Giter VIP home page Giter VIP logo

circuitsim-tester's Introduction

CircuitSim Tester Library

Javadocs for API: https://zucchini.github.io/circuitsim-tester/

This repository aims to bridge the gap between the nitty-gritty details of the CircuitSim API and high-level tests on combinational logic in a CircuitSim file. Until now, we (the CS 2110 staff) handled CircuitSim homeworks with one of the following strategies:

  1. Don't provide a tester because they were scary
  2. Copy and paste some old tests, often missing error handling because the tons of duplicate code and intertwined CircuitSim implementation details made the tests hard to follow

So I have two goals in this template:

  1. Make CircuitSim autograders easy to write, for students too if they want
  2. Insulate tests from the verboseness of the CircuitSim API by wrapping the CircuitSim API, and pass these wrappers to test classes using dependency injection

Getting Started

Note: The CircuitSim.jar in the repo is built against the CS2110 fork of CircuitSim, which is not compatible with files created in "regular" CircuitSim. If you are using vanilla CircuitSim, you probably need to rebuild this library.

To make your own CircuitSim tester, take a look at the sample tester repo.

If you want, you can always include circuitsim-tester yourself to any project using Gradle, making sure that the version is the one you want:

build.gradle:

// ...

dependencies {
    implementation 'io.zucchini.circuitsim-tester:circuitsim-tester:v2.6.1'
}

repositories {
    mavenCentral()

    maven {
        url 'https://maven.pkg.github.com/zucchini/circuitsim-tester'
        credentials {
            username = project.findProperty('gpr.user') ?: System.getenv('USERNAME')
            password = project.findProperty('gpr.key') ?: System.getenv('TOKEN')
        }
    }
}

// ...

gradle.properties:

gpr.user=<github-username>
gpr.key=<personal-access-token>

Note that using GitHub Packages requires authentication using a personal access token (only requires read:packages scope). For more information, see the GitHub documentation for working with the Gradle registry.

Writing Tests

Testing basic combinational logic

Suppose we had a Toy ALU like the following that we wanted to test:

Screenshot of toy-alu.sim

You can find the full test at src/main/java/io/zucchini/circuitsimtester/tests/ToyALUTests.java, but a brief walkthrough follows if you just want the highlights.

Let's start with the boilerplate test class:

@DisplayName("Toy ALU")
@ExtendWith(CircuitSimExtension.class)
@SubcircuitTest(file="toy-alu.sim", subcircuit="ALU")
public class ToyALUTests {
    @SubcircuitComponent(bits=4)
    private InputPin a;

    @SubcircuitComponent(bits=4)
    private InputPin b;

    @SubcircuitComponent(bits=2)
    private InputPin sel;

    @SubcircuitComponent(bits=4)
    private OutputPin out;
}

@DisplayName("Toy ALU") lets students will see the name Toy ALU on test failures (if you don't specify it, they will just see the ugly fully-qualified class name instead), and @ExtendWith(CircuitSimExtension.class) tells JUnit to run some custom CircuitSim setup code first which loads up the subcircuit we want and injects pins in it into the test class instance. Indeed, @SubcircuitTest(file="toy-alu.sim", subcircuit="ALU") tells the CircuitSim JUnit extension to load toy-alu.sim and then open its subcircuit matching the name ALU. As described in the API documentation, this subcircuit name (as well as Pin labels later) will be normalized into a lowercase alphanumeric string. That way, 1-bit ALU matches 1bit alu, 1 bit ALU, and so on.

The Pin fields are pretty self-explanatory: the extension looks up Pin components in the subcircuit based on the name of fields annotated with @SubcircuitComponent. If you want to match on a different label, say label="something else".

Let's start with testing the last operation, Pass A, since it's the easiest:

    @DisplayName("pass A")
    @Test
    public void passA() {
        a.set(0);
        sel.set(0b11);
        assertEquals(0, out.get(), "out");

        a.set(0b1011);
        sel.set(0b11);
        assertEquals(0b1011, out.get(), "out");
    }

Well, that's pretty easy. But what if you need to test more than two cases? That's where JUnit 5 parameterized tests come in:

    @ParameterizedTest(name="a:{0}, b:{1}, sel:00 (a xor b) → out:{2}")
    @CsvSource({
        /*  a      b    |  out */
        "0b1111, 0b0000, 0b1111",
        "0b0000, 0b1111, 0b1111",
        "0b1111, 0b1111, 0b0000",
        "0b1011, 0b0010, 0b1001",
    })
    public void xor(@ConvertWith(BasesConverter.class) int aIn,
                    @ConvertWith(BasesConverter.class) int bIn,
                    @ConvertWith(BasesConverter.class) int outOut) {
        a.set(aIn);
        b.set(bIn);
        sel.set(0b00);
        assertEquals(outOut, out.get(), "out");
    }

JUnit will run this test once for every row in the CSV. Unlike a simple for loop in a test, if one fails, it keeps going. The @ConvertWith(BasesConverter.class) parses Strings to ints based on the prefix. So 0x is hex and 0b is binary, and anything else is decimal.

