CN103729297A - Test case generation method based on hierarchic genetic algorithm - Google Patents

Abstract

The invention provides a test case generation method based on a hierarchic genetic algorithm. The method can be used for effectively preventing and controlling inbreeding, so that the problems of premature convergence and local convergence are effectively solved. The method includes the steps that a program source code is inputted into a path analyzer, and the path analyzer carries out path analysis on the inputted program source code, then obtains a corresponding path list and outputs the path list to a test data generator; a predicate instrumenter carries out predicate instrumentation on the inputted source code and outputs the instrumented code to the test data generator; the test data generator firstly carries out hierarchy on an initial population through the hierarchic genetic algorithm, compares passing paths of individuals in hierarchies with objective paths of the individuals in the hierarchies through the utilization of the instrumented source code to obtain the matching values between the passing paths and the objective paths, and finally the test data generator gives test cases to the paths.

Description

A kind of method for generating test case based on hierarchical genetic algorithms

Technical field

The present invention relates to a kind of method for generating test case based on hierarchical genetic algorithms, belong to software testing technology field.

Background technology

Along with the develop rapidly of computer technology science, as the soul of computing machine, software plays a part very important, and the inefficacy of software may cause even threat to life safety of huge economic loss.People are more and more deep to the understanding of software test importance, and software test stage shared proportion in whole software development cycle increases day by day.Whether existing software testing technology, from needing the angle that is performed tested software to be conventionally divided into static test and dynamic test.Static test is not executive routine code and the mistake that may exist in finder code or the process of appraisal procedure code.Dynamic test is come the dynamic behaviour of check program and operation result to find mistake by working procedure in sample testing data.Dynamic test comprises three part core contents: the operation result of generating test use case, working procedure and proving program.

In test data robotization generation field, comparatively commonly symbol is carried out method, stipulations produce random number method, Korel method etc. according to demand at present.Symbol execution method is static test data generating method, for dynamic object-oriented program, is not suitable for using; By requirements specification, produce random number method and belong to dynamic approach, when adopting test case sample testing, this method is very effective to producing test data, but in the higher method of testing of constraint requirements, such as branch covers or path covering, its efficiency is very low, is difficult to produce satisfactory disaggregation; The Korel dynamic test data method of formation covering in path-oriented owned by France, the information of its branch function during according to program dynamic operation, constantly adjusts the value to the influential input variable of current branch function, and branch function is minimized.But shortcoming is if primary data is selected badly, is easily absorbed in the locally optimal solution in carry out desired path surely that differs.

The existing automatic example generation method based on genetic algorithm, mainly to carry out different improvement for corresponding fitness function and coding method, for example to the improvement of fitness function, can make the evolution in genetic algorithm more accurate, be exactly fitness function because eliminate individual main standard.Appropriate coding method simultaneously also can make follow-up search procedure more quick.Although these improve in test case generates, there is raising in various degree, but still some problems of ubiquity are the problem of " precocity " and local convergence.

Summary of the invention

The present invention is a kind of method for generating test case based on hierarchical genetic algorithms, energy effectively preventing " inbreeding " phenomenon, thereby effectively solved the problem of " precocity " and local convergence, increased the diversity of species simultaneously, be convenient to produce the test case that coverage is higher.

Realize technical scheme of the present invention as follows:

Based on a method for generating test case for hierarchical genetic algorithms, comprise the following steps:

Step 1: give path analysis device part loading routine source code, then path analysis device carries out path analysis to the program source code of input, obtains corresponding path list after analyzing, and path list is exported to test data generating;

Step 2: predicate instrumenter part is carried out predicate plug-in mounting to the source code of input, first instrumenter detects the source code of input, once detect that branch statement and loop statement just insert corresponding predicate after branch statement and loop statement, until predicate instrumenter detects all branches predicate and circulation predicate place, all plug-in mounting is complete, and the code after plug-in mounting is exported to test data generating;

Step 3: the fitness function that mainly utilizes hierarchical genetic algorithms and Huffman encoding in test data generating, first hierarchical genetic algorithms carries out layering to initial population, in layer, make a variation with interlayer, intersect, select genetic manipulation, the fitness function of Huffman encoding builds and utilizes the source code that plug-in mounting is good to compare with destination path by path individual in layer, obtain by the matching value of path and destination path, this matching value is posted the fitness value into this heredity individuality, then according to the rule of elimination of setting, compile, intersect, select genetic manipulation, last test Data Generator provides the test case for every paths.

