A Novel Methodology for Test Scenario Generation From UML 2.x Sequence Diagrams

Transcription

1 A Novel Methodology for Test Scenario Generation From UML 2.x Sequence Diagrams Saroj Kanta Misra Dept. of Computer Science & Engineering, Gandhi Institute of Engineering & technology, Gunupur Odisha , India Suvendu Sekhar Dash Dept. of Computer Science & Engineering, Gandhi Institute of Engineering & technology, Gunupur Odisha , India ABSTRACT Now a days UML is widely used for preparing design documents. It helps to specify, construct, visualize and document artifacts of software systems. This paper presents a novel approach to test the software in the earlier stage (design phase) of software development life cycle, so that it can help the software testers in the later stages. This paper focuses on generating test scenarios from UML 2.0 Sequence diagrams. The most challenging problem in generating test scenarios from UML 2.0 sequence diagram is the presence of fragments such as alt, loop, break, par, opt etc. We propose an intermediate control flow graph in a testable form named Sequence Control Flow Graph (SCFG) resulting from the control flow analysis of UML 2.0 sequence diagrams. We also propose a systematic approach named Sequence Test Scenario Generation Algorithm (STSGA) for generating test scenarios from UML 2.0 Sequence diagrams. The test scenarios generated by our approach are suitable for detection of scenario faults, use case dependency and system testing. General Terms Verification D.2.8 [Software Engineering]: Test Scenario Generation UML 2.x, Software Testing Keywords Test scenarios, UML 2.0 Sequence diagram, Sequence Con- Categories and Subject Descriptors H.4 [Information Systems Applications]: Miscellaneous; Saroj Kanta Misra completed his M.Tech from National institute of technology, Rourkela and completed his B.Tech from Veer Surendra Sai University of Technology, Odisha, India. He joined as a faculty of the Department of Computer Science and Engineering at Gandi institute of Engineering snd Technology, Gunupur in 2014, where he is now working as Assistant Professor. His special fields of interest include Operating System, Computer Architecture, Software Engineering, Software Testing. Suvendu Sekhar Dash completed his M.Tech from Biju Pattnaik University of Technology, Odisha, India.He joined as a faculty of the Department of Computer Science and Engineering at Gandi institute of Engineering and Technology, Gunupur in 2011, where he is now working as a Lecturer. His special fields of interest include Computer Security, Soft Computing, Software Testing. 1. INTRODUCTION Nowadays, Due to rapid increase of size and complexity software applications, more emphasis is given towards object-oriented design strategy, which helps to reduce software cost and increase software reliability, software usability. But in turn introduction of object-oriented design and implementation approach brings out some new difficulties for software testing. Several features of object-oriented approach like polymorphism, dynamic binding, inheritance etc. create certain difficulties in software testing process. To test such object oriented softwares from their implementation code is a very a complex process due to the different features of object oriented approach. Model Based Testing of these object oriented softwares can be beneficial to detect the error in the design phase itself, so that these error do not propagate to other stages of software development life cycle. Control Flow Analysis (CFA) plays a vital role in determining all possible alternative paths a program may fallow during execution. A Control Flow Graph (CFG) is a static representation of a program that represents all alternatives of control flow. For example, both choices for If else statement can be represented in CFG as different control flow paths for both choices. A Loop can be represented as a cycle in a CFG. According to Garousi et al. [1] Control flow information can be derived from two different sources: from software design artifacts and code itself. In Code-based CFA(CBCFA) control flow information is obtained from the available source code. where, in Model-based CFA(MBCFA) control from information is obtained from design models (such as UML). The motivation of our work is to derive control flow informa- 371

