Book recommendations

[Thomas Mercer-Hursh]

I am wondering if anyone has a book recommendation for using UML backwards  By backwards, I mean creating a highly specific PSM from an existing system and then working to create more abstract models based on the highly concrete ones. Best would be a book with a lot of good concrete information about what type of component or link to use for what purpose.  FWIW, the legacy system is not OO.  

The long term goal here is application transformation by abstracting to PIM or CIM and then using MDA to generate a new PSM which is using a different architecture and technology.

[thomaskilian]

Would be nice to have that, but likely there will be none. I'm just going through that interface/component hell. Really funny that expensive consulting companies prefer Excel to document this. Importing the stuff into EA helped really a lot, showing lots of inconsistencies almost immediately.

[Thomas Mercer-Hursh]

Yeah, I'm quite excited by what I think we will be able to do with the tools we are building.  The project is described here http://www.oehive.org/UMLFromABL .  It feels to me like we will have a really quite significant record of the application in EA when this first stage is done ... but I might move a little faster and more confidently if I had a bit of guidance!

[KP]

I am wondering if anyone has a book recommendation for using UML backwards

Put the following into Google:

Code: [Select]

site:www.amazon.com uml+legacy
The first two results look promising.

[Thomas Mercer-Hursh]

"Modernizing Legacy Systems: Software Technologies,  Engineering Processes, and Business Practices" sounds like it might be an apt book for the general modernization topic, although I don't get the impression that it covers the UML in the kind of detail I was hoping for.  "Legacy Systems: Transformation Strategies" which is sold as a companion seems like a potentially good pairing, but again not so much UML. "Advances in UML- and XML-Based Software" doesn't seem to me to be on topic, though.  I'll keep looking farther, but the flavor of many of these seems to be either a very specific focus, e.g., J2EE, which is inappropriate for the 4GL context I am working with, or where the UML contribution is just a chapter or so.  The latter type may be very interesting books to pursue, but don't necessarily deal with the specific issues I was hoping for.  Good thought though.  Thanks.

[sl@sh]

I am currently working on industry software parts of which date back more than 20 years. Needless to say, even though parts have been developed quite recently, more than 95% of the code (including most of the recent 'classes'!) are not OO at all.

One of my tasks is to slowly convert this dinosaur into something remotely resembling a modern system. Unfortunately the code includes heaps of highly specialized algorithms, so a complete redesign is currently not an option - especially when considering that most of the original designers programmers are no longer available.

I got a very good start by simply importing whole directories of code into EA. At least this way I got a basic overview of typedefs, structs, and classes already in existance (and some duplicate names ). EA provides several options to show dependencies not all of which are enabled by default, so be sure to check these before you even start importing. The resulting diagrams will likely be intimidating at first, but if you're lucky you can clean them up pretty fast by hiding some of the central typedefs (and the tons of relations referring to them). At the very least you will have something to show your superior when he asks why you want to redesign the system  

[Thomas Mercer-Hursh]

One of the aspects of our load tool is that it will have done a very exhaustive analysis of call relationships prior to the import, so we will have a very clear idea of relationships co code to code and code to data.  I have also done all I can think of in the profile to define groupings so as to be able to create meaningful packages.  My hope is that one can then have detailed views inside of packages and package level relationships for larger scope, keeping both levels of diagram to manageable complexity.

The book request is to get further guidance to make sure that I am mapping things to the right UML.

[kayani]

See if this article may help, http://webservices.sys-con.com/read/275129.htm

[Thomas Mercer-Hursh]

That is an interesting article, thanks, and very much in line with our own thinking.  However, it doesn't really replace the book I was looking for in terms of going in to the details of what UML to use for what aspect of the legacy application.  I don't know that the book has to be about reverse engineering specifically, but it should deal with platform specific modeling of non-OO environments.

[«Midnight»]

Well Thomas,

This is sounding more and more like a book yet to be written. Perhaps you are among the few in a position to do so. Of course, you are also very aware of the issues involved.

I'll stay tuned to this thread, in the hope that what you are looking for comes to light.

David

[Thomas Mercer-Hursh]

Well, before I can write the book, I have to figure it out!

And, of course, I am going to be figuring it out in the context of applications written in the Progress ABL (Advanced Business Language), a 4GL, so the audience may not justify a book.