Time for an obscure completion question

Time for an obscure completion question

[Bart Schaefer]

Given the completion styles (straight out of compinstall except for moving
the _ignored completer):

zstyle ':completion:*' completer _oldlist _expand _complete _match _ignored _approximate _prefix
zstyle ':completion:*' match-original both
zstyle ':completion:*' matcher-list '' 'r:|[._-,]=* r:|=*' 'm:{a-zA-Z}={A-Za-z} r:|[._-,]=* r:|=*' 'r:|[._-,]=* r:|=* l:|=*'
zstyle ':completion:*' max-errors 2 numeric
zstyle ':completion:*' original true

If I have a directory named "foodatthe/Bar" (to borrow from PWS) and I type:

zsh% cd *at/ba<TAB>

I'm offered the two completions

foodatthe/    *at/ba

This is apparently what I asked for, but it's not what I want. I want the
"original" as considered by _approximate to be the output (for lack of a
better word) of _expand _complete _match.  That is, if there's exactly one
possible expansion of the part with the pattern, then I want that accepted
so that the menu becomes:

foodatthe/Bar   foodatthe/ba

Ignoring that for the moment ... if I type

zsh% cd *at/ba<C-x h>

i.e., invoke completion help, I see a set of contexts and then the possible
completions at the end.  So far so good, but now if I hit TAB the argument
gets erased and I'm left with

zsh% cd 
        ^ cursor here

A similar thing happens here:

zsh% cd *at/ba<TAB><C-e><TAB><TAB><TAB>
zsh% cd  
         ^cursor

The first TAB gives "foodatthe/ba" and lists two completions; I hit C-e to
break out of menu completion (is there a better way?) and TAB again which
gives me the correction "foodatthe/Bar/" and again lists two completions.
Hit TAB again and there's a feep; hit TAB a third time and the argument
vanishes, replaced by a space.  I would have expected it to just keep on
feeping at me for as long as I bang on tab.

-- 
Bart Schaefer                                 Brass Lantern Enterprises
http://www.well.com/user/barts              http://www.brasslantern.com

Re: Time for an obscure completion question

[Sven Wischnowsky]

Bart Schaefer wrote:

> Given the completion styles (straight out of compinstall except for moving
> the _ignored completer):
> 
> zstyle ':completion:*' completer _oldlist _expand _complete _match _ignored _approximate _prefix
> zstyle ':completion:*' match-original both
> zstyle ':completion:*' matcher-list '' 'r:|[._-,]=* r:|=*' 'm:{a-zA-Z}={A-Za-z} r:|[._-,]=* r:|=*' 'r:|[._-,]=* r:|=* l:|=*'
> zstyle ':completion:*' max-errors 2 numeric
> zstyle ':completion:*' original true
> 
> If I have a directory named "foodatthe/Bar" (to borrow from PWS) and I type:
> 
> zsh% cd *at/ba<TAB>
> 
> I'm offered the two completions
> 
> foodatthe/    *at/ba
> 
> This is apparently what I asked for, but it's not what I want. I want the
> "original" as considered by _approximate to be the output (for lack of a
> better word) of _expand _complete _match.  That is, if there's exactly one
> possible expansion of the part with the pattern, then I want that accepted
> so that the menu becomes:
> 
> foodatthe/Bar   foodatthe/ba

[ Aside: whenever I see such requests I wonder what comes next... ;-]

The `foodatthe/' comes from _approximate. Bummer. _match kept itself
from being called more than once (for different matchers), even though 
I changed the completion code to at least partly handle the mixture of 
pattern-matching and match-specs some time ago. The patch comments out 
the match-spec-number test in _match. With that you get `foodatthe/ba' 
on the line. The reason that it doesn't complete the `ba' to `Bar' is
that _path_files stops when the first component with a pattern is
reached and that adding the whole string, with one part needing
pattern matching and the other part needing the match spec can't
behandled by the completion code. But the next TAB gives you
`foodatthe/Bar/'.

Hm.

> Ignoring that for the moment ... if I type
> 
> zsh% cd *at/ba<C-x h>
> 
> i.e., invoke completion help, I see a set of contexts and then the possible
> completions at the end.  So far so good, 

No, the completions (corrections) shouldn't be shown. I'm a bit
confused why I didn't see that before. The patch makes _approximate
don't redefine compadd() if it is already a function. This is ok for
_complete_help, if we ever come up with another function that defines
a compadd() function, we'll have to rethink this.