2 tion and generate test cases in the earlier stage of software development life cycle, after the UML design models of a system becomes available. 2. BASIC CONCEPTS AND DEFINITIONS UML behavioral diagrams such as interaction diagram, activity diagram, state machine diagram describes different functionality of the system and also captures various dynamic behaviors of the system. Interaction diagram illustrates the system functionalities using different fragments such as alternative, loop, break, parallel, etc. In this section we provide a overview of interaction diagram, XMI, Sequence Control Flow Graph (SCFG), how to obtain a SCFG. 2.1 UML 2.0 Interaction Diagram Interaction diagrams describes how a group of objects collaborate in some behavior - typically a single use-case. Interaction diagrams is of two types: Sequence diagram and communication diagram. The basic objective of the diagrams are same. Sequence diagram accentuate on the time sequence of messages passed between communicating objects, and the communication diagram accentuate on the structural organization of the communicating objects that send and receive messages. In our approach, we have used UML 2.0 Sequence diagram to generate test scenarios New Features of UML 2.0 sequence diagram List belows describes some of the feature of UML 2.O Sequence diagram Interaction: Series of messages that are passed between different communicating objects to satisfy some task is called interaction. Interaction Occurrences: When the interactions are used within another interaction or context is called interaction occurrence. Combined Fragments: Combined fragment is an interaction fragment which is a combination of multiple interaction fragments. Each combined fragments has a interaction operator and corresponding interaction operands. Interaction Operand: Interaction Operand shows Grouping of interactions within the combined fragment. Interaction Operator: Interaction Operator for a combined fragments describes how the interaction operands present inside the combined fragments are going to be used. The fallowing are the list of interaction operators defined. Alternative (alt): The interaction operator alt works like if then else structure. At most one operand can be selected based on the guard expression true value. If there is no guard expression then a implicit true guard value is implied. Option (opt): The interaction operator opt used, when the combined fragment represents the operand as a option where the operand is either happens or may not happens. It works similar like a alternative combined fragment where there is one operand is nonempty and the other one is empty. Loop (loop): The interaction operator loop represents a loop structure. The interaction operand present inside the loop combined fragment will be repeated many times. The repetition of loop can be controlled either or both by iteration bound and guard. If a loop combined fragment has no bound specified, then the loop will execute with infinite as upper bound and zero as lower bound. Break (break): The interaction operator break used to represent a breaking or exceptional scenario to be performed instead of the remaining interaction fragment. Parallel (par): The interaction operator par describes parallel execution of behaviors of the interaction operands present inside a combined fragment. 2.2 XMI XMI (XML Metadata Interchange), is a extension of XML that facilitate the standardized way for interchanging object models and metadata. Specifically, XMI is useful programmers using the Unified Modeling Language (UML) with various languages and development tools to exchange their data models with each other. 2.3 Sequence Control Flow Grapg (SCFG) In order to visualize/animate control flow of a sequence diagram, the information contained in the XMI equivalent of the Sequence diagram, is extracted to construct a graph called Sequence Control Flow Graph(SCFG). The fallowing nodes are considered while mapping a Sequence diagram ito a Sequence Control Flow Graph(SCFG). Definition : An Sequence Control Flow Graph is a tuple I ={I,M,F sn,f en,e op,f n} where, I is the starting node of the Sequence Control Graph(SCFG). M is a message node representing a sequence of messages in sequential. F sn is set of nodes representing the beginning of a combined fragments. F en is set of nodes representing the End of a combined fragments. E op is the set of nodes that represents, the next nod to it is a message from nonuser object to user. F n is the set of final nodes represents exit of Sequence Control Flow Graph (SCFG). 2.4 Obtaining the Sequence Control Flow Graph (SCFG) The first step of our approach is to generate a SCFG from the UML sequence diagram. We obtain a SCFG for representing control flow between messages in presence of fragments and nested fragments. In the process we give more emphasis on use scenarios (actions executed by the user and actions viewed by the user). Each message present in sequence diagram is represented by a node in SCFG. We have added some additional nodes for the sake of simplicity in the process generating test scenarios [6].The begining and at the ending of every fragment is denoted by two additional 372

