RE: What do 'hooks' do and how do they do it?

[William Case]

Sorry Drew;

I wish I had calibrated my response to Dirk a bit better.

On Sat, 2009-08-01 at 15:05 -0700, Drew Adams wrote:
> > > I think Drew Adams gave a very clear answer about the Emacs 
> > > implementation,
> > 
> > Yes he did.  However, I tried to point out in my original post that I
> > was aware of how to implement a 'hook' in emacs.
> 
> Well, I did not explain how to implement a hook. I tried to explain what a 
> hook
> is: "a way to allow easy code extension at predefined places".
> 
> I explained that when a hook is run in Emacs, the calling code just calls a
> function (e.g. `run-hooks') that obtains the functions to invoke from a
> variable's value. That's what a hook _is_, for Emacs; that's not how to
> implement a hook.
> 

On rereading your first post, the answer I wanted was, in fact, there.

"A typical example is the code for a major mode, such as
`emacs-lisp-mode'. The last thing such code does, when you turn on the
mode, is to call `run-mode-hooks' to run the hook provided specifically
for that mode, e.g. `emacs-lisp-mode-hook'. That is, it tries to invoke
each of the functions on the list `emacs-lisp-mode-hook', in turn. If
there are no functions on the hook, then nothing is done."


> The general point is that (a) the call (when, why, etc.) is predefined, but 
> (b)
> what to call is determined dynamically (from the value of a variable, in the
> case of Emacs).
> 
> In particular, in Emacs, hooking has _nothing to do with events_. Invocation 
> of
> a hook function in Emacs is _not_ event-driven.
> 
> > > but let me add that the concept of a 'hook' is simply a 
> > > programming technique: the ability to set some function
> > > to be run whenever a particular event happens.
> 
> Nope. (Unless the code that runs the hook variable happens to examine events 
> to
> determine whether to call `run-hooks'.)

The event, so to speak, is turning on the mode. 

> 
> The call to invoke the hook functions is static. The functions to invoke are
> determined dynamically by the value of a variable. The hook is the static 
> part;
> what's hooked is the dynamic part - think of hooking a fish, if you like.
> 
> > I was curious about the programming technique that was used -- or more
> > specifically -- how the hook function was setup so that the hook
> > automagically responded to a program's hook event.
> 
> There is nothing automagical here, and there is no hook event.

Yes, I have tended to over use the term automagical since I first
discovered the word.  Almost always I use it in reference to correcting
my understanding of things that seem automagical.  To me, the
automagical has no place in computing.
> 
> > > The concept is in use in many places (such as SE-Linux), 
> > > but how it's implemented is quite different. In Emacs-Lisp,
> > > the hooks are simply Lisp functions to be called -- no kernel
> > > involved (well...).
> 
> No. The "hooks" are the hook variables plus the calls to `run-hooks' that use
> those variables, both of which are essentially static code. The values of the
> hook variables are dynamic, and they determine which functions (the "hook
> functions") get called. The functions called by the hooks are not the hooks; 
> the
> variables are the hooks.
> 
> More generally, a "hook" is a predefined call to something to be determined
> outside the static calling code. It is, well, a "hook" on which you can hang 
> any
> code you want executed.
> 
> Hooks are static things; dynamic things are hung on the hooks. Think coat 
> hooks
> (stable, static) and coats that might be hung there (movable, dynamic,
> changeable set).
> 
> The hanging typically takes place at a different time from the definition of 
> the
> code that has the hook (the building of the coat rack). That's the point:
> decouple where to do something from what actually gets done there. Those are
> determined at different times: code writing vs runtime.
> 
> Connecting (a) the actual functions to be called to (b) the hook that is made
> available for calling them is, in effect, a form of dynamic or late binding. 
> In
> the case of Emacs, this is simply the binding of a variable to its value.
> 
> This is akin to (funcall foo): the (functional) value of variable `foo' is
> obtained at runtime, and it is called as a function. (funcall foo) is a kind 
> of
> "hook", on which is hung the value of `foo' at runtime. (But in Emacs we
> generally reserve "hook" for a hook variable. Nevertheless, the overall 
> concept
> here is the same.)
> 
> Note that in environments that are much less dynamic than Emacs and Lisp, 
> hooks
> can have greater importance: they are sometimes the _only_ effective way for
> user code to be plugged into the main (binary) code.
> 
> In the early days, giving users a way to branch their own calls into the main
> code was sometimes called a "user exit". (A hook that was hidden was sometimes
> called a "back door".) "User exit" pretty much sums up this claustrophobic 
> state
> of affairs: it was the only way to escape from the program to do something 
> extra
> or different.
> 
I will keep the entirety of this reponse, in my personal emacs
explanations file.

> > > See: http://en.wikipedia.org/wiki/Hooking
> > 
> > The wikipedia site you suggest above covers just about everything I
> > wanted to know.  I don't know how I missed it.
> 
> Most of that page does not particularly apply to Emacs hooks, I'm afraid. The
> general idea applies, but most of the hook mechanisms described on that page
> have little relation to Emacs hooks.
> 

OK.  I'll keep that in mind.

> What is most helpful on that page wrt Emacs hooks are the two sentences that
> constitute the (entire) section "Emacs":
> 
>   "Hooks are an important mechanism for customization of Emacs.
>   A hook is a Lisp variable which holds a list of functions,
>   to be called on some well-defined occasion. (This is called
>   running the hook.)"
> 
> The "well-defined occasion" and the hook variable together constitute the
> stable, "hook" part. The functions on the list at any given time are the
> dynamic, "hung" part.
> 
> The best explanation of Emacs hooks and hooking remains the Emacs doc (which 
> is
> why I pointed you to it, and Wikipedia does likewise).HTH.

All that I can say in my own defence, is that I had somewhat of a
misconception of what hooks were and therefore I DID have good reason to
ask further what hooks were.

-- 
Regards Bill
Fedora 11, Gnome 2.26.3
Evo.2.26.3, Emacs 22.3.1