jpa01 : Testing: test coverage and mutation testing
|num||ready?||description||assigned MW lect||due MW lect||assigned TR lect||due TR lect|
|jpa01||true||Testing: test coverage and mutation testing||Tue 09/27 01:00PM||Sun 10/09 11:59PM|
This assignment is
jpa01, i.e “Java Programming Assignment 01”.
If you find typos or problems with the lab instructions, please report them on the #typos channel on Slack
In this lab:
- using packages
- writing your own JUnit tests
- test coverage using Jacoco
- publishing test coverage using Codecov
- mutation testing with pitest
Before starting this lab, whether you are brand new to Java, or even if you are very experienced with Java, we encourage you to read through parts 1 through 12 of this Java tutorial:
Most of this lab is concerned with really basic object-oriented programming: getters, setters, and simple methods of a class.
However, there are a few concepts that may be new even to the most experienced Java programmers:
It covers some concepts that may be new to you, even if you have worked with Java before, including:
- JUnit testing
- Annotations (things that start with
- Maven (the
pom.xmlfile and the
- Code Coverage (with Jacoco)
- Mutation Testing (with Pitest)
- Exceptions (
catch, and how Exceptions work in Java)
You may also want to read ahead in the Head First Java book (HFJ2), or use the Java in a Nutshell v7 text (JN7) if you encounter Java concepts with which you are unfamiliar.
Working in a pair?
You have the option to work in a pair on this assignment.
To set up a pair repo, see these instructions: Pair Repo Setup.
- Read about Strong Style Pairing and consider trying it.
- If you are working in aSwitch navigator/driver frequently and tradeoff who commits
If you are in your repo directory, and type
git log at the command
line, you’ll see a list of the commits for your repo.
Record that you are pairing on each commit message by putting the initials of the pair partners at the start of the commit message.
E.g. If Selena Gomez is driving, and Justin Timberlake is
navigating, and you fixed a bug in your
getDanceMoves() method, your
commit message should be
SG/JT fixed bug in getDanceMoves()
We should see frequent switches between SG/JT and JT/SG.
When remote pairing, this means switching who is sharing their screen:
- The person driving pushes the repo.
- You switch who is sharing the screen
- Then, the other person pulls from the repo.
Step 0: Set up your repo
When cloning your repo, be sure to use only the ssh links.
- Using the
httpslink may result in an error when you try to push the to the repo, indicating that “a GitHub Action cannot be accessed from an OAuth login” or something like that.
- Using ssh instead of https gets around this problem.
You may work individually or as a pair on this lab. However, if you work as a pair, please:
- Pair with someone from your same team, or at least, from your same discussion section.
- Remember to name the repo correctly, and also to add your pair on Gradescope each time you submit
The starter code is in https://github.com/ucsb-cs156-f22/STARTER-jpa01. Visit that page for the approrpiate URL to add the
To add the starter as a remote, cd into the repo you cloned, then do:
git remote add starter https://github.com/ucsb-cs156-f22/STARTER-jpa01
git checkout -b main git pull starter main git push origin main
That should get you set up with the starter code.
Step 1: Get oriented to packages
A few things to notice:
src, there are two directory trees:
src/main/java/edu/ucsb/cs56/pconrad/menuitemscontains regular Java classes.
src/test/java/edu/ucsb/cs56/pconrad/menuitemscontains the test classes.
Don’t change the package from
pconrad to your name; the Gradescope autograder is looking for the code under the
So each source file:
- must be under that directory path when it is compiled, and
- must have
package edu.ucsb.cs56.pconrad.menuitems;as the first line in the file
Here are the commands you’ll need as you work with the code. Try them out now.
|To do this||Type this command||Notes|
|compile the code||
|run the tests||
|generate a report of test coverage||
||Open the file
|generate a jar file||
|generate a mutation test report||
||Open the file target/pit-reports/yyyymmddhhmm/index.html to see the report; note that yyyymmddhhmm will be replaced by a date/timestamp.|
Step 2: Start writing code to make tests pass
In this lab, you’ll be implementing several methods of a class called
MenuItem that represents
item on a restaurant Menu.