3 Figure 1: Sequence diagram example nodes representing starting and ending of fragment such as alt start, alt end, par start, par end etc. In alt fragment the conditions for control flow is also denoted by additional nodes containing a condition sequence number and condition itself such as condition1 true, condition2 false. The steps to build a SCFG a are presented as fallow. Figure 2: Figure 1 Derived SCFG for sequence diagram in The stating and ending of SCFG is represented by two additional nodes namely start, end. From the initial node start, for each message in sequence diagram is a new node added with into the SCFG with thier value same as message name in sequence diagram. For each message (in order it appearing sequence diagram) if the message is: 1. From user object to non-user object or from nonuser object to non-user object then a new node is added into SCFG with a directed edge from its previous node to itself. 2. From non-user object to user object then two nodes are added into SCFG (1) first node with value expected output and a directed edge from its previos node to itself. (2) Second node with value same as message name and a directed edge from expected output node to itself. 3. PROPOSED METHODOLOGY In this section we disscuss our approach for test case generation from UML 2.x Sequence diagram.our sequence diagram includes combined Fragments using various Intraction Operators such as alt,par,loop,break etc [6]. We propose a mechanism to extract the messages from fragments and combined Fragments precisely. We categorize the messages into two different catagoies,(1) Messages from non user Object to user object.(2) Messages from user Object to non user Object and messages between non user Object. Then we map every message to thair Corresponding fragment.then we generate the Sequence Control Flow Graph (SCFG) Using Sequence Cntrol Flow Generator. The block diagram for generating test case from XMI representation of sequence diagram is given in Figure 1. The major steps of our approach are partitioned into three phses.they are given below: 1. Parsing the XMI representation of UML 2.x Sequence Diagram. 2. Sequence Control Flow Graph Generator 3. Test case Generator The first step in our approach is to parse the XMI representation of UML 2.x sequence diagram.we have used IBM Rational Rose Architecture(RSA) to draw the sequence diagram, and then we have exported the XMI representation 373

4 Figure 3: Block diagram For generating test case from XMI representation of UML 2.x sequence diagram for the sequence diagram, which is used as a input for our process. We propose a parser that parse the XMI file to extract information about messages, structure of fragments and combined fragments. Messages that are sent from any non user object to user object are identified. Using all the extracted information a Sequence Control Flow Graph is generated. In order to generate test case for the sequence diagram the SCFG need to be traversed. We propose a Sequence Test Scenario Generation Algorithm (STSGA), the detailed algorithm is shown in Algorithm 1. This algorithm traverses the Sequence Control Flow Graph(SCFG) to first identify the the messages that are from non user object to user object. Generates test paths from start node to these message and then finally generate test cases for these messages. 4. IMPLEMENTATION OF OUR ALGORITHM Now in this section,we exemplify our approach for generating test scenario form XMI representation of UML sequence diagram by converting XMI representation of UML sequence diagram into a equivalent Sequence Control Flow Graph (SCFG) and then generating test scenarios. For generating test scenarios from UML sequence diagram to test the feasibility and concurrency errors. We have proposed Sequence Test Scenario Generation Algorithm (STSGA) for generating test scenarios for each message from non user object to user object by taking the Sequence Control Graph (SCFG) as input. We have developed a prototype tool called XMI2SCFG (XMI to Sequence Control Flow Graph) for generating test scenarios. XMI2SCFG works in two steps (1) Parsing of XMI representation of UML Sequence diagram.(2) Create a SCFG (Sequence Control Flow Graph) in image format, and generate test scenario from SCFG h. We have implemented XMI2SCFG in java language using Netbeans IDE XMI2SCFG takes the XMI representation of UML 2.x sequence diagram. We have used IBM Rational Software Architecture (RSA) 7.0 to draw the sequence diagram and then exported the XMI representation (XMI equivalent of UML sequence diagram). In the first phase, XMI2SCFG uses SAX parser to parse the XMI representation of sequence diagram. Along with the main class and sequenceparser some auxilary classes listmsg,listalt,listpar,listloop,listbreak are used for this propose. The sequenceparser class implements various methods such as getmsgid(), getmsgname(), getusermsg(), getaltmsg(), getbreakmsg(), getparmsg(), getloopmsg() to interact with SAX parser. These methods process the tagged elements present in XMI representation of UML sequence diagram such as packagedelement, message, fragment, ownedattribute, lifeline, operand, guard, body, ownedoperation to extract various informations like message name, message flow, message dependacies, message type, guard condition, etc. In the second phase,the main tast of XMI2SCFG to visualize the SCFG (Sequence Control Flow Graph) in an image format. Fror SCFG visualization two main classes are used : DotTransformer and Graphvizvisualization. We have used Graphviz. Taking two linked list transource and trandestination as input, the DotTransFormer objet creates a.dot file. After the.dot file is created, the methods present in graphiviz getdotsource(), getgraph(), and writegraph- Tofile() create an image for SCFG (Sequence Control Flow Graph). Then the SCFG is supplied as input to our algorithm STSGA which generates the test scenarios. 4.1 CASE STUDY In this section, we illustrate our approach for generating SCFG and test scenario with the help of a case study pertaining to College Automation System (CAS). The CAS automates various functionalities of a college such as student enrollment, register Student, update Attendance, recive fee,generate recipt etc. Here, we take a particular use case, namely, view Mark use case, a student first needs to get registered in order to view his/her marks.once the student registered he will be provided with a student Id. By using the student ID and name, a student can get his marks. The sequence diagram of view marks is given in Figure 2. The Sequence Control Flow Graph (SCFG) for generating test scenario derived from sequence diagram can be seen in Figure 3. The SCFG is obtained by the transformation procedure presented in section 4. After that, test scenarios are obtained by supplying SCFG as input to our algorithm, STSGA (Sequence Test Scenario Generation Algorithm).Then the generated test scenarios are showen in figure 5. By Analysis of the result, we can observe that our approach generates all test scenarios for each message that is sent from a non user object to user object coverage criteria chosen. 5. COMPARISON WITH RELATED WORK In this section, we provide a comparison of our work with other sequence diagram based test scenario generation methods. In many cases test scenario generation is done manually, as in case of many related work. For large system, it is practically impossible to generate test scenrios from sequence diagram. Sarma et al. [2] proposed an approach to generate test scenarios from uml sequence diagram, by converting sequence 374

