CN106909503B - A simplified method of automated test cases for Android applications - Google Patents

Abstract

The invention discloses a method for simplifying an Android application automatic test case, belonging to the technical field of software testing. The test case that triggers the crash of an Android application is used as an input, and the test case recursively generates a new test case under an in-layer reduction algorithm; If the test case is not in the failed test case set, the new test case is tested. If no Android application error is detected during the testing process or the detected error is inconsistent with the original test case error, the new test case will be added to the failed test case. In the test case set, otherwise, if the error detected by the new test case is consistent with the error of the original Android application, it will be used as a new input for a new round of simplification; this cycle is processed until the new test case satisfies the in-layer simplification algorithm the output conditions. In the present invention, the setting of the set of failed test cases is used to store the failed test cases, which can reduce repeated tests and improve the efficiency of simplifying test cases.

Description

A simplified method of automated test cases for Android applications

technical field

The invention belongs to the technical field of software testing, in particular to a method for streamlining Android application automation test cases.

Background technique

The Android system developed by Google, as one of the smart phone operating systems, is an open source mobile operating system based on the Linux platform, and now it has occupied the majority of the global smart phone operating system market. Users' demand for Android applications is increasing day by day. In order to develop stable applications more quickly and effectively, Android application automation testing technology has been well developed.

Test tools based on automated testing technology improve the efficiency of Android application testing, but also bring some side effects. Among them, the most important problem is that because they do not know the application under test, the test cases generated by these tools have a lot of invalidity and redundancy, which makes the size of test cases relatively large. When a test case causes the Android application to crash, if the program defect analysis is performed according to the original test case, it is usually difficult to accurately locate and locate the application defects, or the range of defects may become very large, and also It consumes a lot of time and manpower. For a debugger, fixing this bug would require manual execution of so many events in hopes of reproducing the bug in the Android app. However, even if reproduced, it may still be difficult to pinpoint the exact sequence of events that triggered the error.

Incremental debugging is a test case minimization method for automation, a greedy search method for automated program debugging using a systematic "what-if-test-result" loop. The initial implementation of the incremental debugging algorithm adopts a character-based pruning strategy, which generates a series of variant test cases by deleting consecutive strings or statement blocks in the test cases to be simplified.

Hierarchical incremental debugging is to generalize the idea of incremental debugging algorithm into the hierarchical structure of the input. Different from the incremental debugging algorithm, hierarchical incremental debugging does not regard the input as a flat block, but first for it. Build a tree structure and then apply an algorithm like incremental debugging to process each level of the tree top-down. Compared with the incremental debugging algorithm, the hierarchical incremental debugging can generate smaller output results through fewer steps, and its simplification efficiency is significantly improved.

SUMMARY OF THE INVENTION

Aiming at the problems existing in the prior art that the defects of the program cannot be accurately located when the test cases are long, and the traditional simplification method is inefficient and time-consuming, the invention proposes an automatic test case simplification method for Android applications.

Specific steps are as follows:

Step 1. For an Android application, build the event sequence tree of the test case according to the state log, state information and migration relationship of a test case that can trigger the application error;

The specific construction process is as follows:

First, starting from the virtual root node i=0 of the tree, each time there is a new event, the corresponding tree node is established;

The set of tree nodes corresponding to the new event is {i ₁ ,i ₂ , _{i 3} ,...,in ,...};

Then, determine the position of the tree node corresponding to the new event in the tree:

For the tree node in of the new event, if the state information of the node in _is the same as the state information of the previous node in _{-1 ,} then the node in _is the sibling node of the node in _-1 , the state of the two nodes, etc. price;

If the state information of the node in is the same as the state information of a certain node _im in the path from the root node _i = ₀ to the previous node in _-1 , then the node in is the _sibling node of the node im;

If the state information of the node in is different from the state information of any node in the path from the root node _i = ₀ to the previous node in _-1 , the node in is a child node of the node in _-1 .

Step 2: After the construction of the event sequence tree is completed, start traversing from the next layer of the root node of the event sequence tree, and obtain the node set of all events of the current traversed level of the event sequence tree;

The current traversal level is initially set to 1;

Step 3: Determine whether the node set of the current traversal level is empty, if so, go to Step 6; otherwise, go to Step 4;

Step 4: Use the in-layer simplification algorithm to simplify the node set of the current traversed layer to obtain the smallest configuration unit that can trigger the error of the original Android application;

Specific steps are as follows:

Step 401, for each node in the node set of the current traversal level, count the number of events of the subsequence blocks corresponding to each node, and generate a public event sequence;

A subsequence block is a block formed by all nodes in the entire subtree rooted at this node;

The common event sequence refers to the combination of events corresponding to all nodes in all levels up to the current traversal level.

Step 402, according to the number of nodes in the node set of the current traversal level, or the number of events of the sub-sequence blocks of each node, group the nodes in the current node set, and each group is used as a sequence block;

The grouping value is initially N=2;

When grouping by the number of nodes, try to keep the number of sub-blocks contained in each sequence block as close as possible;

When grouping by number of events, the total number of events contained in each sequence block is as close as possible.

