PATCH: Re: blah*[TAB] (difference between 3.1.6 and 3.1.9)

[Sven Wischnowsky <wischnow@xxxxxxxxxxxxxxxxxxxxxxx>]

Bart Schaefer wrote:

> On Jun 7,  8:49am, Sven Wischnowsky wrote:
> } Subject: Re: blah*[TAB] (difference between 3.1.6 and 3.1.9)
> }
> } I wrote:
> } 
> } > Bart Schaefer wrote:
> } > 
> } > > Something's definitely gone wrong with expand-or-complete.  This was not
> } > > intentional.
> } > 
> } > Ahem. Changed in 9777 in reply to 9756 (guess the author).
> } > 
> } > 9777 commented out some stuff in doexpansion(). Change it back?
> } 
> } I want to get that from my todo-list. Should we change it back? Bart
> } (9765 was from you)?
> 
> Sorry, I couldn't make up my mind yesterday.
> 
> In 9777 you wrote:
> > About the space: this was obviously intentional, too. I've just
> > changed the test in doexpansion() so we can go back easily if suddenly 
> > we decide the old behaviour was better. But what I would really like
> > to have is a better test: when do we want the space and when not.
> 
> I think it should add the space only when the expansion results in more
> than one word being inserted.  The reasoning is that it probably doesn't
> make sense to continue completing only the last of several words that
> were generated from the same pattern, but if only one word resulted then
> it might reasonably be a prefix of a longer completion.

Sounds sensible.

> The potential issue I see with this is that what a lot of people want is
> really expand-AND-complete (not "or"), which makes them think a space
> should be appended if the result is an existing plain file, or something
> like that.  In fact, I used to be one of them; for a long time I had
> "setopt recexact" just so that two tabs would expand and then insert a
> space rather than expand and then produce a listing (when one file name
> is a prefix of another).  I've since come to the conclusion that recexact
> is almost always more trouble than it is worth.  However, that doesn't
> help resolve this ... perhaps another style?

I don't want to make old completion code use styles, I think (and
somehow I seem to want to avoid using styles in C-code, in the hope
that one day we'll find a way to more-or-less automatically find used
styles).


Bye
 Sven

Index: Src/Zle/zle_tricky.c
===================================================================
RCS file: /cvsroot/zsh/zsh/Src/Zle/zle_tricky.c,v
retrieving revision 1.13
diff -u -r1.13 zle_tricky.c
--- Src/Zle/zle_tricky.c	2000/06/07 08:39:57	1.13
+++ Src/Zle/zle_tricky.c	2000/06/08 07:15:54
@@ -1627,7 +1627,7 @@
 static int
 doexpansion(char *s, int lst, int olst, int explincmd)
 {
-    int ret = 1;
+    int ret = 1, first = 1;
     LinkList vl;
     char *ss;
 
@@ -1678,10 +1678,11 @@
 	if (olst != COMP_EXPAND_COMPLETE || nonempty(vl) ||
 	    (cs && line[cs-1] != '/')) {
 #endif
-	if (nonempty(vl)) {
+	if (nonempty(vl) || !first) {
 	    spaceinline(1);
 	    line[cs++] = ' ';
 	}
+	first = 0;
     }
     end:
     popheap();

--
Sven Wischnowsky                         wischnow@xxxxxxxxxxxxxxxxxxxxxxx