Re: How to enter a floating-point number ?

[Russell Shaw <rjshaw netspace net au>]

David Necas (Yeti) wrote:
> On Sat, May 29, 2004 at 09:40:47PM +1000, Russell Shaw wrote:
>
> > This is a parser problem. Every time a key is pressed, the callback
> > advances thru your implementation of a floating-point number parser.
> > When an invalid character is pressed, your callback can simply refuse
> > to accept the character, make a beep, and display an error message in
> > the status line. The parser is not hard (only 100 lines of code as
> > a switch-statement state machine).
>
> IMHO this is a logical problem.
>
> If you allow only things that correctly parse as floating
> point numbers, the user will have hard time editing the
> numbers, because after each key press the number has to be
> parseable.  This is very annoying, especially for numbers
> with an exponent.
>
> But if you allow any transitional states, you'll lose you
> point because there will be non-number things that can be
> entered then, so you'll have to handle them anyway.

The parsers i've done go forwards and backwards, so you can
use the backspace key. If you change a character midway in the
number, the parser just rescans the entire string (as it does
for all keypresses) and allows or disallows that change,
because it remembers the previous entry.