Open Source Software: A Study of Dynamic Variance of Complexity

Open Source Software: A Study of Dynamic Variance of Complexity

Kanwaljit Singh,

Department of Computer Application.

ACET. Amritsar, India.

Hardeep Singh, Department of Comp. Sc. & Engg.

GNDU. Amritsar, India.

Abstract: Quality and complexity are closely related in software evolution cycle. Evolution period measures the qualities of the software. Software metrics monitor and manage the quality of software. Oscillation in the complexity reflects disparity in the quality. Complexity is one of the indicators of the software quality. Complexity depends upon the size of class, number of statements used and type of statement used in the software development. Video LAN Client (VLC) media player open source software with its 58 versions and 7-Zip open source compression software with 61 versions are used for the quality analysis. This paper measures the complexity of the open source softwares VLC and 7-Zip. The studies of the software complexity are done with comparative analysis on various factors generated by metric tool SourceMonitor. The overview is to calculate the active variation in complexity when compared with functions, class-size, and statements of VLC with 7-ZIP during evolution cycle.

Keywords: Open source software, structured complexity, metrics, evolution, SourceMonitor, VLC, 7-Zip.

1. INTRODUCTION

   Open source software (OSS) is freely available with its source code for study, research, download, modify, and share information [14]. Desirable modules are downloaded quickly from code library and adjusted in the source code of the software to improve its quality with less time and at low cost. This paper introduces two OSS; Video LAN Client (VLC) media player and 7-Zip compression software. VLC consists of clients and server to stream videos across the network. VLC is an open source modular design programmed in an object oriented C++. Incomplete, damaged or unfinished videos can be easily run on VLC.

   Fifty eight versions of VLC have been designed since evolution period of 2001 to 2013. Software evolution is the process of developing the software and then frequently renovating it for various reasons. In this paper evolution is a process of improvement, inheriting version and reformation. By the year Feb, 2001 VLC-0.2.0 was released with its properties and complexities. The evolution process starts from the time slot of Feb April 2001. The evolution in appendix-2 defines the enhancement in the software either by altering the code(s) of previous version as per customer requirements and quality satisfaction. The altitude of complexity gets affected with progression in evolution. Software is distinctly substantial if it fulfills the maximum of seven conditions of Lehmanns Law of

   evolution [2]. The study of all the versions are compared and evaluated with respect to meticulous metric tool. The metric tool will generate the result in two parameters, i.e. source code in the versions v/s complexity in the modules.

   7-Zip is an open source file archive, which may be used to compress and encrypt one or more files for various operating systems. 7-Zip is the conqueror of file archiving and compression tools. It sets the standard for both compression ratio and time with its very own 7z compression format. To compress a file, it manages to beat MagicRAR, WinRAR, and WinZip for the best compression ratio, even with its Fastest Compression setting enabled. 7-Zip takes only 25 seconds to compress target files/folders, WinRAR (44 sec), WinZip ( 51 sec), and MagicRAR (159 sec). Multinational banking, IT, and other organizations use 7-Zip software for compressing and encrypting the software files for data transfer and storage. Appendix-1 defines the enhancement in 7-Zip in its evolution cycle.

2. PROBLEM DEFINITION

   VLC and 7-Zip are open source software with various versions. After the development of one version, another version is ready to release with its own complex properties. This process continues in software evolution course. Each version of software develops with an integration of modules and classes. The statements are the tools used to measure the structure complexity of the software. Difficult and copious statements complicate the modules or classes and inflate the complexity. The quality of the OSS is calculated through complexity value. The organization of elements within the software defines complexity.

   The time gap between release date of 1st version and the final version of software should not be the enlarged. This reduces stability of the software. With the time period of 9 and 13 years more than 58 and 61 versions of VLC and 7- Zip have been developed [8]. 58 times in VLC and 61 times in 7-Zip numerous changes of the software will vibrate the market demand and quality of the software

3. METHODOLOGY

   To read and study the metrics value of the VLC & 7-Zip software, another OSS is executed called SourceMonitor. SourceMonitor is a metric tool that can calculate 14 metrics of java, C#, C++, VB based software with graphic indicator

   Files	Lines	State- ments	% Bnches	% Cmnts	Class Defs	Methods/ Class	Avg Stmnts/ Method	Max Cmplexty	Max Depth	Avg Depth	Avg Cmplxty	Functions
   387	105,153	35,481	19	25.3	37	4.5	8.9	126	9	1.7	4.21	1,396

   and filtering techniques to analyse the results [12]. The SourceMonitor (SM) scrutinizes how much code software has and identifies the relative complexity of statements and modules in software. VLC and 7-Zip are programmed in C++, and SM runs the software code at high speed, thousands of lines of code per second. SM has friendly graphic user interface (GUI). SM presents the metrics in form of tables or charts, to measure software in several phases of the development process and save the resultant metrics in checkpoints. SM helps to find out the changes in the software during the cycle of software evolution by using the Lehman Laws of Software Evolution. Table-1 is the result generated by the SourceMonitor with various attributes or metrics used to measure complexity. Selective metrics are used in this paper, as in Appendix 3, 4.

   Table 1: Study of various attributes by SourceMonitor

   • Files: Total number of files measured in the selected package.

   • Lines: Total number of lines in the selected package, without the blank lines at the end of each included file.

   • Statements: Total number of statements in the selected package.

   • % Branches: Statements such as if, else, for, while, goto, break, continue, switch, case, default and return are measured here as a percentage of the total statements.

   • % Comments: Number of total comments divided by total number of lines. Headers and footers, at the beginning and end of files are not taken into consideration.