(There is a follow up lab in which we will add a
Menu class that uses these menu items; but
we need to discuss sorting,
and Java lambda expressions in lecture first before we get to that.)
MenuItem represents an item on the menu of a restaurant. It has three attributes:
- the menu item name, e.g.
"Small Poke Bowl"
- the price, in cents (e.g an item that costs $1.49 is represented by the integer 149)
- a category such as
Note that the starter code:
- Has stubs for SOME of the needed methods, but NOT ALL of them
- Has unit tests for SOME of the needed methods, but NOT ALL of them
YOU WILL NEED TO WRITE SOME OF YOUR OWN TESTS.
So you’ll need to do a bit more work than you may be used to.
I suggest that you work in this order:
- First, Add stubs for all of the methods that don’t have them yet.
- Until you do this, you won’t be able to run any of the instructor unit tests on Gradescope.
- The reasons is that the instructor tests won’t compile against your code unless and until you have those methods.
- Then, try submitting on Gradescope
- At this point, you should have a clean compile for both the student and instructor code, though you won’t be passing most of the unit tests.
- Then, one at a time, work on each method
- If the method doesn’t have a unit test yet, write the test first and see it fail.
- Then make the test pass.
- Then submit on Gradescope and see if the test for that method passes on Gradescope
- Continue until all of your methods work.
Let me say this again, because I’ve assigned this lab a few times now, and each time, students fail to read this important instruction:
YOU MUST WRITE STUBS FOR ALL METHODS before any of the Gradescope tests will work. Until you write stubs for all methods, the code will FAIL TO COMPILE on Gradescope, and no tests will pass.
Details about methods of MenuItem
The constructor has the signature:
public MenuItem(String name, int priceInCents, String category)
Here are the instance methods you’ll need to implement for
|Modifier and Type||Method||Description|
||Returns the category of the menu item|
||Returns the name of the menu item|
||Returns the price, formatted as a string with a $.|
||Returns the price, formatted as a string with a $, right justified in a field with the specified width.|
||get the price in cents only|
||return a string in csv format, in the order name,price,cateogry.
In this case, the price is unformatted; just an integer number of cents.
Checking the tests
To check the code against the tests that you’ve written, use:
Interpreting the output can be
Step 3: Learning about Test Coverage
Now, did you really write unit tests for all of your code? Let’s check!
We can automatically compute “test case coverage”, using a tool call JaCoCo (Java Code Coverage).
Read these short articles about test coverage before moving to step 4:
Once you’ve looked over those, it’s time to check your test coverage, which we’ll do in Step 4.
Step 4: Checking Test Case Coverage
Be sure that you’ve added your pair partner to your submissions on Gauchospace
Then, check your test coverage!
Checking Test Case Coverage Locally.
mvn test jacoco:report
- Then, open the file
target/site/jacoco/index.htmlin a browser.
How you go about that depends a lot on your setup.
- On CSIL, we suggest mounting your home directory with Samba so that you can navigate to the
target/site/jacoco/index.htmlfile from your computer, and double click on it to open it.
- Instructions for that are here for Windows and MacOS
How to read the test coverage reports
- If any line of code is red, that means it is not tested at all—it is being missed by line coverage
- If a line of code is yellow, it means there are multiple ways to execute the line.
- it may have an if/else, or a boolean expression involving
||, and thus there are multiple paths through the code (multiple branches).
- Yellow means it is being missed by branch coverage; some branches are covered, and others are not.
- Think about the multiple paths through the code and be sure your tests are coverage all of them.
- it may have an if/else, or a boolean expression involving
Step 5: Try to get as close to 100% coverage as you can
Keep reworking your code until you get as close as you can to 100% test coverage.
As we’ll discuss in lecture, 100% coverage isn’t always necessary or even desirable, but in this case you should be able to get there, or at least pretty close.
Try to get as close to 100% test coverage as you can. Then, once you’ve gotten to 100%, we’ll check the quality of your tests with an even more rigorous method, called “mutation testing”.
What is mutation testing?
Here is an explanation of mutation testing from the website for pitest which is the package we are using for mutation testing.
Explanation of Mutation Testing (click the arrow to reveal)
What is mutation testing?
How it works in 51 words
Mutation testing is conceptually quite simple.
Faults (or mutations) are automatically seeded into your code, then your tests are run. If your tests fail then the mutation is killed, if your tests pass then the mutation lived.
The quality of your tests can be gauged from the percentage of mutations killed.
Really it is quite simple
To put it another way - PIT runs your unit tests against automatically modified versions of your application code. When the application code changes, it should produce different results and cause the unit tests to fail. If a unit test does not fail in this situation, it may indicate an issue with the test suite.
What’s wrong with line coverage?
Traditional test coverage (i.e line, statement, branch, etc.) measures only which code is executed by your tests. It does not check that your tests are actually able to detect faults in the executed code. It is therefore only able to identify code that is definitely not tested.
The most extreme examples of the problem are tests with no assertions. Fortunately these are uncommon in most code bases. Much more common is code that is only partially tested by its suite. A suite that only partially tests code can still execute all its branches (examples).
As it is actually able to detect whether each statement is meaningfully tested, mutation testing is the gold standard against which all other types of coverage are measured.
So, to clarify, you will run mutation testing once you’ve gotten to 100% line/branch coverage, or as close to it as you can get. At that point, you run:
mvn test org.pitest:pitest-maven:mutationCoverage
(Yeah, no one is going to expect you to memorize that command.
We might expect you to know
mvn package, and
mvn test, and maybe even
mvn test jacoco:report, but not this one.)
When you run that command, your file
MenuItem.java will be mutated; the pitest software will create mutant versions of that code in an attempt to deliberately introduce bugs. A mutant is killed when your test suite finds the bug. It survives when your test suite doesn’t kill the bug.
The output will show you how many mutants are killed.
You can see more detailed output by using a web browser to open up the file:
Note that yyyymmddhhmm will be replaced by a date/timestamp; each time you run the Maven command to run pitest, it will produce a new version of the report with a new timestamp.
It may be convenient to use
mvn clean before running a pitest mutation report; that way there will only be one such directory rather than multiple ones.
Another way to see the Pitest report: GitHub Actions artifacts
Another way to see the pitest report is to let the GitHub Actions run for your repo (which happens each time you push a change to GitHub). When you do, you should see that there is a link for “Artifacts” when you examine the Github Actions results.
Note: If you don’t see the Artifacts as shown below, click the arrow for hints on what might be wrong.
First, you must be passing all of the regular JUnit tests, or you won’t get a mutation testing report from pitest.
So check that you are passing all of the regular JUnit tests first.
Second, there is a chance that you picked up an old version of the starter code—and to be clear, you didn’t necessarily do anything wrong;
the first version of the instructions had
master instead of
main, and that branch didn’t have everything in it.
To check, compare your
.github/workflows/maven.yml file with the listing of the file here:
.github/workflows/maven.yml from starter code
Specifically, make sure the files ends with these lines, indented so that all of the
- name commands line up (just like in the starter code at the link above)”:
- name: Pitest run: mvn test org.pitest:pitest-maven:mutationCoverage - name: Upload Pitest to Artifacts uses: actions/upload-artifact@v2 with: name: pitest-mutation-testing path: target/pit-reports/**/*
If you download the artifacts, you’ll get a .zip file that you can download, and
open. Inside that zip file, you’ll find an
index.html file that you can open in a
web browser. That will show you the Pitest mutation report.
Mutation testing is 30% of your grade
30% of your grade for this lab will come from the number of mutants you kill. Kill 100% of the mutants, you get 30/30. Kill only half of them, you get 15/30.
Gradescope will run the mutation testing on your solution, and award points proportionately depending on how many you kill.
Now: since the mutants are based on your your individual solution, it is impossible to know in advance how many there will be, or what they will be. But you can investigate by looking at the report.
Step 6: Before final submission on Gradescope
Look for jpa01 on Gradescope.
If working in a pair:
- Make sure that your README.md file has your github id(s) and name(s)
- Make sure that both of your names are attached to your Gradescope submission.