5 Algorithm 1 Sequence Test Scenario Generation Algorithm (STSGA) Input: Sequence Control Flow Graph (SCFG). Output: Set of test scenario. 1: runningstack=φ 2: decisionstack=φ 3: resultstack=scfg.rootnode 4: for all nodes of SCFG do 5: while SCFG.node==expextedOutput.node do 6: usermessagestack.push(child node of SCFG.node) 7: end while 8: end for 9: for all elements of usermessagestack do 10: repeat 11: if runningstack[top] alt.node loop.node par.node break.node SCFG.EndNode then 12: resultstack.push(runningstack.pop) push runningstack top element in resultstack and pop the top element from runningstack 13: resultstack.push(child node of resultstack[top] in SCFG) push child node of resultstack top element into resultstack 14: else if runningstack.[top] == SCFG.EndNode then 15: Mark the last decision node in resultstack as visitd 16: while resultstack[top] decisionstack[top] do 17: resultstack.pop pop the top element fom resultstack 18: end while 19: decisionstack.pop Pop the top element from decisionstack 20: else if runningstack[top] == alt.node break.node loop.node then 21: decisionstack.push(runningstack.pop) pop top element of runningstack and push it into decisionstack and resultstack 22: 23: for all Child nodes of resultstack[top] in SCFG do 24: if Child node is not Marked Visited then runningstack.push(child Node) push all child nodes of resultstack[top] in SCFG, if marked as visited insert into runningstack 25: end if 26: end for 27: else if runningstack[top] == par.node then 28: resultstack.push(resultstack[top] and all its child nodes in SCFG) 29: runningstack.push(child nodes of resultstack[top] in SCFG) 30: else if runningstack.top == UserMessageStack.CurrentNode then 31: resultstack.push(runningstack.pop) Pop the top element from runningstack and push it into resultstack 32: Print resultstack 33: if decisionstack φ then 34: while resultstack[top] decisionstack[top] do 35: resultstack.pop pop the top element fom resultstack 36: end while 37: end if 38: if decisionstack φ then decisionstack.pop pop the top element from decisionstack 39: end if 40: end if 41: until runntingstack φ & decisionstack φ 42: delete all elements from resultstack 43: delete all elements from decisionstack 44: delete all elements from runningstack 45: end for 46: Exit diagram into an directed graph called Sequence Diagram Graph (SDG), where nodes in SDG represent a message in sequence diagram and a directed edge represent control folw between nodes. SDG is then used to generate test scenarios. The difference between Sarma et al. work and our work is as fallows. Sarma et al. used UML 1.x sequence diagram for their work, which did not support fragments such as alt, par, loop, break etc whereas our approach considered these fragments by using UML 2.x sequence diagram. In our work 375