The step that hierarchical genetic algorithms hierarchical block described in step 3 is concrete is as follows:

A. first produce random initial population GA;

B. initial population is carried out to layering and be respectively GA111, GA112, GA121, GA122, GA211, GA212, GA221, GA222;

C. set genetic algebra counting variable t=0, Group Hierarchical counter m=3;

D. first calculate the individual fitness GA(t in t generation the 3rd layer of GA population), judge whether to satisfy condition, meet just output as test case; Do not meet and enter step e;

E. between the sub-population of the 3rd layer, select, variation, interlace operation, and then calculate fitness function value, see and whether satisfy condition, if meet, export as test case, otherwise m-1, in the second layer, select again, variation, interlace operation;

If f. still do not satisfy condition at m=1 layer fitness function, counter adds 1 and enters into the second generation, and then repeating step d, till so circulation obtains the test case in coverage goal path always.

In the fitness function module of the Huffman encoding described in step 3, remember that t is x for the individuality of Advanced group species, x is the input data of tested program, and the path string that x passes through is designated as f (x), and the length of f (x) is designated as | f (x) |, total n bar path to be measured, f _jrepresent j article of path to be measured, wherein j=1,2 ..., n, by calculating f (x) and f _jmatching degree obtain the fitness function P (x) of x, specific formula for calculation is as follows:

P _j(x)＝(C _j(f(x)))+1)*P _j′(x);

$P_j^{\′}\left(x\right)=\underset{k=1}{\overset{\min \left(\right|f\left(x\right),\left|f_j\right|)}{\mathrm{\Σ}}}{m}_{\mathrm{jk}}*d_{\mathrm{jk}}\left(f\left(x\right)\right);$

Wherein: d _jk: reflection f (x) and f _jk position whether identical, if identical value is 1, otherwise be 0;

C _j(f (x)): since first consecutive identical figure place;

M _jk: f (x) and f during to k position _jthe number of identical bits;

By calculating P _j(x) we just can obtain f _jmatching degree with f (x).

Compared with existing best technique, the invention has the advantages that:

1. the present invention has realized the automatic generation of test case, and the program that is suitably for source code rank generates corresponding test case, and utilizes hierarchical genetic algorithms to have speed and restrain preferably.

2. the framework of general frame of the present invention is simple, and stability is high, and execution efficiency is high;

3. the present invention has realized the model of a test case Auto-Generation Tool.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of a kind of dirty data travel path discover method based on complex network of the present invention;

Fig. 2 is that the present invention collects, the procedure chart of simplification and visual track path;

Fig. 3 is the function call result schematic diagram of application program in the embodiment of the present invention;

Fig. 4 is Huffman coding schematic diagram of the present invention.

Embodiment

Below in conjunction with accompanying drawing, further the present invention is described in detail.

In dirty data travel path discover method based on complex network of the present invention, mainly comprise path analysis device, predicate instrumenter, three parts of test data generating.First the code of the source program of path analysis device to input is analyzed, thereby obtains the routing table of source code program.Predicate instrumenter carries out predicate plug-in mounting to the source code of input simultaneously, code after plug-in mounting and path list are exported to test data automatic generator simultaneously, in test data generating, mainly utilize the fitness function of hierarchical genetic algorithms and Huffman encoding, first hierarchical genetic algorithms carries out layering to initial population, makes a variation, intersects, the genetic manipulation such as selection in layer with interlayer.And the structure of the fitness function of Huffman encoding is to utilize plug-in mounting good source code carries out.Individuality in layer has been passed through to path compares with destination path, thereby obtain by the matching value of path and destination path, this matching value is posted the fitness value into this heredity individuality, then according to the rule of elimination of setting compile, intersect, the genetic manipulation such as selection, last test Data Generator provides the test case for every paths.