You can also source your parameterized test from a method as follows:

    @ParameterizedTest(name="a:{0}, b:{1}, sel:01 (a + b) → out:{2}")
    @MethodSource
    public void add(int aIn, int bIn,
                    int outOut) {
        a.set(aIn);
        b.set(bIn);
        sel.set(0b01);
        assertEquals(outOut, out.get(), "out");
    }

    public static Stream<Arguments> add() {
        List<Arguments> args = new LinkedList<>();

        for (int a = 0; a < (1 << 4); a++) {
            for (int b = 0; b < (1 << 4); b++) {
                args.add(Arguments.of(a, b, (a + b) % (1 << 4)));
            }
        }

        return args.stream();
    }

This allows for a much easier way to do an exhaustive test than a CSV! Note that JUnit locates the second (static!) method because it has the same name as the test.

You can see the finished product in src/main/java/io/zucchini/circuitsimtester/tests/ToyALUTests.java.

Restrictors (Banning Components)

Suppose in the previous example, we wanted students to build the XOR themselves, without using a XOR gate component. Then we can change the test as follows:

@DisplayName("Toy ALU")
@ExtendWith(CircuitSimExtension.class)
@SubcircuitTest(file="toy-alu.sim", subcircuit="ALU",
                restrictors={ToyALUTests.BannedGates.class})
public class ToyALUTests {
    public static class BannedGates extends Restrictor {
        @Override
        public void validate(Subcircuit subcircuit) throws AssertionError {
            blacklistComponents(subcircuit, "XOR");
        }
    }
    // ...

That is, we can provide restrictors to @SubcircuitTest. The fact that it's a class is useful for when you have a bunch of subcircuits with the same restrictions. In such a case, you can reduce code duplication by simply referencing the same Restrictor class in all of the tests.

Consult the Restrictor documentation for more information on restrictions. There is also support for whitelists.

Testing sequential logic

Combinational logic is easy enough, but what about a state machine? Consider the following:

fsm

which corresponds to this one-hot circuit:

screenshot of fsm.sim

The register here is kind of a pain, because we want to test the combinational logic on each side of it. Luckily, we can use a feature in the tester which replaces the register with some Pin components before running the tests, (very) roughly like this:

screenshot of fsm.sim after mocking the register

The syntax for this is similar to pins:

@DisplayName("Finite State Machine")
@ExtendWith(CircuitSimExtension.class)
@SubcircuitTest(file="fsm.sim", subcircuit="fsm")
public class FsmTests {
    @SubcircuitComponent(bits=1)
    private InputPin g;

    @SubcircuitComponent(bits=1)
    private InputPin clk;

    @SubcircuitComponent(bits=1)
    private InputPin rst;

    @SubcircuitComponent(bits=1)
    private InputPin en;

    @SubcircuitComponent(bits=1)
    private OutputPin a;

    @SubcircuitComponent(bits=2, onlyInstance=true)
    private MockRegister stateReg;
}

So we specify the input/output pins we want as usual, but then also require there to be exactly 1 2-bit register in the subcircuit: the state register.

Testing the output a is fairly simple, so let's start with that:

    @DisplayName("output a")
    @Test
    public void outputA() {
        stateReg.getQ().set(0b00);
        assertEquals(0, a.get(), "output a in state 00");
        stateReg.getQ().set(0b01);
        assertEquals(1, a.get(), "output a in state 01");
        stateReg.getQ().set(0b10);
        assertEquals(0, a.get(), "output a in state 10");
    }

MockRegister.getQ() returns an InputPin that the tester placed where the out port of the register used to be, so we can set it to an arbitrary value just like any other input pin.

MockRegister.getD(), on the other hand, is an OutputPin:

    @ParameterizedTest(name="state:{0}, g:{1} → next state:{2}")
    @CsvSource({
        /* state  g | next state */
        "   0b00, 0,        0b01",
        "   0b00, 1,        0b01",
        "   0b01, 0,        0b01",
        "   0b01, 1,        0b10",
        "   0b10, 0,        0b01",
        "   0b10, 1,        0b10",
    })
    public void transition(@ConvertWith(BasesConverter.class) int stateIn,
                           int gIn,
                           @ConvertWith(BasesConverter.class) int nextStateOut) {
        stateReg.getQ().set(stateIn);
        g.set(gIn);
        assertEquals(nextStateOut, stateReg.getD().get(), "next state");
    }

You can see the finished product at src/main/java/io/zucchini/circuitsimtester/tests/FsmTests.java.