   • Class Size: Total number of operations and attributes that are encapsulated in method or class.

   • Method per Class: Total number of complex methods in a class.

   • Functions: Total number of functions existing in the selected package.

   • Average statements/method: Total number of statements inside methods in a selected package divided by the number ofmethods in the package.

   • Maximum complexity: Value of the coupling and cohesion of the most complex function in the selected package.

   • Maximum depth: Maximum nested methods are depth in the selected package. At the beginning of each file its value is zero. It must be pointed out that statements at levels 1 to 8 are recorded, while statements at deeper levels are counted depth 9.

   • Average depth: Depth is total number of methods starting from root method to leaf method in the execution path. Average depth of software is sum of depth of all execution paths divided by total number of execution paths.

   • Average complexity: The average value of all complexity values in the selected package

   After measuring the metrics of VLC and 7-Zip software, the paper will perform the comparative analysis of the metrics of two softwares and find that during evolution period which metric(s) follow evolution laws.

4. RESULT AND ANALYSIS

Various versions of VLC and 7-Zip have been generated during the evolution period. Each version has hundred of packages in it. The individual version of the software is evaluated with SourceMonitor that generates attributes as shown in table-1. The data is generated, collected and analysed through all versions (shown in Appendix 3, 4).

Table-2: Correlation value of metrics in VLC and 7-Zip

In the study comparative analysis of the metrics of VLC and 7-Zip software has been done. Correlation among metrics is revealed in table-2. Using complexity metric, software team has the capability to indicate problems of software, guide software testing, and estimate software maintenance efforts [15]. Formats designed for comparative analysis in this paper are:

1. Average complexity vs. Functions

   Structure of the functions and their interrelation that are used to avoid statement redundancy describes the complexity of the software. Relational analysis classifies that the level of average complexity fluctuates with amendment in number of functions. The relationship is explained with the help of data collection for VLC and 7- Zip analysis in fig.1 and fig.2. There is a huge gap between complexity and function metrics values in VLC

   and 7-Zip. To recover, complexity value is multiplied by an independent constant value to make it relevant to represent two-dimensional graph of complexity and function.

   At the initial development in VLC, after the rise in functions at initial stage with fall in functions, there is equivalent fall in average complexity. There is fluctuation in function due to which average complexity descends. Where as in 7-ZIP application there is linear rise in functions and calm complexity at initial level. At the mid stage of evolution, there is strong boost in functions, due to which complexity level increases with low growth rate in VLC application. In 7-ZIP there is low density, high wavy shade enhancement in functions during evolution period at mid stage, due to which there is less frequency change in complexity. At the final stage, with the small rise in number of functions in VLC, there is narrow increment in complexity. In case of 7-ZIP application, at final stage with slight fall in number of functions there is increase in complexity. In table-2 the correlation among average complexity and function is positive in both the software but low in VLC and high in 7-Zip.

   Finally there are three parameters found in the relationship among functions and complexity;

   1. Sharp Increase in functions Gentle Rise in complexity.

   2. Aslant increase in functions Tilt Fall in complexity.

   3. Syrupy Rise or Fall in functions Calm complexity.

   12000

2. Class Size vs Average Complexity

   Class is the core of the object oriented program, and size of a class has mammoth credence on software output. The status of class structure is measured with the variable called complexity. Class size is calculated by the multiplication of method/class and statements/method. An analysis is being done on the relation between class size and the complexity in fig. 3. In VLC and 7-Zip, the complexity metric value is multiplied by an independent constant value to make it equivalent to class size for graph development. In VLC, at the early evolution stages, the complexity reduces as the class size remains same. With decrease in class size, average complexity also reduces. In the middle stage of evolution there is a steep fall in class size. With fall in class size, the complexity increases calmly. At the last mode of evolution, there is constant flow of class size. In constant mode of class size, the complexity reduces. This clarifies that the number of statements used in methods at different class remain same but the format of the statements varies.

   As compare to VLC in fig.3, 7-Zip application in fig.4 has very narrow variation in complexity with rise in evolution. At the early stage with rise in class size, there is rise is complexity. This results in directs relations. At the mid of 7-Zip evolution period, the complexity increases steadily with bit by bit increase in class size. The complexity gets consistent with change in class size. At the last stage, with minor rise in size, there is narrow increase in complexity. Table2 calculates the correlation among class size and

   10000

   8000

   6000

   4000

   2000

   0

   VLC

   Early Stage

   Middle Stage

   Last Stage

   Functions

   AvgCmplx

   average complexity, it is ve in VLC and +ve in 7-Zip software. The various parameters found in the relationship among class size and complexity are:

   1. Sharp Boost in class size – Minor fall in complexity

   2. Fall in class size Angled increase in complexity

   3. Calm in class size In control complexity

3. Statement vs Average complexity

The instruction processed by compiler is called statement. Set of statements is called method. Program is collection of statements, functions and classes. Set of programs develop

1 4 7 101316192225283134374043464952555

Fig1: Relation among Complexity and Function in VLC

the software. Statements are the core of the software. Line graph explains the relation between number of statement and complexity level. The complexity metric values of

VLC and 7-Zip software are multiplied by different independent constant values to make them equivalent to

1600

1400

1200

1000

800

600

400

200

0

7-Zip

Early Stage

Mid Stage Last Stage

Avg Complexity

Functions

their corresponding statement metric value.

At the initial evolution development stage of VLC the complexity decreases with increase in number of statements as in fig.5. At middle stage, a sky scraper is generated by number of statements, and a bit increase in complexity.

1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61

Fig2: Relation among Complexity and Function in 7-Zip

10000

8000

6000

4000

Initial Stage

Middle Stage