Below in conjunction with Fig. 1, describe workflow of the present invention in detail:

Step 1: give path analysis device part input source code, then path analysis device carries out path analysis to the program source code of input, obtains corresponding path list after analyzing, and path list is exported to test data generating.

Step 2: predicate instrumenter part is carried out predicate plug-in mounting to the source code of input, plug-in mounting is mainly to carry out at the branch function of program and cyclical function place, first instrumenter detects the source code of input, once detect branch statement and loop statement just branch statement and loop statement below insert corresponding predicate, until predicate instrumenter detects all branches predicate and circulation predicate place, all plug-in mounting is complete.Code after plug-in mounting is exported to test data generating and is used.

Step 3: test data generating part mainly comprises the fitness function design of hierarchical genetic algorithms layering and Huffman encoding.Wherein the layering of hierarchical genetic algorithms we take points three layers as example, each individuality divides two to carry out as shown in Figure 2 in every layer.Ground floor in Fig. 2 has 2 sub-populations, and every sub-population has again 2 sub-populations at the 2nd layer, and the second layer has 2 × 2 sub-populations,, every sub-population of the 2nd layer has again 2 sub-populations at the 3rd layer, and such the 3rd layer has 2 × 2 × 2 sub-populations.By that analogy, if ground floor has K1 sub-population, the sub-population in n-1 layer has Kn sub-population at n layer, the total sub-population of n layer.

Wherein as shown in Figure 3, its concrete step is as follows for the overall flow of the hierarchical genetic algorithms in test data generating:

A. first produce random initial population GA.

B. initial population is carried out to layering and be respectively GA111, GA112, GA121, GA122, GA211, GA212, GA221, GA222.

C. set genetic algebra counting variable t=0, Group Hierarchical counting technology device m=3, to divide three layers as example;

D. first calculate the individual fitness GA(t in t generation the 3rd layer of GA population), judge whether to satisfy condition, meet just output as test case.

If e. step 4 do not meet between the sub-population of the 3rd layer, select, variation, interlace operation, and then calculate fitness function value, and see and whether satisfy condition, if meet, export as test case, otherwise m-1 selects, variation, interlace operation in the second layer again.

If f. still do not satisfy condition at m=1 layer fitness function, algebraically counter adds 1 and enters into the second generation, and then repeating step 4, till so circulation obtains the test case in coverage goal path always.

The fitness function of the Huffman path code adopting designs as shown in Figure 4, Huffman encoding is mainly to be encoded in path, thereby the binary code string list that has obtained every paths by path being carried out to binary coding shows so that we carry out the structure of fitness function.

In the fitness function module of described Huffman encoding, remember that t is x for the individuality of Advanced group species, x is the input data of tested program, and the path string that x passes through is designated as f (x), and the length of f (x) is designated as | f (x) |, total n bar path to be measured, f _jrepresent j article of path to be measured, wherein j=1,2 ..., n, by calculating f (x) and f _jmatching degree obtain the fitness function P (x) of x, specific formula for calculation is as follows:

P _j(x)＝(C _j(f(x)))+1)*P _j′(x);

$P_j^{\′}\left(x\right)=\underset{k=1}{\overset{\min \left(\right|f\left(x\right),\left|f_j\right|)}{\mathrm{\Σ}}}{m}_{\mathrm{jk}}*d_{\mathrm{jk}}\left(f\left(x\right)\right);$

Wherein: d _jk: reflection f (x) and f _jk position whether identical, if identical value is 1, otherwise be 0;

C _j(f (x)): since first consecutive identical figure place;

M _jk: f (x) and f during to k position _jthe number of identical bits;

By calculating P _j(x) we just can obtain f _jmatching degree with f (x).