Step 403: Acquire each sequence block in turn, and merge the events corresponding to all nodes in the current sequence block, or the events corresponding to all nodes in the complement sequence block of the current sequence block, respectively, and the events in the public event sequence are merged in order, as a set of event sequences to be tested;

The first sequence of events to be tested is initially selected as the current sequence of events to be tested, and there are two types of sequence of events to be tested: the set of the last node in the node set that does not include the current traversal level, and the set of nodes that include the current traversal level. The collection of the last node.

Step 404, determine whether the current sequence of events to be tested includes the last node in the node set, if so, go to step 405; otherwise, go to step 406;

Step 405, determine whether the current sequence of events to be tested appears in the set of failed test cases, if not, go to step 407; otherwise, go to step 406;

The failed test case set stores the sequence of events to be tested that proved that the error of the original Android application could not be triggered in the previous operation; the failed test case set is initially empty.

Step 406, skip the current event sequence set to be tested without executing the test, select the next event sequence set to be tested, and return to step 404;

Step 407: Transfer the script corresponding to the current event sequence set to be tested to the original Android application testing machine to run, and determine whether the original Android application error is successfully triggered. If so, go to step 408; otherwise, go to step 409;

Step 408: Determine whether the number of nodes in the node set of the current traversal level corresponding to the current event sequence set to be tested is greater than 1, if so, regroup and return to step 403; otherwise, the node set of the current traversal level cannot be grouped again, and the obtained node A collection is the smallest unit of configuration.

When the current sequence of events to be tested is a complementary sequence, the nodes in the current sequence block corresponding to the sequence of events to be tested are merged into a set of nodes; and the number of groups in a new round should be reduced by 1 from the original value N; otherwise, the value of N The value is set to the default value of 2.

Step 409: Store the current sequence of events to be tested in the set of failed test cases, select the next sequence of events to be tested, and return to step 404;

Step 410: Determine whether the sequence of events to be tested corresponding to all sequence blocks and the sequence of events to be tested corresponding to all complements fail to trigger the original Android application error, and whether the grouping value N is less than the number of nodes in the node set, if Yes, the grouping number N of the current node set needs to be reset equal to the smaller value of 2N and the number of nodes in the node set; return to step 403; otherwise, the grouping value N is equal to the number of nodes in the node set, and the minimum configuration unit is obtained.

Step 5: Perform subtree clipping on the event sequence tree according to the minimum configuration unit, and use the clipped event sequence tree as a new event sequence tree to be simplified, increment the current traversal level by 1, and return to step two;

Step 6: The node sets obtained in the current layer are all empty, and the event sequence tree of the smallest test case that successfully triggers the error of the original Android application is obtained.

The advantages of the present invention are:

1) A method for streamlining test cases for Android application automation, which provides great convenience for program testers to simplify test cases and precisely locate program errors.

2), a simplified method of Android application automated test cases, compared with the traditional incremental debugging method, the efficiency is greatly improved. For most test cases, the simplified time can be shortened to 20%-50% of the traditional method, Save a lot of time.

Description of drawings

Fig. 1 is the schematic diagram of Android application automation test case simplification of the present invention;

Fig. 2 is a flow chart of a method for streamlining Android application automation test cases according to the present invention;

Fig. 3 is the schematic diagram of the principle of the reduction algorithm in the layer adopted by the present invention;

Fig. 4 is the schematic flow chart of the reduction algorithm in the layer adopted by the present invention;

5-1 to 5-8 are simplified schematic diagrams of specific embodiments of the present invention;

Fig. 6 is that the present invention selects 11 test cases to adopt the effect and time statistics table of this method and traditional method;

FIG. 7 is a schematic diagram showing the time-consuming comparison between the present method and the traditional method using 11 test cases selected by the present invention.

Detailed ways

The present invention will be further described in detail below with reference to the accompanying drawings and implementation examples.

The present invention is based on the automatic testing technology of Android application, which provides conditions for the error location and debugging of Android application software; the basic principle is shown in Figure 1. Under the given test environment, a simplified model of the application test cases is designed, and a tree-like hierarchical structure-based test case reduction algorithm is proposed to simplify the test cases, hereinafter referred to as the hierarchical reduction algorithm.

The purpose of this simplified model design is to simplify the test cases that trigger Android application crashes, reduce the size of the test cases as much as possible, so that the simplified test cases can still trigger the crash of the original program, that is, the simplified test cases It is equivalent to the original test case in terms of test effect.

Specifically, the test case that triggers the crash of the Android application is used as the input of the model, and the test case recursively generates a new test case under the action of the specified reduction strategy; if the new test case is not in the set of failed test cases, the new test case Carry out the test. If the Android application error is not detected during the test or the detected error is inconsistent with the error of the original test case, the new test case is added to the set of failed test cases. However, if the new test case detects an error If the error is consistent with the original Android application, it will be used as a new input for a new round of simplification; this cycle is processed until the new test case satisfies the output conditions of the simplification strategy. In this model, the setting of the set of failed test cases is used to store the failed test cases, which can reduce repeated tests and improve the efficiency of test case reduction.

