企業應用系統測試資料庫之研究

Abstract

軟體開發生命週期 (SDLC) 中，測試工作佔用了大部份的資源，因此測試 工作之完整實係一個系統良莠之分際。而在過去的測試過程中，多採用技 術導向，在各階段有不同標的重點的測試，過程中運用如白箱測試、黑箱 測試、灰箱測試、由上往下法及由下往上法的測試策略，其目的在追求軟 體與規格之一致性，只要軟體經測試後一如規格所定，即認為軟體為正確 無誤，此為過去的軟體導向開發策略。但經由學者對系統運作後所做的維 護工作的分析來看，使用者的知識 (user knowledge) 能解釋其中 60% 以上的差異，因此在系統開發過程中若能增加使用者的參與，提高使用者 與系統間的互動，將可使完成後的系統更接近使用者的需要。過去雖有從 雛型法、改善人機界面及教育訓練三方面著手，唯不論用何種方法，使用 者要能一窺系統之全貌、或要評估系統的執行績效，往往都需等到真實資 料陸續進到系統中才得竟全功，在時效上經常緩不濟急。因此本研究乃提 出一個資訊系統導向的開發策略，除達到以上的需求外，也希望能及早發 現在系統中隱藏的問題，以彌補過去軟體導向未週全考慮的缺憾。本研究 認為過去使用者之所以對測試工作持消極態度的主因在於使用者無良好的 工具參與測試工作，而且過去對測試的探討多屬技術層次，使用者無容置 喙。因此本研究乃提出一套程序方法，依據 E-R Model，分析將來在系統 中運作的真實資料、記錄其特性，而後藉助資料產生器，產生出仿真且足 量的測試資料庫，供系統在正式上線前測試功能之用。如此則不僅使整合 測試得以完整、驗收測試也可因使用者有所據而得以遂行，將有助於整個 資訊系統之完成。 Testing consumes most of the resources in the SDLC, so the quality of a software product is dependent largely on the completeness of testing. We adapted several known testing strategies in testing process like white-box testing, black-box testing, etc. which are so called technique orientation. The objective is to pursue the conformity between software and specification. We conceive software is certified as if the software tested is just the same as the specification defied. This was the software development strategy. In the past study of the maintenance components of the implemented system, user knowledge accounted for more than 60 percent of the common problem variance. If we can improve the user involvement in system development process, enhance the interaction between user and system, both make the system completed more near the need of end-user. We started from prototyping, human-machine interface, and training. No matter which method used, users cannot see through the whole features of the system, or evaluate the system performance until most of the real data loads into the system. It's always too late. So we proposed an information system orientation development strategy to fulfill the need above, in addition to discovering the problems hidden in system, to supplement the missing in software orientation. We found the reason why users passively participated in testing is they had no appropriate tools in hand, and all the issues were around technique, so unprofessional users are excluded. This paper proposed a set of procedures based on E-R model to analyze the characteristics of the real data later be loaded into the system, and by means of the generator to generate test data with fidelity as many as users want. We hope this procedure will be help to complete the system test and acceptance test, and information system wholelly.