Templar：基于时态逻辑的软件规范语言解析

下载需积分: 5 | PDF格式 | 2.42MB | 更新于2024-07-09 | 143 浏览量 | 举报

"Templar是一种基于知识的软件规范语言，它使用时态逻辑来构建清晰易懂的需求规范。设计者Alexander Tuzhilin在纽约大学研发了这种语言，旨在实现以下目标：规范语法简洁明了，形式语义明确，便于系统分析师编写，同时也能让最终用户理解。此外，Templar的规范易于转换成多种设计规格。 Templar的核心是建立在时间逻辑和‘活动-事件-条件-活动’（Activity-Event-Condition-Activity, A-E-C-A）模型上的规则系统，这实际上是活动数据库中E-C-A模型的扩展。这种模型强调了时间维度，使得对系统行为的描述更为精确。语言包含丰富的建模元素，如规则、过程、时间逻辑操作符、事件、活动以及活动的层次分解。这些元素支持并行处理和决策机制，能够构建出结构紧密的系统。在Templar中，规则是核心组件，用于定义系统的运行时行为。它们结合了事件（表示系统状态的改变）、条件（用于检查特定情况是否满足）和活动（系统响应事件执行的操作）。时间逻辑运算符则提供了描述事件随时间发生顺序的能力，这对于描述软件系统随时间演变的行为至关重要。过程是可重复执行的单元，可以嵌套和组合，以创建更复杂的系统行为。活动的层次分解允许将大任务分解为可管理的小部分，提高规格的可读性和可维护性。并行性支持多个活动同时执行，反映了现代多线程和分布式系统的特点。通过决策机制，Templar可以表达在不同条件下的不同系统响应，增强了规范的灵活性和适应性。分类和主题描述符包括软件工程中的需求工程（D.2.1），表明Templar特别关注于软件开发的早期阶段，即需求的捕获和表述，以确保软件系统在设计和实现阶段就能准确反映用户的需求和期望。 Templar是一种强大的工具，它将时态逻辑的精确性和易理解性结合，提供了一种有效的途径来规范和验证软件系统的行为。它的设计考虑了实际开发和用户理解的需求，有助于提高软件质量，降低维护成本，并促进系统分析员与最终用户之间的沟通。"

展开

Templar: AKnowledge-Based Language .

275

RDL satisfies our third and fourth requirements for the same reasons as

TRIO does. However, RDL is less powerful than TRIO because it is based on

propositional logic, while TRIO is based on predicate logic. Furthermore, RDL

does not satisfy our first and second requirements for the same reasons as for

TRIO.

RML [Borgida et al. 1985; Greenspan 1984] is a requirements specification

language based on the object-oriented framework and multisorted first-order

logic. An RML specification consists of a set of interrelated object definitions.

RML distinguishes three types of objects, i.e., entity, activity, and assertion.

Also, RML supports time, but unlike TRIO, ERAE, INFOLOG, and RDL, it is

based on first-order rather than on temporal logic. Moreover, RML has a

formal semantics,

as described by Greenspan [1984].

However, RML does not fully satisfy some of the objectives stated in the

introduction. Greenspan tries to design RML so that specifiers could organize

knowledge in a natural and convenient fashion and make RML specifications

easily understood by end-users. Although he achieves his objective in many

respects, RML specifiers still have to use some encoding techniques in their

specifications. For example, the statement

“a new patient’s location after he

has been admitted is the ward

to which he is being admitted” [Borgida et al.

1985, p. 87] has to be rephrased as “at the end of an ADMIT event, the value

of the toWard property of the ADMIT event equals the value of the location

property of the patient being admitted” [Borgida

et al. 1985, p. 87, footnote]

and is expressed in RML as:

toWard of ADMIT at end(ADMIT) =

location of (newPatient of ADMIT at end(ADMIT)) at end(ADMIT)

We believe that this kind of statement would require less encoding if RML

were based on temporal logic, especially since the English statement has

conjunction “after” in it.1 Also, RML does not fully satisfy our third objective

because it is more difficult to map requirements specifications written in

RML into a broad range of design specification languages than for some of the

previously considered specification languages, such

as ERAE or TRIO, mainly

because RML supports

a wide range of knowledge representation primitives.