The present invention adopts the idea of hierarchical simplification algorithm, improves the incremental debugging algorithm by using the state information of the test case in the execution process, and introduces the relatively mature incremental debugging technology and hierarchical idea into the field of Android application test case simplification, and On this basis, a hierarchical reduction algorithm is proposed and implemented; its core idea is to use the equivalent state after the event execution in the test case to mine the hierarchical relationship between events, and convert the time-based event sequence into an event sequence with a hierarchical structure. tree, and then simplifies the subtree set of the event sequence tree hierarchically, in order to improve the simplification efficiency. The workflow is shown in Figure 2, and the specific steps are as follows:

Step 1. For an Android application, build the event sequence tree of the test case according to the state log, state information and migration relationship of a test case that can trigger the application error;

The specific construction process is as follows:

First, starting from the virtual root node i=0 of the tree, each time there is a new event, the corresponding tree node is established;

The set of tree nodes corresponding to the new event is {i ₁ ,i ₂ , _{i 3} ,...,in ,...};

Then, determine the position of the tree node corresponding to the new event in the tree:

For the tree node in of the new event, if the state information of the node in _is the same as the state information of the previous node in _{-1 ,} then the node in _is the sibling node of the node in _-1 , the state of the two nodes, etc. price;

If the state information of the node in is the same as the state information of a certain node _im in the path from the root node _i = ₀ to the previous node in _-1 , then the node in is the _sibling node of the node im;

If the state information of the node in is different from the state information of any node in the path from the root node _i = ₀ to the previous node in _-1 , the node in is a child node of the node in _-1 .

Analyze the state log of the input test case, and then construct an event sequence tree according to the state equivalence and its migration relationship. The state log, state information and migration relationship are all attributes of the application running process; Captured and recorded while executing the test case, taking the case where the state is equivalent to using the event of the application runtime or the activity information of the node as an example, for each event in the test case, there is an application corresponding to the Activity after the event is responded to. information. Nodes, events, and Activity information are in a one-to-one correspondence.

Starting from the virtual root node i=0 of the tree, for each new event i=1, 2, 3..., a new tree node i is established, and according to the Activity information of the event, it is determined that the node i is in the tree Position in: If the activity of node i is the same as that of the previous node i-1, then node i is the sibling node of node i-1; If the activity of a node in the path of 1 is the same, then node i is the sibling node of the node; if the activity of node i is the same as that of any node in the path from the root node i=0 to the previous node i-1 Activity is different, then node i is the child node of node i-1.

The method for constructing the event sequence tree of the test case in the present invention ensures that the child nodes of each node have the same Activity information. Since the subtree is closed and independent, that is, the activity of the top node of the subtree is Activity0, the corresponding nodes generated are all in the subtree before the event sequence reaches a new Activity that is the same as Activity0. Conversely, when the event sequence reaches a new Activity that is the same as Activity0, the corresponding node generated must not be in the previous subtree; in this way, it is ensured that when processing the in-layer reduction algorithm: 1) When trying to reduce the nodes of each layer , the nodes and corresponding subtrees in one layer must be independent and relatively unrelated to each other. 2) When making reservations or discarding attempts in units of nodes and corresponding subtrees, the smallest unit must be a closed, set of Activity emitted and reached from the top Activity of the same subtree, and the corresponding event sequence block. The same is true when doing subtree pruning; it is guaranteed that the reduction algorithm in the layer is efficient and has no omissions.

Step 2: After the construction of the event sequence tree is completed, start traversing from the next layer of the root node of the event sequence tree, and obtain the node set of all events of the current traversed level of the event sequence tree;

After the tree is built, set the current traversal level to 1, that is, start traversing from the next level of the root node of the event sequence tree;

Step 3: Determine whether the node set of the current traversal level is empty, if so, go to Step 6; otherwise, go to Step 4;

Obtain the node set of all events in the current traversal level of the event sequence tree. If the node set is not empty, use the in-layer reduction algorithm to reduce the node set to obtain the smallest configuration unit that can trigger the error of the original Android application;

Step 4: Use the in-layer simplification algorithm to simplify the node set of the current traversed layer to obtain the smallest configuration unit that can trigger the error of the original Android application;

Similar to the incremental debugging algorithm, the intra-layer reduction algorithm is a greedy search algorithm that uses the system's "hypothesis-test-result" cycle for automatic program debugging, as shown in Figure 3 and Figure 4. The specific steps are as follows:

Step 401, for each node in the node set of the current traversal level, count the number of events of the subsequence blocks corresponding to each node, and generate a public event sequence;

The public event sequence refers to the combination of events corresponding to all nodes in all levels up from the current traversal level; all levels up from the current traversal level refer to level 0 to the current traversal level -1 level;

A subsequence block is a block formed by all nodes in the entire subtree rooted at that node; there is at least one element in a subsequence block; counting events in a subsequence block refers to that node and all its children. The total number of event statistics corresponding to the node.

Step 402, according to the number of nodes in the node set of the current traversal level, or the number of events of the sub-sequence blocks of each node, group the nodes in the current node set, and each group is used as a sequence block;

According to the number of nodes in the node set or the number of events in the subtree, the nodes in the node set are grouped by N to form N sequence blocks, where the default grouping value of N is initially 2, and each group is used as a sequence block; When grouping by the number of nodes, the grouping strategy is to make the number of sub-blocks contained in each of the N sequence blocks as close as possible, or when grouping by the number of events, the total number of events contained in each sequence block is as close as possible. interface features to choose from.

