Re: [Springframework-developer] RE : [Springframework-developer] volunteers to test beandoc

[Darren D. <da...@da...>]

On Fri, January 14, 2005 11:08, tho...@ac... said:

> I gave beandoc a try on a pretty large project with 150 beans spread ov=
er 18
> bean definition files.  It worked very well.  I would make three sugges=
tions:

Thanks Thomas, I appreciate your help.

> 1) Add an option to deactive the generation of graphs.  In some situati=
ons it
> may not be possible to install Graphviz.  A consolidated graph may also=
 become
> unworkable when you have hunderds of beans.

This will certainly be done - at the moment if GraphViz is not installed =
or
not correctly configured, the documentation should still work (with warni=
ngs
in the output).  The HTML generation just needs to be made aware of the l=
ack
of graphs.

> 2) Integrate this as a Maven report.  (There's something going on along=
 those
> lines @
> http://maven-plugins.sourceforge.net/maven-springgraph-plugin/index.htm=
l)

I'm not a Maven user personally and don't know much about this, but I'll
certainly take a look at some point.  The one you pointed to seems to be =
a
wrapper around Mike Thomas' SpringViz XSL transformation which simply han=
dles
graphing of a single context file.

> 3) Right now, beans that are not linked to other beans are added horizo=
ntally
> to the top of the page one next to the other.  This makes the graph pre=
tty
> wide when you have a lot of single beans.  Would there be a way to rear=
range
> them so that they takes up less width?

yes, that's a bit annoying.  GraphViz can be configured to rank nodes in =
a
certain way, the difficulty is coding this in a sensible manner so that g=
raphs
look good regardless of the number of unlinked beans.  I could probably f=
ix it
so that you can define application layers based on bean name patterns (or
class name patterns) and have the layer attribute used to rank them.  The
problem would still manifest though if you had a lot of unlinked beans in=
 the
same application layer.

There is a facility (not fully implemented yet) to specify a pattern of b=
ean
names (or class names) that you wish to be ignored.  These will still be
documented in the HTML but not included in the graphs - potentially handy=
 for
things like post processing beans, message sources and so on that don't r=
eally
benefit from being graphed.

I'll think about this some more.

Thanks once again for your feedback.

--=20
Darren Davison
Public Key: #DD356B0D