38
Explain the implementation principle of the Jest framework in a simple way
This article mainly provides you with an in-depth understanding of the operating principles behind Jest, which is convenient for responding to interviews and actual business needs. I believe we are already familiar with the preparation of Jest, but we may be very unfamiliar with how Jest works. Let us walk into Jest together. Inwardly, explore together. First attach the code to students in need, welcome to pay attention: https://github.com/Wscats/jest-tutorial
Jest is a Javascript testing framework developed by Facebook. It is a JavaScript library for creating, running and writing tests.
Jest is released as an NPM package and can be installed and run in any JavaScript project. Jest is currently one of the most popular test libraries for the front-end.
In technical terms, testing means checking whether our code meets certain expectations. For example: a function called sum (sum
) should return the expected output given some operation result.
There are many types of tests, and you will soon be overwhelmed by the terminology, but the long story short tests fall into three categories:
- unit test
- Integration Testing
- E2E test
In terms of testing, even the simplest code block may confuse beginners. The most common question is "how do I know what to test?".
If you are writing a web page, a good starting point is to test every page of the application and every user interaction. However, the web page also needs to be composed of code units such as functions and modules to be tested.
There are two situations most of the time:
- You inherit the legacy code, which has no built-in tests
- You must implement a new feature out of thin air
so what should I do now? In both cases, you can think of the test as: checking whether the function produces the expected result. The most typical test process is as follows:
- Import the function to be tested
- Give the function an input
- Define the desired output
- Check if the function produces the expected output
Generally, it's that simple. Master the following core ideas, writing tests will no longer be scary:
Input -> Expected output -> Assertion result.
We will create a simple Javascript function code for the addition of 2 numbers and write a corresponding Jest-based test for it
const sum = (a, b) => a + b;
Now, for testing, create a test file in the same folder and name it test.spec.js
. This special suffix is a Jest convention and is used to find all test files. We will also import the function under test in order to execute the code under test. Jest tests follow the BDD style of tests. Each test should have a main test
test block, and there can be multiple test blocks. Now you can write test blocks for the sum
method. Here we write a test to add 2 Number and verify the expected result. We will provide the numbers 1 and 2, and expect 3 to be output.
test
It requires two parameters: a string to describe the test block, and a callback function to wrap the actual test. expect
wraps the objective function and combines it with the matcher toBe
to check whether the calculation result of the function meets expectations.
This is the complete test:
test("sum test", () => {
expect(sum(1, 2)).toBe(3);
});
We observe the above code and find two points:
The test
block is a separate test block, which has the function of describing and dividing the scope, that is, it represents a general container for the test we want to write for the calculation function sum
. -expect
is an assertion. This statement uses inputs 1 and 2 to call the sum
method in the function under test, and expects an output of 3. -toBe
is a matcher, used to check the expected value, if the expected result is not met, an exception should be thrown.
The test block is actually not complicated. The simplest implementation is as follows. We need to store the callback function of the actual test of the test package, so we encapsulate a dispatch
method to receive the command type and the callback function:
const test = (name, fn) => {
dispatch({ type: "ADD_TEST", fn, name });
};
We need to create a callback function called state
globally to save the test. The callback function of the test is stored in an array.
global["STATE_SYMBOL"] = {
testBlock: [],
};
The dispatch
method only needs to identify the corresponding commands at this time, and store the test callback function in the global state
.
const dispatch = (event) => {
const { fn, type, name } = event;
switch (type) {
case "ADD_TEST":
const { testBlock } = global["STATE_SYMBOL"];
testBlock.push({ fn, name });
break;
}
};
The assertion library is also very simple to implement. You only need to encapsulate a function to expose the matcher method to satisfy the following formula:
expect(A).toBe(B)
Here we implement the commonly used method toBe
, when the result is not equal to the expectation, just throw an error:
const expect = (actual) => ({
toBe(expected) {
if (actual !== expected) {
throw new Error(`${actual} is not equal to ${expected}`);
}
}
};
Actually, try/catch is used in the test block to catch errors and print stack information to locate the problem.
In simple cases, we can also use the assert
module that comes with Node to make assertions. Of course, there are many more complex assertion methods, and the principles are similar in essence.
