From help-octave-request at bevo dot che dot wisc dot edu  Thu Oct 12 08:31:58 1995
Subject: Re: BAR not Matlab 4+ compatible
From: Ted dot Harding at nessie dot mcc dot ac dot uk (Ted Harding)
To: jwe at bevo dot che dot wisc dot edu (John Eaton)
Cc: help-octave at bevo dot che dot wisc dot edu
Date: Thu, 12 Oct 1995 14:19:07 +0200 (BST)

> I had no idea.
> 
> Matlab is apparently placing the points like this:
> 
> 	         o---o
> 	         |   |
>        o---o   |   |
>        |   |   |   |
>        |   |   |   |   o---o
>        |   |   |   |   |   |
>      o-o   o-o-o   o-o-o   X
> 
> where an `o' represents a single point and an `X' represents a
> duplicate point.
> 
> This seems very odd to me and probably not worth duplicating exactly.

This is not how Matlab USED to do it (see bar.m from MatLab, vintage 1988,
below). This would draw the following graph (where "+" marks a data
point):

 	   o-+-o
 	   |   |
       o-+-o   |
       |   |   |
       |   |   o-+-o
       |   |   |   |
       o   o   o   o

It basically follows the principle of octave's "bar".

I, too, think that this is how I like to see "bar"s, because I really
use histograms and have little use for the spreadsheet kind (in fact the
Matlab [1988] "hist" function uses "bar" to draw the histogram). However,
there is a place for the bar-chart as such (like the one drawn by John
Eaton); but to be fully useful this needs an additional parameter to
control the "mark-space" ratio (i.e. what proportion of the distance
between index marks is occupied by the width of the bar, relative to the
proportion occupied by the space between the bars).

Ted.                                    (Ted dot Harding at nessie dot mcc dot ac dot uk)