RE: keypad-*-setup

[Drew Adams]

I've tried to improve the doc for this feature (library keypad.el) - both
the Emacs manual and the doc strings and Customize tags.

I have not tried to mess with texinfo. I will make any changes people think
are needed to the attached doc and code files, and I will then pass the
result on to someone else to modify the texinfo and install in CVS.

Attached is a patch for keypad.el. I did not change the behavior of the
code; I made only cosmetic changes, e.g. doc and Commentary.

Attached is a text file for the Emacs manual. I propose these changes:

- Node "Function Keys": Remove all text starting with "Many keyboards have a
"numeric keypad...".  Put that topic in its own node (see next).

- New node, "Numeric Keypad", child of "Key Bindings", Next node after
"Function Keys". See attached file keypad-doc.txt for the text.

Some questions:

1. A nil ("No change (keep existing bindings)") value for the setup options
seems to be a no-op. What is its use; what does it do (change)? If it does
nothing, then perhaps we should remove it (and modify the doc accordingly).
Here is the :set code for `defcustom' - it seems to do nothing if `value' is
nil. Am I reading this wrong?

:set (lambda (symbol value) (when value (keypad-setup value nil nil value)))

2. Do we really need a separate `numeric' choice? Why not always specify an
explicit decimal-point? That is, why not get rid of the value `numeric'
("Numeric with standard decimal point") and just use the version that
specifies a decimal-point character - it uses `.' as the default value,
which is the same as choosing the option value `numeric'.

3. Shouldn't `integerp' be `char-valid-p' in the test for the value? :match
(lambda (widget value) (integerp value)). In function `keypad-setup',
`char-valid-p' is used to test the character.

keypad-2007-09-12.patch
Description: Binary data

keypad-doc.txt
Description: Text document