04-14 11:27 PATCH: fix for (#s) and (#e) in param substs Peter Stephenson
2000-04-14 11:42 ` Andrej Borsenkow
@ 2000-04-14 12:00 ` Zefram
2000-04-14 13:03 ` Peter Stephenson
1 sibling, 1 reply; 5+ messages in thread
From: Zefram @ 2000-04-14 12:00 UTC (permalink / raw)
To: Peter Stephenson; +Cc: Zsh hackers list
Peter Stephenson wrote:
>By the way, I inserted a hostage to fortune in the manual by claiming that
>`(^(#s))' etc. also work, i.e. you can assert that you are not at the start
>of the string.
Careful. If that's been implemented correctly, `(^(#s))' will match
an arbitrarily long substring anywhere, in addition to a zero-length
substring anywhere other than the start of the string. The semantics of
negation mean that it will match any substring other than what `(#s)'
will match; i.e., it will match anything except for a zero-length
substring at the start of the string. The documentation should be made
clearer on this point. We could also do with proper negative anchors,
that will match only a zero-length string in places other than the
{start,end} of the string. Or if someone feels like implementing my
generalised anchors proposal (I don't have the time at the moment), then
`(#Aan:...)' could be used to negate the sense of an anchor.
-zefram
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: PATCH: fix for (#s) and (#e) in param substs
2000-04-14 11:42 ` Andrej Borsenkow
@ 2000-04-14 12:15 ` Peter Stephenson
0 siblings, 0 replies; 5+ messages in thread
From: Peter Stephenson @ 2000-04-14 12:15 UTC (permalink / raw)
To: Zsh hackers list
> Just a question - is there any speed penalty in using / instead of # or
> %?
There will be some effect, since there is no optimisation to look for an
initial (#s) or final (#e) in the parameter substitution code, unlike # and
% where the start/end anchor is hard-coded. I had to think about this...
Ksh form Form with explicit anchors
${..#..} ${(S)../(#s)..}
${..##..} ${../(#s)..}
In the second form in each case, we still advance up the string looking for
matches, so it's slower, particularly for a long test string. However, the
test will always fail on the first attempt, so it doesn't depend on the
complexity of the pattern after the anchor.
${..%%..} ${../..(#e)}
Not much difference, since in the former case we march up the string until
we find a match, pretty much as in the second case.
${..%..} ${(S)../..(#e)}
These could be quite a lot slower with a fixed tail for a long string,
e.g. ${(S)foo/.c(#e)} where $foo is long, since that has to march up the
string to find a match, then shorten from the end (though the latter will
always fail on the second match so isn't really a problem). But with a
fixed string you wouldn't bother with the (S) parameter flag.
Anyway, I'd be interested to hear if you can notice any difference using /.
It wouldn't be that hard to do special cases. I don't know if it's worth
it. I expect these bits are only going to be used by the real zsh nutcases
:-).
--
Peter Stephenson <pws@cambridgesiliconradio.com>
Cambridge Silicon Radio, Unit 300, Science Park, Milton Road,
Cambridge, CB4 0XL, UK Tel: +44 (0)1223 392070
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: PATCH: fix for (#s) and (#e) in param substs
2000-04-14 12:00 ` Zefram
@ 2000-04-14 13:03 ` Peter Stephenson
0 siblings, 0 replies; 5+ messages in thread
From: Peter Stephenson @ 2000-04-14 13:03 UTC (permalink / raw)
To: Zsh hackers list
> Peter Stephenson wrote:
> >By the way, I inserted a hostage to fortune in the manual by claiming that
> >`(^(#s))' etc. also work, i.e. you can assert that you are not at the start
> >of the string.
>
> Careful. If that's been implemented correctly, `(^(#s))' will match
> an arbitrarily long substring anywhere, in addition to a zero-length
> substring anywhere other than the start of the string.
Yes, this was at the back of my mind but I didn't really work it all out.
(''~(#s)) works, though is a bit clumsy.
Index: Doc/Zsh/expn.yo
===================================================================
RCS file: /cvsroot/zsh/zsh/Doc/Zsh/expn.yo,v
retrieving revision 1.5
diff -u -r1.5 expn.yo
--- Doc/Zsh/expn.yo 2000/04/14 11:49:31 1.5
+++ Doc/Zsh/expn.yo 2000/04/14 13:01:06
@@ -1318,7 +1318,10 @@
provides a single simple and memorable method.
Note that assertions of the form `tt((^(#s)))' also work, i.e. match
-anywhere except at the start of the string.
+anywhere except at the start of the string, although this actually means
+`anything except a zero-length portion at the start of the string'; you
+need to use `tt((""~(#s)))' to match a zero-length portion of the string
+not at the start.
)
enditem()
--
Peter Stephenson <pws@cambridgesiliconradio.com>
Cambridge Silicon Radio, Unit 300, Science Park, Milton Road,
Cambridge, CB4 0XL, UK Tel: +44 (0)1223 392070
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2000-04-14 13:04 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2000-04-14 11:27 PATCH: fix for (#s) and (#e) in param substs Peter Stephenson
2000-04-14 11:42 ` Andrej Borsenkow
2000-04-14 12:15 ` Peter Stephenson
2000-04-14 12:00 ` Zefram
2000-04-14 13:03 ` Peter Stephenson
Code repositories for project(s) associated with this public inbox
https://git.vuxu.org/mirror/zsh/
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).