For the situation that the nodes in the subsequence blocks in the current traversal level are evenly distributed, either grouping by the number of nodes or grouping by the number of events can be selected.

According to the distribution of nodes in the subsequence blocks in the current traversal level, it is obvious that there are many nodes in the subsequence blocks in the back and few nodes in the subsequence blocks in the front.

The transfer relationship between the interfaces in the application program will lead to the tendency of the node distribution after the tree is built.

Step 403: Acquire each sequence block in turn, and merge the events corresponding to all nodes in the current sequence block, or the events corresponding to all nodes in the complement sequence block of the current sequence block, respectively, and the events in the public event sequence are merged in order, as a set of event sequences to be tested;

Sequential merging refers to ensuring that the events in the event sequence are in order during the merging process. In fact, the events in the sequence block are often inserted into a certain position in the middle of the common event sequence.

The expression form of the sequence of events to be tested is the set formed by the root node and all child nodes. Initially, the sequence of events to be tested in the first sequence block is selected as the current sequence of events to be tested, and there are two types of sequence of events to be tested. The set of the last node in the node set of the traversal level, and the set of the last node in the node set of the current traversal level, and the event sequence to be tested does not contain the sequence block of the last node in the node set or its complement, no need to obtain Subsequent tests, skip directly.

Complement For the current sequence block, the complement refers to all other sequence blocks remaining after the current sequence block is removed.

The sequence of events generated by the complement of the sequence block is still a common sequence of events to be tested, which is equal to the sequence of events generated by the non-complement.

If the event sequence is a complement, the value of N in the new round is equal to the value of the original N minus 1, otherwise the value of N is set to the default value of 2.

Step 404, determine whether the current sequence of events to be tested includes the last node in the node set, if so, go to step 405; otherwise, go to step 406;

Step 405, determine whether the current sequence of events to be tested appears in the set of failed test cases, if not, go to step 407; otherwise, go to step 406;

The failed test case set stores the sequence of events to be tested that proved that the error of the original Android application could not be triggered in the previous operation; the failed test case set is initially empty.

The current sequence of events to be tested does not appear in the set of failed test cases, indicating that the sequence of events to be tested has not been processed, and then the sequence of events to be tested is executed.

Step 406, skip the current event sequence set to be tested without executing the test, select the next event sequence set to be tested, and return to step 404;

The current set of events to be tested that is skipped because it does not contain the last node does not need to be added to the set of failed test cases, because in fact all sets of events to be tested that do not contain the last node will be skipped, and the set of failed test cases is only used for Saved in the previous run, it proves that the sequence of events under test cannot trigger the error of the original Android application.

Step 407: Transfer the script corresponding to the current event sequence set to be tested to the original Android application testing machine to run, and determine whether the original Android application error is successfully triggered. If so, go to step 408; otherwise, go to step 409;

Step 408: Determine whether the number of nodes in the node set of the current traversal level corresponding to the current event sequence set to be tested is greater than 1, if so, regroup and return to step 403; otherwise, the node set of the current traversal level cannot be grouped, and the obtained node A collection is the smallest unit of configuration.

The condition for regrouping is: the number of nodes in the current node set is greater than 1. Because the number of non-complement groups is 2, and the number of nodes is greater than 1, it can be grouped; for the number of complement groups is N-1, but the number of nodes in the complement node set must be greater than or equal to N-1, and it can also be grouped.

When the current sequence of events to be tested is a complementary sequence, the nodes in the current sequence block corresponding to the sequence of events to be tested are merged into a set of nodes; and the number of groups in a new round should be reduced by 1 from the original value N; otherwise, the value of N The value is set to the default value of 2.

Step 409: Store the current sequence of events to be tested in the set of failed test cases, select the next sequence of events to be tested, and return to step 404;

Step 410: Determine whether the sequence of events to be tested corresponding to all sequence blocks and the sequence of events to be tested corresponding to all complements fail to trigger the original Android application error, and whether the grouping value N is less than the number of nodes in the node set, if Yes, the grouping number N of the current node set needs to be reset equal to the smaller value of 2N and the number of nodes in the node set; return to step 403; otherwise, the grouping value N is equal to the number of nodes in the node set, and the minimum configuration unit is obtained.

If the event sequence fails to trigger the original Android application error, it will be added to the failure set; if the event sequence can trigger the original Android application error, the event sequence is a successful streamlining, as a new round of input to simplify.

If the event sequence of N sequence blocks and the corresponding event sequence of N complements fail to trigger the original Android application error, and the value of N is less than the number of nodes in the node set, it indicates that the original test case event sequence There is also the possibility of simplification in further fine-grained division. At this time, the value of N takes the minimum value between 2N and the number of nodes in the node set to ensure the effective division of a new round of event sequences; If the number of set nodes is equal, it means that each sequence block contains only one node and its subtree, and there is no possibility of finer-grained division. At this time, the event sequence to be simplified is the one obtained under the action of the in-layer simplification algorithm that can successfully trigger the original The minimum event sequence set of Android application errors is called the minimum configuration unit, and the in-layer reduction algorithm under this layer ends.

