bug#13687: /srv/bzr/emacs/trunk r111878: * lisp/replace.el(read-regexp): Let-bind `default' to the first

[Drew Adams]

> > E.g., in the code I cited, if a user does not want the same 
> > defaulting behavior for commands `occur', `how-many', etc.,
> > she can set option `search/replace-default-fn' to a function
> > that distinguishes them (e.g., using `this-command', as
> > Jambunathan suggested).
> 
> Interesting suggestion there.
> 
> This makes me think that there is no need for multiple
> `hi-lock-read-regexp-defaults-function' and a separate
> `occur-read-regexp-defaults-function' etc.  But a single
> `read-regexp-defaults-function' that cases on `this-command'.

The question, as I said, is whether it makes sense, for the particular commands
that we group to use the same option, to provide the default regexp (or other
string) in the same way.

I can't speak to whether that is the case for hi-lock, occur, etc.  But if it is
true, then yes, a single option for such a group of commands makes sense.

> The function can return a symbol token like `t' for
> `this-command's which it doesn't want to meddle with but
> return nil or a regexp or list of regexps for commands it
> wants to insinuate.

That is not what I suggested.  I suggested that the option value be a function
that returns a string to use as the default value when reading user input.

What I said in the passage you cite is that that function (the value of the
option) could, if the user so wants, itself test `this-command' and provide a
different string depending on the current command.

> Is there any problem with this 
> `read-regexp-defaults-function' approach?

I think you're suggesting that the option value be a function that returns t or
nil, instead of returning a default-value string.  It's not clear to me how a
given command such as `occur' would make use of that Boolean return value.

As I noted before, I would not _encourage_ users to use a dispatching function
as the option value, but that would not (could not) be prevented.  They can do
anything they want using any function they want.

The out-of-the-box design should make a reasonable assumption about which
commands to group (i.e., which should use the same option).

If it is expected that some command that reads a regexp would generally be
better off with a different defaulting behavior, then that command should not
use the group option.  It could use its own, similar option, with a different
default option value (a different default-value-providing function).  Or it
could hard-code its defaulting, or whatever.

The use of a function to dispatch according to the current command should be
exceptional, IMO - only a fallback possibility and not something to be
encouraged.