For example, it would be more difficult

to map RML requirements specifica-

tions into the design specification language Z [Spivey 1988] than it would be

to map TRIO specifications into Z. The reason for that is that object-oriented

constructs of RML are mapped into set-theoretic constructs of Z, and this

requires a paradigm shift from the rich knowledge representation world of

RML objects to the simpler world of Z values.

Telos [Myloupoulos et al. 1990] is an extension of RML and, therefore, is

also based on the object-oriented framework. Telos extends RML by improv-

ing RML facilities for representing and reasoning about temporal knowledge,

provides more general forms of generalization and classification abstractions

than RML does, supports linguistic extensions through the definition of

lSee Prior [1967] for an argument as to why temporal logic provides a more user-friendly

approach to time than the first-order 10gic.

ACM Transactions on Information Systems, Vol. 13, No 3, July 1995.

276 . Alexander Tuzhilin

metaattributes, and provides support for deductive rules and integrity con-

straints. All these features added to RML make Telos a powerful require-

ments specification lang-sage that is relatively easy for the specifier to use.

Also, Telos has a rigorously defined semantics.

However, Telos does not fully satisfy some of the objectives stated in the

introduction. First, it is more difficult to map requirements specifications

written in Telos into a broad range of design specification languages than for

some of the previously considered specification languages, such as ERAE or

TRIO, for the same reasons as it is for RML. Second, Telos does not fully

satisfy requirements (1) and (2) because the SA still has to do a

certain

amount of encoding. Although Telos supports Allen’s [1984] time interval

temporal logic, it does not support point-based operators of temporal logic, as

INFOLOG, ERAE, TRIO, and RDL do. For instance, the example in

Mylopoulos et al. [1990, p. 333] saying that an author cannot referee his own

paper is stated in Telos as:

Y / F’erson)(y = paper. author - =(3 t/ Time)y = paper. referee [at t])

This expression is contrasted with an equivalent temporal logic expression

that provides a more user-friendly treatment of time:

(V Y / Person)(y c paper. author = always_ in_the_future y @paper.referee)

Furthermore, Telos rules have the if-then structure and do not support

when, while, before, and after clauses. Without these clauses and without

the full support of temporal logic, Telos specifications require various encod-

ing techniques to specify end-user requirements involving time.

Tempera [Loucopoulos et al. 1990] is still another specification language

supporting time, complex objects, an extended entity-relationship (E-R) data

model, and deductive rules. As in Tel OS, it also represents a rich modeling

language. However, it also does not fully satisfy some of the requirements

stated in the introduction. The rule structure of Tempera supports temporal

logic, events, the when clause, and is based on the Event-Condition-Action

model of a rule [McBrien et al. 1991]. Therefore, it provides a more user-

-friendly rule structure than other languages considered so far. However,

Tempera supports only events and conditions and does not support activities.

This means that activities occurring over time require some type of encoding

in Tempera. For example, Tempera rules have to use some encoding tech-

niques to model statement “while an activity lasts . . . ,“ such as the one

presented in the introduction. Therefore, Tempera violates, to a certain

extent, our second objective because the end-user has to understand encoding

techniques used by the SA. Moreover, Tempera depends heavily on the E-R

data model and complex objects. This makes it more difficult to map require-

ments specifications written in Tempera into design specifications that use

other paradigms, such as the object-oriented paradigm (e.g., language TDL

[Borgida et al. 1993]),

than for such language as TRIO (because TRIO is

based only on temporal predicates that can be easily simulated in most of the

modeling paradigms). Therefore, Tempera specifications do not fully satisfy

our third objective to provide a specification language that can be mapped

into a broad range of design specification languages.

ACM Transac’uons on Information Systems, Vol. 13, No. 3, July 1995

剩余35页未读，继续阅读

身份认证购VIP最低享 7 折!

30元优惠券

weixin_38717171

粉丝: 1

Templar：基于时态逻辑的软件规范语言解析

最新资源