Step 5: Perform subtree pruning on the event sequence tree according to the minimum configuration unit, and use the pruned event sequence tree as a new event sequence tree to be simplified, and return to step 2 after increasing the current traversal level by 1;

The event sequence tree is trimmed according to the minimum configuration unit, and the trimmed event sequence tree is used as a new event sequence tree to be simplified. The current traversal level is incremented by 1 and the streamlining process is repeated.

Step 6: The node sets obtained in the current layer are all empty, and the event sequence tree of the smallest test case that successfully triggers the error of the original Android application is obtained.

Until the node set obtained in a certain layer is empty, the event sequence tree at this time is the smallest test case that can successfully trigger the error of the original Android application.

Specific examples:

First, for an Android application, build the event sequence tree of the test case according to the status log, Activity information and migration relationship of a test case that can trigger the application error;

As shown in Figures 5-1 to 5-8, the root node of the event sequence tree is 1, and the child nodes under the root node are: {2, 7, 17, 18, 21, and 22}; the child nodes under node 2 are: {3, 4, 5 and 6}; the child nodes under node 7 are: {8-16}; there is no child node under node 17; the child nodes under node 18 are {19 and 20}; there is no child node under node 21 Child nodes; the child nodes under node 22 are {23-36}.

In a sense, a state is a node, and an event is a transition relationship between nodes, that is, between states, because the state of the application may change every time an event executes. In order to facilitate processing, the node, state and event are bound. For example, the root node is 1, its state is state 1, and the corresponding event 1 is connected between the root node 1 and the node 2 of the first layer, indicating that after event 1 is executed, The state of the application changes from state 1 to state 2. The relationship between state i and event i is: state i is the state of the application before event i occurs. For example, in the first simplification, the node {2, 7, 17} and its subtrees are removed, and {1, 18, 19-20, 21, 22, 23-36} are retained, which means that the original state 1 passed through the event 1- 17 reaches state 18, but with events 2-17 removed, it is still possible to get from state 1 to state 18 by event 1 alone.

Then, the next layer of the root node 1 is traversed as the first layer, and the node set {2, 7, 17, 18, 21, 22} of all events of the first layer is obtained.

Use the in-layer reduction algorithm to reduce the node set {2, 7, 17, 18, 21, 22} of the first layer to obtain the minimum configuration unit;

The specific intra-layer reduction processing is as follows:

First, the node set {2, 7, 17, 18, 21, 22} is grouped according to the number of nodes, and divided into two sequence blocks according to the number of nodes. The number of groups is N=2, which are {2, 7, 17} and {18, 21, 22};

Then, the sequence blocks are sequentially acquired to generate the sequence of events to be tested. The sequence of events to be tested {1, 2, 3-6, 7, 8-16, 17} corresponding to the first sequence block {2, 7, 17} Contains the last node 22 in the node set and is skipped; the second event sequence set to be tested is: {1, 18, 19-20, 21, 22, 23-36}; contains the last node in the node set 22, and does not appear in the set of failed test cases;

Then, the script corresponding to the second event sequence set to be tested is transferred to the original Android application testing machine to run, the original Android application error is successfully triggered, and the simplification attempt is successful. Then, the number of nodes in the node set is {18, 21, 22 } is 3, greater than 1; to regroup and try to simplify again;

The new node set {18, 21, 22} is regrouped by the number of nodes, the number of groups is N=2, and is divided into sequence blocks {18, 21} and sequence blocks {22} based on the greedy strategy; sequentially obtain the sequence blocks to generate pending The sequence of events to be tested is: the sequence of events to be tested {1, 18, 19-20, 21} corresponding to the first sequence block is skipped because it does not contain the last node 22; the sequence of events to be tested corresponding to the second sequence block {1, 22, 23-36} includes the last node 22 and does not appear in the set of failed test cases; then the script corresponding to the set of events to be tested is transferred to the original Android application testing machine to run, and the result of the event sequence is executed In order to successfully trigger the original Android application error, the simplification attempt was successful; so far, the node set of the first layer has only one node {22}, which cannot be further reduced, so the first layer has reached the simplest state; the obtained minimum configuration unit is {22};

Then, cut the subtree according to the minimum configuration unit, continue to traverse the level to 2, obtain the node set of all events of the current traversal level, and start repeating.

The node set of the second layer includes {23-36}; according to the number of nodes, it is divided into 2 groups to obtain sequence blocks {23-29} and sequence blocks {30-36}, the first sequence block {1, 22, 23-29 } The corresponding test event sequence is skipped because it does not contain the last node 36; the second test event sequence set is the sequence {1, 22, 30-36}, which includes the last node 36 and appears in the failed test case The collection is unprocessed, and the corresponding script is passed to the original Android application testing machine to run, and the execution result is that the original Android application error was not successfully triggered.

