{smcl}
{* 21jun2004}{...}
{hline}
{center:{hi:The undocumented classes underlying {cmd:graph}}}
{hline}

		{hi:Lions,}
		      {hi:and Tigers,}
				 {hi:and Bears}

{hline}
{p 4 4 4}{it}
We are going to introduce some undocumented features of Stata'a graphics.
Given Stata's history of openness, and that even the {cmd:gph} commands
underlying the previous graphics system were documented, it may be surprising
that the facilities underlying the {cmd:graph} command are undocumented.  There
are several reasons for this.  Whereas the {cmd:gph} system encompassed only a
es.smcl of commands, the facilities underlying the new {cmd:graph} command
are big ...{sf}{p_end}
{hline}

{p 0 2 0}o {view big.smcl:>> Big}

{p 0 2 0}o Intermediate layer need for widespread adoption

{p 0 2 0}o {cmd:twoway} is very, very flexible

{hline}
{p 4 4 4}{it}
While documenting such a system is a large task, that wasn't the reason it was
not documented.  That we expected significant changes to this system based on
your comments and input was.  We expected that these changes would likely
break code written to the original API of the system and that has indeed
turned out to be true.

{p 4 4 4}{it}
Second, we believed, and still do believe, that although the underlying
faculties beneath graphics can be accessed by users, an intermediate layer
will be needed for widespread use of these facilities.

{p 4 4 4}{it}
Third, and most important the {cmd:twoway} command, combined with Stata's data
management commands is flexible enough to produce a host of custom graphs.  As
Nick Cox, Ulrich Kohler, Michael Blasnik, and many other have shown, this is
typically much easier than rolling your own graph using low-level facilities.

{p 4 4 4}{it}
Here are a few examples using this approach and I heartily recommend it as the
first, best course of action.
{sf}


{hline}
{center:{view cover3.smcl:<<}   {view tindex.smcl:index}   {view twoways.smcl:>>}}