title: Number Separator Characters
description:
 ANS Forth specifies that '.' (dot) must be used in both double
 numbers and floating-point numbers, whatever notation is accepted
 in the country where the program is used. In addition, there is
 no way to separate groups of characters, for example, can you
 determine from the first sight whether fffffff is a 32-bit value
 with all bits set?

 DP-CHAR ( -- a-addr )
 A variable, stores ASCII value of double number separator.
 Default: '.'

 FP-CHAR ( -- a-addr )
 A variable, stores ASCII value of float number separator.
 Default: '.'

 NU-CHAR ( -- a-addr )
 A variable, stores ASCII value of the separator that is ignored.
 'NU' stands for 'not used'.
 Default: '_'

Examples:
 hex fff_ffff is clearly different from ffff_ffff

 after the assignment CHAR , FP-CHAR !
 123,4 denotes a floating-point number while
 123.4 is a double number.

original-messages:
 From: sfp@mpeltd.demon.co.uk (Stephen Pelc)
 Message-ID: <3bd6b8e2.394732224@192.168.0.1>

 [...]

 There is a ... strategy which MPE has used for several
 years, and which satisfies most international users.
 Define two system variables DP-CHAR and FP-CHAR to hold the
 double number and floating point number separators. Use these
 variables instead of constants in your text scanning routines.

 For ANS use set both of these to '.'

 For some countries use:
   char , dp-char !
   char . fp-char !
 For others use:
   char . dp-char !
   char , fp-char !
 and so on.

 I am also considering adding a character which can be ignored
 during scanning for numbers. This is useful for masks and literal
 addresses, especially for embedded systems, e.g.
   $BF00:0040
 for a 32 bit address. I suggest the use of ':' as the default, and
 the system variable will be called NU-CHAR for not-used.

 Stephen

 From: Keith Wootten <Keith@wootten.demon.co.uk>
 Message-ID: <gT+1NEAIEu17EwYs@wootten.demon.co.uk>

 >I am also considering adding a character which can be ignored
 >during scanning for numbers.
 [...]

 Can I suggest _ as a default instead?  I use this a lot for binary
 particularly when bits are grouped for different purposes in a
 configuration register.  I tried a few, but _ is IMO the clearest.

 eg

 %1011:0000:101010:001000:111010001000

 or

 %1011_0000_101010_001000_111010001000

page-added-by: mlg