At this time, {1, 22, 30-36} is stored in the failed test case set, and the sequence of events to be tested corresponding to all sequence blocks {23-29} and sequence blocks {30-36}, as well as the corresponding The sequence of events to be tested fails to trigger the original Android application error, and N=2, which is less than the number of nodes in the node set, it is necessary to reset the grouping number N=4 for the node set {23-36} of the current traversal level. Try parsing again.

The node set of layer 2 includes {23-36}, which is divided into 4 groups {23-26}, {27-29}, {30-33} and {34-36};

Then, sequentially acquire sequence blocks or their complements to generate the sequence of events to be tested, which are the event sequence {1, 22, 34-36} of sequence block 4, the complement of sequence block 4, and the event sequence {1 of sequence block 3, respectively , 22, 30-33}, the sequence of events for the complement of sequence block 3 {1, 22, 23-29, 34-36}, etc. In the actual test process, select from the last sequence of events to be tested {1, 22, 34-36}, and test in sequence;

The sequence set of events to be tested is the sequence {1, 22, 34-36}, which includes the last node 36 and does not appear in the set of failed test cases. The corresponding script is passed to the original Android application test machine to run, and the execution result is parallel. The original Android application error was not successfully triggered; the sequence of events to be tested is set as the sequence {1, 22, 34-36} and stored in the set of failed test cases; the complement of sequence block 4 does not contain the last node 36 is skipped, Sequence block 3 {1, 22, 30-33} does not contain the last node 36 and is skipped; when the event sequence to be tested is the complement of sequence block 3, the event sequence {1, 22, 23-29, 34-36} , including the last node 36 in the node set, and does not appear in the failed test case set, when the corresponding script is passed to the original Android application test machine to execute the test, and the original Android application error is successfully triggered;

The number of nodes in the complement set {23-29, 34-36} of sequence block 3 is greater than 1, and it is re-reduced as a new node set, because the sequence of events to be tested that triggers the error of the original Android application successfully is the complement sequence , so the number of groups in a new round should be reduced by 1 from the original value N=4 to become N=3; the new node set {23-29, 34-36} is divided into 3 sequence blocks {23-26}, {27 -29} and {34-36};

Similarly, the sequence set of events to be tested is the sequence {1, 22, 34-36}, including the last node 36 and appearing in the set of failed test cases, select the complement of the sequence block {34-36}, excluding the last one The node 36 is skipped, and the sequence block {1, 22, 27-29} does not contain the last node 36 to be skipped; the event sequence to be tested is the complement of the sequence block {27-29}. 23-26, 34-36}, including the last node 36 in the node set, and does not appear in the failed test case set, the corresponding script is passed to the original Android application test machine to execute the test, and the original Android application test machine is successfully triggered. Android application bug;

Then, the nodes in the complement sequence blocks {23-26} and {34-36} are combined as a node set for streamlining, because the sequence of events to be tested that triggers the original Android application error and success is the complement sequence, so a new round The number of groups should be reduced by 1 from the original value N=3 to become N=2; the sequence blocks {23-26} and {34-36} generate the corresponding test event sequences {1, 22, 23-26} and {1, 22, 34-36};

Similarly, the sequence set of events to be tested is the sequence {1, 22, 34-36}, which appears in the set of failed test cases, and the sequence block {1, 22, 23-26} does not contain the last node 36 and is skipped; so , the sequence of events to be tested corresponding to all sequence blocks {23-26} and {34-36} failed to trigger the original Android application error, and the grouping value 2 is less than the number of nodes in the node set, then reset the grouping number 2N=4 ;

Then the number of nodes {23-26, 34-36} in the node set of the second layer is grouped, and the number of groups is N=4; it is divided into 4 sequence blocks {23-24}, {25-26}, {34-35 } and {36};

Obtain the sequence blocks or their complements in turn, and generate the sequence of events to be tested, which are the event sequences {1, 22, 36} of sequence block 4, which cannot successfully trigger the original Android application error; the complement of sequence block 4 is skipped, Sequence block 3 is skipped, and the event sequence {1, 22, 23-26, 36} of the complement of sequence block 3, including the last node 36 in the node set, and does not appear in the set of failed test cases, the corresponding The script is passed to the original Android application test machine to run, and the original Android application error is successfully triggered;

If the node set {23-26, 36} of the complement of the sequence block {34-35} is greater than 1, it needs to be merged into a new node set and regrouped for simplification. Because it is the complement sequence that triggers the error of the original Android application, the number of groups in a new round is N=3; {23, 24}, {25, 26} and {36};

Obtain sequence blocks or their complements in turn, and generate a sequence of events to be tested, which are event sequences {1, 22, 36} of sequence block {36}. In the set of failed test cases, the complement of sequence block {36} is skipped , the sequence block {25, 26} does not contain the last node 36 is skipped, the sequence of events {1, 22, 23-24, 36} of the complement of the sequence block {25, 26}, containing the last node 36, does not appear in In the failed test case set, the script corresponding to {1, 22, 23-24, 36} is passed to the original Android application test machine to run, and the original Android application error is successfully triggered;

If the number of nodes in the node set {23-24, 36} of the complement of the sequence block {25, 26} is greater than 1, it is re-used as a new node set, and the grouping is simplified. Because it is the complement sequence that triggers an error in the original Android application, the number of groups in a new round is N=2; {23, 24} and {36};