6 Figure 4: Sequence Diagram of View Marks 376

7 Figure 5: Sequence Diagram of View Marks we have generated a directed graph called sequence Control Flow Graph(SCFG), and then test scenarios are generated from SCFG automatcally. Cartaxo et al. [3] proposed a model a approach to generate test paths for mobile application using sequence diagram. They, constructed a intermediate model call Labeled Transition System (LTS) from sequence diagram, where directed edges are used to represent control flow, expected output and steps. Then, they applly a depth first search (DFS) algorithm to traverse the LTS model to generate test paths. However their approach did not support fragments present in UML 2.x sequence diagram. Khandai et al. [4] proposed an approach to generate test cases from sequence diagram. They have used a intermediate graph called Concurrent Composite Graph (CCG) generated from sequence diagram, which is a variant of activity diagram. Then they traverse the CCG by applying depth first search (DFS) and breath first search (BFS) to generate test cases. BFS algorithm is used between fork and join construct to explore all concurrent nodes. We proposed a novel testing methodology to object-oriented software based on UML Sequence Diagrams. Our approach is a systematic approach to automatic test scenario generation from UML 2.x Sequence Diagram, which supports various fragment types. The test scenarios generated are suitable for system testing and to detect operational, intraction faults. 6. CONCLUSION Existing test scenario generation techniques from UML diagrams limit their applicability on specifying and transforming a single path generation information about the test scanario. They do not consider fragments and nesting of fragments into test scenarios. Their techniques can be automated as their input incorporates statements. However, their techniques are becomes more complex for UML 2.x sequence diagrams. Our approach for test scenario generation based on the on the basic control flow primitives of UML 2.X sequence diagram.our approach is a systematic interpretation of control flow for various fragments as well as nested fragments presents in UML 2.x sequence diagram. We also proposed a directed graph presenting all control flow primitive resulting from UML 2.x sequence diagram. The test scenarios thus generated are suitable for system testing and detect operational, use case dependency, interaction and scenario faults. 7. REFERENCES [1] Garousi, Vahid, Lionel C. Briand, and Yvan Labiche. Control flow analysis of UML 2.0 sequence diagrams. In Model Driven ArchitectureÂŰFoundations and Applications, pp Springer Berlin Heidelberg, [2] Sarma, Monalisa, Debasish Kundu, and Rajib Mall. Automatic test case generation from UML sequence diagram. In Advanced Computing and Communications. ADCOM International Conference on, pp IEEE, [3] Cartaxo, Emanuela G., Francisco G. Oliveira Neto, and Patricia DL Machado. Test case generation by means of UML sequence diagrams and labeled transition systems. In Systems, Man and Cybernetics. ISIC. IEEE International Conference on, pp IEEE,

8 [4] Khandai, Monalisha, Arup Abhinna Acharya, and Durga Prasad Mohapatra. A novel approach of test case generation for concurrent systems using UML Sequence Diagram. In Electronics Computer Technology (ICECT), 3rd International Conference on, Vol. 1, pp IEEE, [5] Mall, Rajib. An approach to convert XMI representation of UML 2. x interaction diagram into control flow graph. ISRN Software Engineering 2012 (2012). [6] Misra, Saroj Kanta. A novel methodology for test scenario generation based on control flow analysis of uml 2.x sequence diagrams. ICICT