Although combine accompanying drawing, the specific embodiment of the present invention has been described; but to those skilled in the art; under the premise without departing from the principles of the invention, can also make some distortion, replacement and improvement, these also should be considered as belonging to protection scope of the present invention.

Claims (3)

1. the method for generating test case based on hierarchical genetic algorithms, is characterized in that, comprises the following steps:

Step 1: give path analysis device part loading routine source code, then path analysis device carries out path analysis to the program source code of input, obtains corresponding path list after analyzing, and path list is exported to test data generating;

Step 2: predicate instrumenter part is carried out predicate plug-in mounting to the source code of input, first instrumenter detects the source code of input, once detect that branch statement and loop statement just insert corresponding predicate after branch statement and loop statement, until predicate instrumenter detects all branches predicate and circulation predicate place, all plug-in mounting is complete, and the code after plug-in mounting is exported to test data generating;

Step 3: the fitness function that mainly utilizes hierarchical genetic algorithms and Huffman encoding in test data generating, first hierarchical genetic algorithms carries out layering to initial population, in layer, make a variation with interlayer, intersect, select genetic manipulation, the fitness function of Huffman encoding builds and utilizes the source code that plug-in mounting is good to compare with destination path by path individual in layer, obtain by the matching value of path and destination path, this matching value is posted the fitness value into this heredity individuality, then according to the rule of elimination of setting, compile, intersect, select genetic manipulation, last test Data Generator provides the test case for every paths.

2. a kind of method for generating test case based on hierarchical genetic algorithms as claimed in claim 1, is characterized in that, the step that the hierarchical genetic algorithms hierarchical block described in step 3 is concrete is as follows:

A. first produce random initial population GA;

B. initial population is carried out to layering and be respectively GA111, GA112, GA121, GA122, GA211, GA212, GA221, GA222;

C. set genetic algebra counting variable t=0, Group Hierarchical counter m=3;

D. first calculate the individual fitness GA(t in t generation the 3rd layer of GA population), judge whether to satisfy condition, meet just output as test case; Do not meet and enter step e;

E. between the sub-population of the 3rd layer, select, variation, interlace operation, and then calculate fitness function value, see and whether satisfy condition, if meet, export as test case, otherwise m-1, in the second layer, select again, variation, interlace operation;

If f. still do not satisfy condition at m=1 layer fitness function, counter adds 1 and enters into the second generation, and then repeating step d, till so circulation obtains the test case in coverage goal path always.

3. a kind of method for generating test case based on hierarchical genetic algorithms as claimed in claim 1 or 2, it is characterized in that, in the fitness function module of the Huffman encoding described in step 3, remember that t is x for the individuality of Advanced group species, x is the input data of tested program, the path string that x passes through is designated as f (x), the length of f (x) is designated as | f (x) | and, total n bar path to be measured, f _jrepresent j article of path to be measured, wherein j=1,2 ..., n, by calculating f (x) and f _jmatching degree obtain the fitness function P (x) of x, specific formula for calculation is as follows:

P _j(x)＝(C _j(f(x)))+1)*P _j′(x);

$P_j^{\′}\left(x\right)=\underset{k=1}{\overset{\min \left(\right|f\left(x\right),\left|f_j\right|)}{\mathrm{\Σ}}}{m}_{\mathrm{jk}}*d_{\mathrm{jk}}\left(f\left(x\right)\right);$

Wherein: d _jk: reflection f (x) and f _jk position whether identical, if identical value is 1, otherwise be 0;

C _j(f (x)): since first consecutive identical figure place;

M _jk: f (x) and f during to k position _jthe number of identical bits;

By calculating P _j(x) we just can obtain f _jmatching degree with f (x).

Images