> but now if I hit TAB the argument
> gets erased and I'm left with
> 
> zsh% cd 
>         ^ cursor here
> 
> A similar thing happens here:
> 
> zsh% cd *at/ba<TAB><C-e><TAB><TAB><TAB>
> zsh% cd  
>          ^cursor
> 
> The first TAB gives "foodatthe/ba" and lists two completions; I hit C-e to
> break out of menu completion (is there a better way?) and TAB again which
> gives me the correction "foodatthe/Bar/" and again lists two completions.
> Hit TAB again and there's a feep; hit TAB a third time and the argument
> vanishes, replaced by a space.  I would have expected it to just keep on
> feeping at me for as long as I bang on tab.

Both should be fixed now. The first one was caused by _oldlist which
took the list from _complete_help. For now I made _oldlist test if the 
last widget was _complete_help and ignore the list in that case. It
was already testing widget names, but for a somewhat cleaner solution
we probably should use a new element in $_lastcomp that is set if the
list is generated by some `special' function and should not be used by 
any other completion function.
The second one I couldn't reproduce after having fixed the other
things...

Bye
 Sven

Index: Completion/Core/_approximate
===================================================================
RCS file: /cvsroot/zsh/zsh/Completion/Core/_approximate,v
retrieving revision 1.2
diff -u -r1.2 _approximate
--- Completion/Core/_approximate	2000/04/01 20:43:43	1.2
+++ Completion/Core/_approximate	2000/04/11 09:35:42
@@ -10,7 +10,7 @@
 
 [[ _matcher_num -gt 1 || "${#:-$PREFIX$SUFFIX}" -le 1 ]] && return 1
 
-local _comp_correct _correct_expl comax cfgacc
+local _comp_correct _correct_expl comax cfgacc redef
 local oldcontext="${curcontext}" opm="$compstate[pattern_match]"
 
 zstyle -s ":completion:${curcontext}:" max-errors cfgacc || cfgacc='2 numeric'
@@ -40,17 +40,20 @@
 # to stick the `(#a...)' in the right place (after an
 # ignored prefix).
 
-compadd() {
-  [[ ${argv[(I)-[a-zA-Z]#U[a-zA-Z]#]} -eq 0 &&
-     "${#:-$PREFIX$SUFFIX}" -le _comp_correct ]] && return
-
-  if [[ "$PREFIX" = \~*/* ]]; then
-    PREFIX="${PREFIX%%/*}/(#a${_comp_correct})${PREFIX#*/}"
-  else
-    PREFIX="(#a${_comp_correct})$PREFIX"
-  fi
-  builtin compadd "$_correct_expl[@]" "$@"
-}
+if (( ! $+functions[compadd] )); then
+  redef=yes
+  compadd() {
+    [[ ${argv[(I)-[a-zA-Z]#U[a-zA-Z]#]} -eq 0 &&
+       "${#:-$PREFIX$SUFFIX}" -le _comp_correct ]] && return
+
+    if [[ "$PREFIX" = \~*/* ]]; then
+      PREFIX="${PREFIX%%/*}/(#a${_comp_correct})${PREFIX#*/}"
+    else
+      PREFIX="(#a${_comp_correct})$PREFIX"
+    fi
+    builtin compadd "$_correct_expl[@]" "$@"
+  }
+fi
 
 _comp_correct=1
 
@@ -81,7 +84,7 @@
       [[ "$compstate[list]" != list* ]] &&
           compstate[list]="$compstate[list] force"
     fi
-    unfunction compadd
+    [[ -n "$redef" ]] && unfunction compadd
     compstate[pattern_match]="$opm"
 
     return 0
@@ -91,7 +94,7 @@
   (( _comp_correct++ ))
 done
 
-unfunction compadd
+[[ -n "$redef" ]] && unfunction compadd
 compstate[pattern_match]="$opm"
 
 return 1
Index: Completion/Core/_match
===================================================================
RCS file: /cvsroot/zsh/zsh/Completion/Core/_match,v
retrieving revision 1.2
diff -u -r1.2 _match
--- Completion/Core/_match	2000/04/01 20:43:43	1.2
+++ Completion/Core/_match	2000/04/11 09:35:42
@@ -9,7 +9,7 @@
 # expand-or-complete function because otherwise the pattern will
 # be expanded using globbing.
 
-[[ _matcher_num -gt 1 ]] && return 1
+### Shouldn't be needed any more: [[ _matcher_num -gt 1 ]] && return 1
 
 local tmp opm="$compstate[pattern_match]" ret=0 orig ins
 
Index: Completion/Core/_oldlist
===================================================================
RCS file: /cvsroot/zsh/zsh/Completion/Core/_oldlist,v
retrieving revision 1.2
diff -u -r1.2 _oldlist
--- Completion/Core/_oldlist	2000/04/01 20:43:43	1.2
+++ Completion/Core/_oldlist	2000/04/11 09:35:42
@@ -13,7 +13,8 @@
 # Do this also if there is an old list and it was generated by the
 # completer named by the oldlist_list key.
 
