I’m an biz/sys analyst. I just discovered that I’m actually doing QVT the whole day (in some way):
Q uery: I’m discovering rules, valid FOR the model. (Aspects, CONSTRAINTS, Checks, …)
V iew: I’m creating views on the world, that contribute to one overall model. (Workflow, Screenflow, Dataflow, …)
T ransformation: I need some glue to keep it all together consistently: the views among each other as well as the views with the queries.
Plainly put, this is the analysts view on Model Transformation … – is it?
|=
Could we say that views are the result of executing query transformations (i.e. transformations that filter the input model and only copy the parts that satisfy the query conditions)?
Yep! Analysts use in-place transformations without refactoring anything – in contrast to people like designers. We don’t touch, just watch ;o)
PS
also queries ARE just transformations FROM model TO a a SET OF tuples. Thus QVT could be abstracted TO TTT.
Think, aside the World-to-Model Tranformation, analysis actually does by creating views, there are 2 ways of M2M Transformation in Analysis:
1. Applying rules to models
For checking or applying aspects of/to models.
2. Keeping views consistent
Creating views to understand the world is nice, but keeping them consistent is hell.
That’s it – did I forget anything?
|=
Version WITH ‘pretty’ pictures: http://BIT.ly/iMxuA9
Just an idea.
Analyzing a model may also needs TO compare this model WITH a reference, say a meta-model by example. TO find, instead OF apparent rules/relations, missing rules/relations. IN that CASE we need TO bring outside DATA INTO the “query/filter/transformation” TO enlighten the analysis IN the resulting “view/result model”. It’s equally non intrusive…
as far as I understood it, I think we have the same idea. In this case:
http://bit.ly/jHT9Y3
the metamodel would be “linearly ordered range”, right? The “out side data” would correspond to the understanding of the structure, as far as I understood it? Of course this is, as you mean it, about much more than just completeness.
Regards
|=
Just twitted your post!
I agree for meta-model at primary interpretation, I think but there is also more than that.
The initial post talks about financial analysis. Financial domain has rules (even if tooooo much often not applied). A complete analysis on something financial should have to check if rules are applied, which will rarely appears in the rough data. This is maybe not exactly the same thing as checking inherent data structure.