午夜拍键惊奇
子夜 编程 代码与我同在
posts - 48,comments - 118,trackbacks - 79

Str#1d. "Invest an Hour" Strategy // activities and model components

- Rather than philosophize endlessly, invest an hour in each of several different ways of modeling a particularly challenging area. Compare your results -- and decide which way to go (based upon actual results, rather than the outcome of a multiweek debate).


Str#1e. "Consider the Domain First, Artifacts After That" Strategy // activities and model components

- Build an object model with a domain expert first. Then add-in content that you can extract from artifacts (existing data models, source code, whatever).

- Reason why: you need the benefit of the former (fresh insights, new ideas) to help you grapple with the latter (what to include, what to exclude).


Str#1f. "Extract Useful Content From An Existing Data Model" Strategy // activities and model components

- Yes, it can be done.

- Best practice: build an initial object model with a domain expert first. Then use that model to help you filter out the classes and attributes (in an previous data model) that are no longer needed. Why: the added domain understanding will help you do a better job leaving unneeded things behind, rather than dragging everything from the past along with you once again.

- For the entities:

. List them. Delete correlation tables. Delete (or revise) names that do not fit the problem domain vocabulary (words that a domain expert uses and understands). Collapse supertypes-subtypes that do not express domain-based generalization-specialization.

- Then, when you work on attributes:

. List them. Delete (or revise) names that do not fit the problem domain vocabulary (words that a domain expert uses and understands). Delete flags, indicators, sequence numbers, and unique keys -- nearly all of which are simply leftover implementation mechanisms from a previous system.

posted on 2005-09-19 15:23 ^ Mustang ^ 阅读(789) 评论(0)  编辑  收藏 所属分类: OO

只有注册用户登录后才能发表评论。


网站导航: