|Version 41 (modified by 8 years ago) ( diff ),|
PathMap is an extended unit test runner that provides valuable feedback for software maintenance tasks such as debugging, testing, and program comprehension. This tool that does not only verify test cases but also analyze them in order to provide feedback about test quality, identify failure causes, derive likely invariants, and trace requirements.
Test cases possess a valuable source of information as they implicitly define expected and unexpected behavior all over the system. During the execution of their exemplary assertions, they do not merely cover directly-tested methods but rather large call sequences with corresponding states through internal parts of the system. Based on this idea, we can inductively derive from the specific run-time values of executed test cases more generalized and helpful program properties. For example, our type harvesting exploits test cases to obtain type information for a code base automatically. We derive type data from the concrete objects used during test case executions and provide developers this hidden information to assist in navigating source code and using application programming interfaces (APIs) correctly. By making this hidden test knowledge available, developers have additional information for debugging, testing, and understanding their programs.
PathMap is our extended test runner that reveals the hidden knowledge of test cases and presents it to developers. It provides a low overhead analysis framework that computes coverage data, likely invariants, and static metrics and displays the results in form of a scalable tree map visualization representing the system under observation.
For more specific use cases, see also the following tutorials: Test-driven Fault Navigation (Debugging), Test Quality Feedback (Testing), Type Harvesting (Program comprehension), and Requirements Traceability (Program comprehension).
How to Use
- First, you need a proper Path project that is selected as current. For more information see PathProject.
- To start PathMap choose the tool either in the world docking bar at the top (Apps -> PathMap) or in the world menu (open... -> PathMap).
- Depending on the project size, it can take some seconds for computing the static analysis in background.
- After that, you will see a test runner that looks in parts similar to the original one.
- You can select tests from the list on the left and run them with the buttons below.
- Test results are then presented in the middle. At the top, you will see the summarized test results as usual, in the middle a large tree map represents the system under observation (green, yellow, and red rectangles represent the test results and can interactively be explored), at the bottom you will find some static project metrics.
- On the right side, you will find some flaps that allow PathMap to reveal hidden test knowledge within the next run (for more information see below).
PathMap supports several analysis modes that can be activated by selecting and extending the corresponding flap on the right side. After that, running test cases will automatically include the chosen analysis.
With no extended flap, PathMap acts like an ordinary test runner (no analysis in background without any performance slow down). Its main components are the test control panel (left), the tree map in the middle, and its specific flaps on the right.
- Left pane (list)
- Displays all test cases (suites) within the current Path project. Developers can select, which of them should be executed by clicking on the Run Selected button.
- Left pane (buttons)
- Run Selected: Executes the selected tests with activated analysis if a flap is open. Otherwise, tests are executed without any analysis.
- Run with Contracts: If contracts are available, tests can be run and checked for contract violations (see Induction Flap below). Violations are shown in the tree map with small exclamation marks. The corresponding background analysis still depends on opened flaps.
- Run Failures: Executes only (yellow) tests that have failed before. Analysis depends on opened flaps.
- Run Errors: Executes only (red) tests that have an error before. Analysis depends on opened flaps.
- Text box (top)
- Summarizes test results (passed, failed, errors, and expectations) as usual.
- Tree map (middle)
- Represents a structural system overview and its relation to test execution. This visualization represents categories/packages as columns and their classes as rows. Each class represents each of its methods as a box. In addition to it, methods are optionally ordered in columns with respect to their method category. The allocated space is proportional to the number of methods per node. Packages, classes, and methods are sorted alphabetically and for a clear separation we distinguish between test classes on the left-hand side and application classes on the right-hand side. Developers can systematically explore the visualization and interactively obtain more details about specific method boxes such as its name and covering tests. Moreover, each method can be colored with a hue element for reflecting specific analysis metrics. In addition to that, the visualization colors a test method in green, yellow, or red if the test respectively succeeds or fails.
- Interactive elements on tree map (morphs)
- Each small rectangle morph represents a method of the system under observation. Methods with a u mark unit test methods. By interactively hovering over the visualization, labels present the current selected source code entity. With a click on a box the following buttons can appear on a method.
- The standard buttons that are mostly available. Browse opens a source code browser for this specific method (Note: If a coverage analysis has been chosen (test quality or fault localization flap), this will also refine the coverage results at the statement-level. The browser underlines the corresponding statements). Information opens a new dialog that shows additional data about the computed analysis metrics (color in the method). Rerun allows developers to rerun one specific test case. This button is only available for unit test methods (u mark). Debug sets a breakpoint in the chosen method and runs a specific test that covers this method. Thus, a symbolic debugger opens at the chosen method and allows developers to explore its run-time behavior. Pathfinder opens our lightweight back-in-time debugger at the chosen method. This allows developers to explore the entire execution history of a specific test case that the chosen method is involved.
- If contracts are available and a specific method includes violations, this button allows developers to further inspect the violated assertions. The same icon highlights methods in the tree map, if violations have been occurred.
- If the tree map has not enough space for displaying all methods, it automatically summarizes them into their classes. Only a large rectangle is available that can be further explored at the method level with this button.
- Text box (bottom)
- Summarizes some static project metrics such as number of classes and methods or lines of code (LoC).
In the following, we will present each flap on its own.
Test Quality Flap
The test quality flap analyzes the (method) coverage of test cases (effectivity) and relates it to different static metrics. This allows developers to answer questions such as which methods require more tests and which author is responsible for missing tests.
See also: Test Quality Feedback
- No Static Metric drop down list (top)
- Choose the static metric that colors the underlying tree map (hue) value. We offer author, last change, lines of code, complexity, and variable access metrics. The coverage of following analyzes only influences the saturation color so that both (coverage and static metric) can be seen at the same time. Thus, it is obvious which parts of the system are covered with a specific metric value and which are not.
- Sort Methods by Metric button
- Restructures the tree map layout by ordering methods with their static metric value first instead of the original alphabetical order.
- Methods drop down list
- Choose between pure method coverage (method was executed by any test or not), test intensity (number of tests that cover a method), or statement coverage (percentage of all executed statements within a method).
- Color legend (bottom)
- Interactive legend that relates coverage data and static metric values in order to explain what each color means.
The profiling flap measures the run-time characteristics of test cases (efficiency) and their relation to the methods of the system under observations. Performance bottlenecks, memory leaks, and massively called methods are revealed with this analysis.
See also: Test Quality Feedback
- Method Calls drop down list (top)
- Choose the profiling measurement. We offer number of method calls, number of created instances (in this case, the tree map hides all methods and limits the visualization to classes), pure run-time, and run-time per method call.
- Profile Leafs button
- With this option, we measure the exact run-time of each method for its own (ignoring the run-time of sub-calls). This means the sum of all called methods equals to the entire run-time. Without this option, we measure for each method its entry and exit so that following method calls are also included in the results.
- Color legend
- Shows the required performance spectrum from blue as less to red as high.
- Highscore button
- Shows a sorted list of all performance measurements.
Fault Localization Flap
The fault localization flap reveals spectrum-based anomalies (methods) that have a high probability to include failure causes. By comparing the coverage of all failed and passed test cases, we offer several starting points for debugging failing test cases and creating initial failure cause hypotheses.
See also: Test-driven Fault Navigation
- Ochiai drop down list (top)
- Choose a spectrum-based fault localization metric. We have implemented Ochiai, Tarantula, and Jaccard. Nevertheless, Ochiai has been shown to be the best choice in most cases. So, this metric is a proper default value.
- Color legend
- Shows the failure cause probabilities and their mapping to colors. From green to red, it represents the suspicious score and from colorful to white the confidence score. For a developer, you need to know that colors express the failure cause probability (from red as high to green as low).