Debugging This Library

If an autograder is producing weird failures (for instance, the error message printed for some test is null), you can pass --verbose to the grader jar to make the custom JUnit Launcher in this library print the full stack trace for the Throwable corresponding to every test failure.

If worst comes to worst, you can clone this repository locally and build it alongside a grader. The following steps are an example workflow:

  1. Clone this repository locally
  2. In the autograder directory itself:
    1. Modify settings.gradle to register the local clone:
      include ":circuitsim-tester"
      project(":circuitsim-tester").projectDir = file("path/to/circuitsim-tester")
      
    2. Remove the dependency on io.zucchini.circuitsim-tester:circuitsim-tester inside the dependencies {} block in build.gradle, and add a new dependency on the locally-cloned project. The result might look like this:
      dependencies {
          // implementation 'io.zucchini.circuitsim-tester:circuitsim-tester:v2.4.0'
          implementation project(":circuitsim-tester")
      }
      
  3. In the cloned copy of this repository, comment out these two lines related to git version in build.gradle:
    // id 'com.palantir.git-version' version "0.12.3"
    ...
    // version gitVersion()
    
    For autograders (consumers) using newer versions of Gradle, you may also need to comment out the sourcesJar and javadocJar tasks as well as the artifacts {} and publishing {} sections. Isn't software "engineering" fun?
  4. Run ./gradlew jar in both the cloned copy of this repository and in the top-level autograder directory, in that order. If you modify the code in the copy of this repository, you will have to rerun the command to rebuild the jar.

Caveats

Because CircuitSim is a JavaFX application and this tester simply runs CircuitSim, this tester tries to attach to a display — like an X11 $DISPLAY on GNU/Linux. If you want to run this grader in a headless environment, you can get around this by standing up a dummy X11 server.

For GNU/Linux, zucchini installs xf86-video-dummy and then uses the following run_graphical.sh:

#!/bin/bash
cat >xorg.conf <<'EOF'
# This xorg configuration file is meant to be used by xpra
# to start a dummy X11 server.
# For details, please see:
# https://xpra.org/Xdummy.html
Section "ServerFlags"
  Option "DontVTSwitch" "true"
  Option "AllowMouseOpenFail" "true"
  Option "PciForceNone" "true"
  Option "AutoEnableDevices" "false"
  Option "AutoAddDevices" "false"
EndSection
Section "Device"
  Identifier "dummy_videocard"
  Driver "dummy"
  Option "ConstantDPI" "true"
  VideoRam 192000
EndSection
Section "Monitor"
  Identifier "dummy_monitor"
  HorizSync   5.0 - 1000.0
  VertRefresh 5.0 - 200.0
  Modeline "1024x768" 18.71 1024 1056 1120 1152 768 786 789 807
EndSection
Section "Screen"
  Identifier "dummy_screen"
  Device "dummy_videocard"
  Monitor "dummy_monitor"
  DefaultDepth 24
  SubSection "Display"
    Viewport 0 0
    Depth 24
    Modes "1024x768"
    Virtual 1024 768
  EndSubSection
EndSection
EOF
/usr/lib/xorg/Xorg -noreset -logfile ./xorg.log -config ./xorg.conf :69 \
    >/dev/null 2>&1 &
xorg_pid=$!
export DISPLAY=:69
"$@"
exitcode=$?
kill "$xorg_pid" || {
    printf 'did not kill Xorg!\n' >&2
    exit 1
}
exit $exitcode

which allows you to say ./run_graphical.sh java -jar hwX-tester.jar on a headless system.

Zucchini Support

A zucchini backend exists, CircuitSimGrader, which reads the JSON printed to stdout for

java -jar hwX-tester.jar --zucchini FsmTests

given some test FsmTests. Each subcircuit you want to test (aka each test class) should correspond to a different assignment component, and each test method is a part of that component. Here's a zucchini.yml for the FSM example above:

name: Homework X
author: Austin Adams
components:
- name: Finite State Machine
  weight: 1
  backend: CircuitSimGrader
  backend-options:
    grader-jar: hwX-tester.jar
    test-class: FsmTests
  files: [fsm.sim]
  grading-files: [hwX-tester.jar]
  parts:
  - {test: clockConnected,  weight: 1}
  - {test: resetConnected,  weight: 1}
  - {test: enableConnected, weight: 1}
  - {test: outputA,         weight: 5}
  - {test: transition,      weight: 10}

Note that since hwX-tester.jar is in grading-files, it'll need to be in grading-files/ in the zucchini assignment repository.

circuitsim-tester's People

Contributors

ausbin avatar kanwren avatar cgokmen avatar mbrickell3 avatar sameer-s avatar manleyroberts avatar skyman avatar endorpersand avatar dummycode avatar adriencao avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.