Starting from the last sequence block {36}, when all the events to be tested {1, 22, 36} are in the set of failed test cases, {1, 22, 23-24} does not contain the last node 36; therefore, all sequences If the block fails to trigger the original Android application error, and N is 2 at this time less than the number of nodes in the node set, then reset the value of N equal to the smaller value of 4 and the number of nodes in the node set, because at this time The element in {23-24, 36} is 3, so N is set to 3; the number of nodes in the node set of the second layer is further fine-grained and simplified;

The node set of the second layer is {23, 24, 36}, which are divided into 3 groups to form 3 sequence blocks: {23}, {24} and {36} correspond to the generated sequence of events to be tested as {1, 22, 36} and complement {1, 22, 23, 24}, {1, 22, 24} and complement {1, 22, 23, 36}; {1, 22, 23} and complement {1, 22, 24, 36};

The sequence of events to be tested is {1, 22, 36} in the set of failed test cases, the complement {1, 22, 23, 24} does not include the last node 36 to skip; {1, 22, 24} does not include the last One node 36 is skipped; the complement {1, 22, 23, 36} contains the last node 36 that does not appear in the failed test case set, and the corresponding script is passed to the original Android application test machine to run, and the original Android application is successfully triggered. Application error; therefore, the node set {23, 36} of the complement of the sequence block {24} is used as a new node set, which is greater than 1 and regrouped for reduction.

Because it is the complement sequence that triggers the error of the original Android application, the number of groups in a new round is N=2; sequence block {23} and sequence block {36};

Starting from the last sequence block {36}, the sequence of events to be tested {1, 22, 36} of sequence block {36} has already appeared in the failure set, and the sequence of events to be tested {1, 22 of sequence block {23} , 36} The corresponding script is passed to the original Android application test machine to run, and the original Android application error is not successfully triggered. At this time, the value of N is 2, which is equal to the number of nodes in the node set, indicating that the level has not been more fine-grained and simplified is possible, so the second layer has reached the simplest state, and the obtained minimum configuration unit is {23, 36}.

Then, cut the subtree according to the minimum configuration unit, continue to traverse the level to 3 and start repeating, because there is no third-level node in the tree, the simplification process ends, and the event sequence tree of the minimum test case that successfully triggers the original Android application error is obtained. {1, 22, 23, 36};

Generate an event sequence script from the event sequence tree of the smallest test case and save it.

According to the automatic reduction method of the present invention, the event sequence with the original length of 36 is reduced to a result with a length of only 4, and it is ensured that the reduced event sequence can still trigger the same program error as the original sequence. This simplification method is generally applicable. For event sequences with a length of more than 1000, it can also achieve good results, and obtain event sequences that only retain single-digit events and can trigger the same error.

Compared with the test cases in other fields, the test cases of Android applications have its particularity, it is a collection of ordered events, each event represents a user operation. The present invention generates a series of variant test cases by strategically deleting one event or a plurality of consecutive events in the original test case. In the variant program set, variant test cases that contain undefined behavior or cannot trigger the original error are considered as unsuccessful variants and will be discarded after being marked; successful variants that can trigger the original error will be considered as unsuccessful variants. will be used as a new input for the next reduction.

The hierarchical reduction algorithm of the present invention is the application and improvement of hierarchical incremental debugging in the field of Android application test cases. Because the Android application test case is a sequence of events based on time sequence, there is no obvious hierarchical structure, and its user interface has a certain level of hierarchy. The hierarchical reduction algorithm builds a tree-like hierarchy of event sequences by recording the user interface state after the execution of each event, and then uses the intra-layer reduction algorithm hierarchically to delete one event or multiple consecutive events in the test case to be reduced. event subtree to generate a series of variant test cases.

Through the test of a historical version of the Android application DalvikExplorer, the test object is 11 event sequences randomly generated by Monkey that can trigger the program crash; as shown in Figure 6, the error type that triggers the crash is shown in the third column. The number of events is shown in the fourth column, the number of events in the result after the reduction method of the present invention is reduced, such as the fifth column, and the number of events between the two is shown in the sixth column. It can be seen that the reduction process greatly shortens the length of the sequence. The seventh, eighth, ninetieth, eleventh and twelve columns are respectively the traditional incremental debugging method as a comparison, the uniform grouping in this method is the grouping scheme according to the number of nodes, and the non-uniform grouping in this method is the grouping scheme according to the number of events. time usage. For the convenience of comparison, another line graph is drawn for the time information, as shown in Figure 7. The horizontal axis is the 11 test event sequences, and the vertical axis is the processing time (unit: second) of the three methods. It can be seen that both the uniform grouping scheme and the non-uniform grouping scheme in this method can greatly shorten the time compared with the traditional method. Using this method to simplify the test cases can greatly improve the processing efficiency.