- Single Fault button
- If you can ensure that you debug only one failure at a time, then this button filters some failure cause probabilities and simplifies the debugging process. All methods that are not covered by all failing tests are hidden because it is necessary that for one root cause all failing test cases overlap (origination condition of single faults).
- Highscore button
- Lists all failure cause probabilities and their methods in descending order.
- Developer Ranking button
- Recommends experienced developers for helping with this failure.
The induction flap derives likely invariants from passing test cases. We expect that every state (parameter, receiver, return value) of a passing test is also a valid data point. Based on this assumption, we generalize concrete run-time values all over the system into assertions. Such likely invariants can help in program comprehension as well as debugging if they are compared to failing test cases.
- No Harvester drop down list (top)
- Choose the inductive analysis mode. So far, we support the analysis of type information and value ranges.
- Color legend
- Depending on the analysis, this legend shows the tree map colors of the summarized results.
- Add Contracts button
- After the inductive analysis is done and results are available, this button allows developers to add derived likely invariants as assertions and contracts to the involved methods. This works completely automatic and can be seen afterwards in our PathBrowser. The contracts will be built with all available likely invariants and not only with the last analysis run.
- Remove Contracts button
- Remove all existing contracts from the source code. Also manually adapted contracts will be deleted.
The traceability flap connects arbitrary concerns and software artifacts with test cases in order to analyze their relationship with the code base. For example, requirements can be related to their acceptance tests and we compute methods that overlap or are completely specific to requirements.
See also: Requirements Traceability
- No Menu Entry drop down list (top)
- After defining groups of related concerns, you can choose what you are interested in and should be analyzed later on.
- Selection list
- With a chosen group, all concerns are listed here. After the analysis run, you can select and deselect specific concerns and the tree map updates immediately.
- Concerns... button
- A new dialog opens and allows developers to define groups, concerns, concern colors for the tree map, and a mapping to related test cases.
- Color legend
- Depending on the available concerns, this legend summarizes related colors.
- Specific Only button
- Hide all methods that are related to multiple concerns so that only 1:1 mappings between concerns and methods remain.
- Multiple Colors button
- If methods are covered by multiple concerns, this button switches between gray scale and all concern colors within a method. However, as too many colors can confuse, this button can hide these details.
Global Options Flap
The global options flap allows developers to configure analysis and visualization properties.
- Show Meth. Categories button
- Switch the tree map layout with and without Smalltalk's method categories. Without method categories the layout is more compact and more methods can be drawn.
- Show All Methods button
- Usually, we hide simple methods (quick methods) such as accessors. With this button also these methods can be shown in the tree map.
- Ignore Setup Code button
- Our analyzes consider each method call of a test case. However, sometimes, it is meaningful to not trace methods that are called within setUp and tearDown methods.
- Visualize Inter-Steps button
- During our analyzes we update the tree map each time new results are available. This is more a features for demonstration than for production purposes. It can be disabled here.
- Show Test Results button
- The default behavior in our PathMap is that test methods (u mark) are colored with their test results. As this hides the metric value, this option allows developers to ignore the test result visualization and to see the underlying metric score.
- pathMap.png (62.1 KB ) - added by 8 years ago.
- testQualityFlap.png (11.4 KB ) - added by 8 years ago.
- profilingFlap.png (10.7 KB ) - added by 8 years ago.
- faultLocalizationFlap.png (20.0 KB ) - added by 8 years ago.
- inductionFlap.png (10.7 KB ) - added by 8 years ago.
- traceabilityFlap.png (12.5 KB ) - added by 8 years ago.
- optionsFlap.png (13.4 KB ) - added by 8 years ago.
- pathMapMorphButton1.png (5.7 KB ) - added by 8 years ago.
- pathMapMorphButton2.png (3.6 KB ) - added by 8 years ago.
- pathMapMorphButton3.png (3.7 KB ) - added by 8 years ago.
Download all attachments as: .zip