-if [[ -n $compstate[old_list] && $list != never ]]; then
+if [[ -n $compstate[old_list] && $list != never &&
+      $LASTWIDGET != _complete_help ]]; then
   if [[ $WIDGET = *list* && ( $list = always || $list != shown ) ]]; then
     compstate[old_list]=keep
     return 0
@@ -33,8 +34,10 @@
 # and the style :oldlist:old-menu is `true', then we cycle through the
 # existing list (even if it was generated by another widget).
 
-if [[ -z $compstate[old_insert] && -n $compstate[old_list] ]]; then
+if [[ -z $compstate[old_insert] && -n $compstate[old_list] &&
+      $LASTWIDGET != _complete_help ]]; then
   compstate[old_list]=keep
+  return 0
 elif [[ $WIDGET = *complete(|-prefix|-word) ]] &&
      zstyle -t ":completion:${curcontext}:" old-menu; then
   if [[ -n $compstate[old_insert] ]]; then

--
Sven Wischnowsky                         wischnow@informatik.hu-berlin.de

Re: Time for an obscure completion question

[Bart Schaefer]

On Apr 11, 11:36am, Sven Wischnowsky wrote:
} Subject: Re: Time for an obscure completion question
}
} Bart Schaefer wrote:
} 
} > foodatthe/    *at/ba
} 
} I changed the completion code to at least partly handle the mixture of 
} pattern-matching and match-specs some time ago. The patch comments out 
} the match-spec-number test in _match. With that you get `foodatthe/ba' 

That'll be close enough; thanks.

} > but [after C-x h] if I hit TAB the argument
} > gets erased and I'm left with
} > 
} > zsh% cd 
} >         ^ cursor here

This one is fixed now.

} > A similar thing happens here:
} > 
} > zsh% cd *at/ba<TAB><C-e><TAB><TAB><TAB>
} > zsh% cd  
} >          ^cursor
} > 
} > The first TAB gives "foodatthe/ba" and lists two completions; I hit C-e to
} > break out of menu completion (is there a better way?) and TAB again which
} > gives me the correction "foodatthe/Bar/" and again lists two completions.
} > Hit TAB again and there's a feep; hit TAB a third time and the argument
} > vanishes, replaced by a space.  I would have expected it to just keep on
} > feeping at me for as long as I bang on tab.
} 
} Both should be fixed now.

This second one still happens to me.  I've set a few more styles; what I
get now is:

zagzig[72] cd *at/ba<TAB><TAB>
zagzig[72] cd foodatthe/Bar/
Completing `local directories', `directories in cdpath', or `corrections'

There are no completions beneath the verbose output there.  At this point
if I hit TAB a third time, "foodatthe/Bar/" becomes " ".  I tried unsetting
cdpath, but that just changes the message to:

Completing `directory' or `corrections'

Here's the entire list of styles again:

zstyle :completion::complete:cd:: tag-order local-directories path-directories
zstyle ':completion:*' completer _oldlist _expand _complete _match _ignored _approximate _prefix
zstyle ':completion:*' file-sort modification reverse
zstyle ':completion:*' format 'Completing %d'
zstyle ':completion:*' group-name ''
zstyle ':completion:*' ignore-parents parent pwd ..
zstyle ':completion:*' list-colors ''
zstyle ':completion:*' match-original both
zstyle ':completion:*' matcher-list '' 'r:|[._-,]=* r:|=*' 'm:{a-zA-Z}={A-Za-z} r:|[._-,]=* r:|=*' 'r:|[._-,]=* r:|=* l:|=*'
zstyle ':completion:*' max-errors 2 numeric
zstyle ':completion:*' menu 'select=6'
zstyle ':completion:*' original true
zstyle ':completion:*' verbose true

And completion-related options (kshoptionprint format):

noalwayslastprompt    off
alwaystoend           on
noautolist            off
noautomenu            off
noautoparamkeys       off
noautoparamslash      off
noautoremoveslash     off
bashautolist          off
nobeep                off
chasedots             off
chaselinks            off
completealiases       off
completeinword        on
globcomplete          off
nolistambiguous       off
nolistbeep            off
listpacked            off
listrowsfirst         off
nolisttypes           off
magicequalsubst       off
menucomplete          off

-- 
Bart Schaefer                                 Brass Lantern Enterprises
http://www.well.com/user/barts              http://www.brasslantern.com