[ oops, forgot to reply to this one when I first saw it ]

"Perry E. Metzger" wrote:

> Simon Burge <simonb@netbsd.org> writes:
> > Module Name:	src
> > Committed By:	simonb
> > Date:		Fri Dec 26 11:30:36 UTC 2003
> >
> > Modified Files:
> > 	src/lib/libc/gen: humanize_number.c
> >
> > Log Message:
> > Don't ever add a ".0" to a single digit number.
> 
> Why is .0 wrong when .1 would be okay? I can see not wanting to do 1.0
> bytes (where it would be silly), but if a decimal makes sense in some
> contexts, why not in others? Shouldn't we just be following normal
> scientific significant digits rules?

Maybe the wording of the commit message wasn't the best, but a "single
digit number" was an _input_ number of less than 10 - not any _output_
number less than 10.  Other numbers will still be decimalised if
necessary.  In the following, the size of the file is the same as the
filename:

	-rw-r--r--  1 simonb  simonb    0B Jan  3 13:07 0
	-rw-r--r--  1 simonb  simonb    1B Jan  3 13:07 1
	-rw-r--r--  1 simonb  simonb   10B Jan  3 13:08 10
	-rw-r--r--  1 simonb  simonb  100B Jan  3 13:08 100
	-rw-r--r--  1 simonb  simonb  1.0K Jan  3 13:08 1000
	-rw-r--r--  1 simonb  simonb  9.8K Jan  3 13:08 10000

Also, in a later message you say:

> Perhaps I'm not human, but I find randomly shifting numbers of
> significant digits confusing...

The old output of ls was:

	-rw-r--r--  1 simonb  simonb  0.0B Jan  3 13:07 0
	-rw-r--r--  1 simonb  simonb  1.0B Jan  3 13:07 1
	-rw-r--r--  1 simonb  simonb   10B Jan  3 13:08 10
	-rw-r--r--  1 simonb  simonb  100B Jan  3 13:08 100
	-rw-r--r--  1 simonb  simonb  1.0K Jan  3 13:08 1000
	-rw-r--r--  1 simonb  simonb  9.8K Jan  3 13:08 10000

I find the jump from "1.0B" to "10B" confusing...  Maybe if
humanize_number() had a floating point version the output 0.0B
would mean more.

Simon.
--
Simon Burge                            <simonb@wasabisystems.com>
NetBSD Support and Service:         http://www.wasabisystems.com/