Claims (4)

1. an Android application automation test case reduction method, is characterized in that, concrete steps are as follows: Step 1. For an Android application, build the event sequence tree of the test case according to the state log, state information and migration relationship of a test case that can trigger the application error; Step 2: After the construction of the event sequence tree is completed, start traversing from the next layer of the root node of the event sequence tree, and obtain the node set of all events of the current traversed level of the event sequence tree; The current traversal level is initially set to 1; Step 3: Determine whether the node set of the current traversal level is empty, if so, go to Step 6; otherwise, go to Step 4; Step 4: Use the in-layer simplification algorithm to simplify the node set of the current traversed layer to obtain the smallest configuration unit that can trigger the error of the original Android application; Step 5: Perform subtree clipping on the event sequence tree according to the minimum configuration unit, and use the clipped event sequence tree as a new event sequence tree to be simplified, increment the current traversal level by 1, and return to step two; Step 6: The node sets obtained in the current layer are all empty, and the event sequence tree of the smallest test case that successfully triggers the error of the original Android application is obtained. 2. a kind of Android application automation test case reduction method according to claim 1, is characterized in that, in described step 1, the construction process of event sequence tree is as follows: First, starting from the virtual root node i=0 of the tree, each time there is a new event, the corresponding tree node is established; The set of tree nodes corresponding to the new event is {i ₁ ,i ₂ , _{i 3} ,...,in ,...}; Then, determine the position of the tree node corresponding to the new event in the tree: For the tree node in of the new event, if the state information of the node in _is the same as the state information of the previous node in _{-1 ,} then the node in _is the sibling node of the node in _-1 , the state of the two nodes, etc. price; If the state information of the node in is the same as the state information of a certain node _im in the path from the root node _i = ₀ to the previous node in _-1 , then the node in is the _sibling node of the node im; If the state information of the node in is different from the state information of any node in the path from the root node _i = ₀ to the previous node in _-1 , the node in is a child node of the node in _-1 . 3. a kind of Android application automation test case reduction method according to claim 1, is characterized in that, described step 4 is as follows: Step 401, for each node in the node set of the current traversal level, count the number of events of the subsequence blocks corresponding to each node, and generate a public event sequence; A subsequence block is a block formed by all nodes in the entire subtree rooted at this node; Common event sequence refers to the combination of events corresponding to all nodes in all levels up to the current traversal level; Step 402, according to the number of nodes in the node set of the current traversal level, or the number of events of the sub-sequence blocks of each node, group the nodes in the current node set, and each group is used as a sequence block; The grouping value is initially N=2; When grouping by the number of nodes, make the number of subsequence blocks contained in each sequence block close; When grouped by the number of events, the total number of events contained in each sequence block is close to each other; Step 403: Acquire each sequence block in turn, and merge the events corresponding to all nodes in the current sequence block, or the events corresponding to all nodes in the complement sequence block of the current sequence block, respectively, and the events in the public event sequence are merged in order, as a set of event sequences to be tested; The first sequence of events to be tested is initially selected as the current sequence of events to be tested, and there are two types of sequence of events to be tested: the set of the last node in the node set that does not include the current traversal level, and the set of nodes that include the current traversal level. The collection of the last node; Step 404, determine whether the current sequence of events to be tested includes the last node in the node set, if so, go to step 405; otherwise, go to step 406; Step 405, determine whether the current sequence of events to be tested appears in the set of failed test cases, if not, go to step 407; otherwise, go to step 406; The failed test case collection stores the sequence of events to be tested that proved that the original Android application error could not be triggered in the previous operation; the failed test case collection is initially empty; Step 406, skip the current event sequence set to be tested without executing the test, select the next event sequence set to be tested, and return to step 404; Step 407: Transfer the script corresponding to the current event sequence set to be tested to the original Android application testing machine to run, and determine whether the original Android application error is successfully triggered. If so, go to step 408; otherwise, go to step 409; Step 408: Determine whether the number of nodes in the node set of the current traversal level corresponding to the current event sequence set to be tested is greater than 1, if so, regroup and return to step 403; otherwise, the node set of the current traversal level cannot be grouped again, and the obtained node A collection is the smallest configuration unit; When the current sequence of events to be tested is a complementary sequence block, the nodes in the current sequence block corresponding to the sequence of events to be tested are merged into a node set; and the number of groups in a new round should be decremented by 1 from the original value N; otherwise, N The value is set to the default value of 2; Step 409: Store the current sequence of events to be tested in the set of failed test cases, select the next sequence of events to be tested, and return to step 404; Step 410: Determine whether the sequence of events to be tested corresponding to all sequence blocks and the sequence of events to be tested corresponding to all complementary sequence blocks fail to trigger the original Android application error, and whether the grouping value N is less than the number of nodes in the node set , if yes, the grouping number N of the current node set needs to be reset equal to: 2N and the smaller value of the number of nodes in the node set; return to step 403; otherwise, the grouping value N is equal to the number of nodes in the node set, and the minimum configuration is obtained unit. 4. a kind of Android application automation test case simplification method according to claim 1, is characterized in that, in described step 402, for the situation that the nodes in the subsequence blocks in the current traversal hierarchy are evenly distributed, select according to the number of nodes Grouping or grouping by the number of events; According to the distribution of nodes in the subsequence blocks in the current traversal level, it is obvious that there are many nodes in the subsequence blocks in the back and few nodes in the subsequence blocks in the front.

Images