Option Options

Option Options

[Wayne Davison]

One thing I think that would be nice to get put into the 3.0 release is
the option naming cleanup that Zefram (I believe) is working on.  Even
if it means delaying the release a little, I like the idea of having
better-named options and some aliases for backward compatibility.

I also like the idea that Bart has espoused of having little or no options
show up in setopt's output if the shell is running in a default manner.
The following patch will undoubtedly be controversial, but it changes the
(un)setopt commands to take into account if this option is on by default
in the current emulation mode.  Here's the output I get running it with
"zsh -f":

% setopt
interactive
monitor
norcs
shinstdin
zle

Note that all the "special" options show up.  This could be changed if
someone wished to mark them as OPT_SPECIAL|OPT_ALL (or whatever).

This does cause all non-default options to be referred to by their
"no" name (like nonomatch).

Finally, I corrected a comment in globals.h that referred to OPT_SET,
which doesn't seem to exist.

..wayne..
---8<------8<------8<------8<---cut here--->8------>8------>8------>8---
Index: builtin.c
@@ -433,13 +433,22 @@
 	/* ksh-style option display -- list all options, *
 	 * with an indication of current status          */
 	printf("Current option settings\n");
-	for(optno = 1; optno < OPT_SIZE; optno++)
-	    printf("%-20s%s\n", optns[optno].name, isset(optno) ? "on" : "off");
-    } else
+	for(optno = 1; optno < OPT_SIZE; optno++) {
+	    if (defset(optno))
+		printf("no%-20s%s\n", optns[optno].name, isset(optno) ? "off" : "on");
+	    else
+		printf("%-22s%s\n", optns[optno].name, isset(optno) ? "on" : "off");
+	}
+    } else {
 	/* list all options that are on, or all that are off */
-	for(optno = 1; optno < OPT_SIZE; optno++)
-	    if (set == isset(optno))
+	for(optno = 1; optno < OPT_SIZE; optno++) {
+	    if (set == (isset(optno) ^ defset(optno))) {
+		if (set ^ isset(optno))
+		    fputs("no", stdout);
 		puts(optns[optno].name);
+	    }
+	}
+    }
 }
 
 /**** job control builtins ****/
Index: globals.h
@@ -451,7 +451,7 @@
 
 EXTERN int emulation;
  
-/* the options; e.g. if opts[SHGLOB] == OPT_SET, SH_GLOB is turned on */
+/* the options; e.g. if opts[SHGLOB] != 0, SH_GLOB is turned on */
  
 EXTERN char opts[OPT_SIZE];
  
Index: init.c
@@ -173,7 +173,7 @@
     for(optno = OPT_SIZE; --optno; )
 	if((fully && !(optns[optno].flags & OPT_SPECIAL)) ||
 	    	(optns[optno].flags & OPT_EMULATE))
-	    opts[optno] = !!(optns[optno].flags & (1 << emulation));
+	    opts[optno] = defset(optno);
 }
 
 static char *cmd;
Index: zsh.h
@@ -1159,6 +1159,7 @@
 #undef isset
 #define isset(X) (opts[X])
 #define unset(X) (!opts[X])
+#define defset(X) (!!(optns[X].flags & (1 << emulation)))
 
 #define interact (isset(INTERACTIVE))
 #define jobbing  (isset(MONITOR))
---8<------8<------8<------8<---cut here--->8------>8------>8------>8---

Re: Option Options

[Zoltan Hidvegi]

> I also like the idea that Bart has espoused of having little or no options
> show up in setopt's output if the shell is running in a default manner.
> The following patch will undoubtedly be controversial, but it changes the
> (un)setopt commands to take into account if this option is on by default
> in the current emulation mode.  Here's the output I get running it with
> "zsh -f":
> 
> % setopt
> interactive
> monitor
> norcs
> shinstdin
> zle

I like this idea.

> Note that all the "special" options show up.  This could be changed if
> someone wished to mark them as OPT_SPECIAL|OPT_ALL (or whatever).

That's fine because they have no real default value.

Zoltan

Re: Option Options

[Zefram]

>One thing I think that would be nice to get put into the 3.0 release is
>the option naming cleanup that Zefram (I believe) is working on.

I'm not actually working on it at the moment.  I decided to wait until
after the 3.0 release before making functional changes like this.  But
if the consensus is in favour, it's really quite easy to produce a
patch.

>I also like the idea that Bart has espoused of having little or no options
>show up in setopt's output if the shell is running in a default manner.
>The following patch will undoubtedly be controversial, but it changes the
>(un)setopt commands to take into account if this option is on by default
>in the current emulation mode.

That's going to break things.  Sometimes one *wants* a list of all
options -- not least in the completion for {un,}setopt.  If this change
is to be made, other things will have to change too.  This definitely
should wait.

Not that I'm implacably opposed to the idea -- I just think that such a
radical change to the behaviour of existing builtins in the default
case is a bad move right before a production release.

-zefram

Re: Option Options

[Wayne Davison]

Zefram writes:
> That's going to break things.  Sometimes one *wants* a list of all
> options -- not least in the completion for {un,}setopt.

All the options are still there.  It's just that some of them are listed
by their "no" name as an option to turn off rather than an "option" that
is already turned on.  This makes the output more compatible with prior
zsh releases, so I don't see what the problem is.  Perhaps I am missing
something.

..wayne..

Re: Option Options

[Zefram]

>All the options are still there.  It's just that some of them are listed
>by their "no" name as an option to turn off rather than an "option" that
>is already turned on.  This makes the output more compatible with prior
>zsh releases, so I don't see what the problem is.  Perhaps I am missing
>something.

Oh, I see.  You list options that default on under their NO_ name.  I
suppose that's sensible.  It doesn't fundamentally change the behaviour
-- at least, no more than my option patch did.

The setopt completion example does need to be reverted to its previous
form though.

-zefram