After writing the test, we need to enter the command in the command line to run the single test. Normally, the command is similar to the following:
node jest xxx.spec.js
The essence here is to parse the parameters of the command line.
const testPath = process.argv.slice(2)[0];
const code = fs.readFileSync(path.join(process.cwd(), testPath)).toString();
In complex situations, you may also need to read the parameters of the local Jest configuration file to change the execution environment, etc. Here, Jest uses third-party libraries yargs
, execa
and chalk
, etc. to parse, execute and print commands.
In complex test scenarios, we must not avoid a Jest term: mock (mock
)
In the Jest documentation, we can find that Jest has the following description of simulation: "The simulation function erases the actual implementation of the function, captures the call to the function, and the parameters passed in these calls, so that the link between the test codes becomes easy"
In short, a simulation can be created by assigning the following code snippets to functions or dependencies:
jest.mock("fs", {
readFile: jest.fn(() => "wscats"),
});
This is a simple simulation example that simulates the return value of the readFile function of the fs module in testing specific business logic.
Next, we will study how to implement it. The first is jest.mock
. Its first parameter accepts the module name or module path, and the second parameter is the specific implementation of the module’s external exposure method.
const jest = {
mock(mockPath, mockExports = {}) {
const path = require.resolve(mockPath, { paths: ["."] });
require.cache[path] = {
id: path,
filename: path,
loaded: true,
exports: mockExports,
};
},
};
Our solution is actually the same as the implementation of the above test
test block. You only need to find a place to save the specific implementation method, and replace it when the module is actually used later, so we save it in require In .cache
, of course we can also store it in the global state
.
The implementation of jest.fn
is not difficult. Here we use a closure mockFn
to store the replaced functions and parameters, which is convenient for subsequent test inspections and statistics of call data.
const jest = {
fn(impl = () => {}) {
const mockFn = (...args) => {
mockFn.mock.calls.push(args);
return impl(...args);
};
mockFn.originImpl = impl;
mockFn.mock = { calls: [] };
return mockFn;
},
};
Some students may have noticed that in the testing framework, we don’t need to manually introduce the functions of test
, expect
and jest
. Each test file can be used directly, so we need to create a run that injects these methods here. surroundings.
Since everything is ready, we only need to inject the methods required for testing into the V8 virtual machine, that is, inject the testing scope.
const context = {
console: console.Console({ stdout: process.stdout, stderr: process.stderr }),
jest,
expect,
require,
test: (name, fn) => dispatch({ type: "ADD_TEST", fn, name }),
};
After injecting the scope, we can make the code of the test file run in the V8 virtual machine. The code I passed here is the code that has been processed into a string. Jest will do some code processing, security processing and SourceMap here. For sewing and other operations, our example does not need to be so complicated.
vm.runInContext(code, context);
Before and after the code is executed, the time difference can be used to calculate the running time of a single test. Jest will also pre-evaluate the size and number of single test files here, and decide whether to enable Worker to optimize the execution speed.
const start = new Date();
const end = new Date();
log("\x1b[32m%s\x1b[0m", `Time: ${end - start}ms`);
After the execution of the V8 virtual machine is completed, the global state
will collect all the packaged test callback functions in the test block. Finally, we only need to traverse all these callback functions and execute them.
testBlock.forEach(async (item) => {
const { fn, name } = item;
try {
await fn.apply(this);
log("\x1b[32m%s\x1b[0m", `√ ${name} passed`);
} catch {
log("\x1b[32m%s\x1b[0m", `× ${name} error`);
}
});
We can also add life cycles to the single test execution process, such as hook functions such as beforeEach
, afterEach
, afterAll
and beforeAll
.
Adding the hook function to the above infrastructure is actually injecting the corresponding callback function in each process of executing the test. For example, beforeEach
is placed before the traversal execution test function of testBlock
, and afterEach
is placed on testBlock
After traversing the execution of the test function, it is very simple. You only need to put the right position to expose the hook function of any period.
testBlock.forEach(async (item) => {
const { fn, name } = item;
beforeEachBlock.forEach(async (beforeEach) => await beforeEach());
await fn.apply(this);
afterEachBlock.forEach(async (afterEach) => await afterEach());
});
And beforeAll
and afterAll
can be placed before and after all tests of testBlock
are completed.
beforeAllBlock.forEach(async (beforeAll) => await beforeAll());
testBlock.forEach(async (item) => {}) +
afterAllBlock.forEach(async (afterAll) => await afterAll());
At this point, we have implemented a simple test framework. Based on this, we can enrich the assertion method, matcher and support parameter configuration, and read the personal notes of the source code below.
Download Jest source code and execute it in the root directory
yarn
npm run build
It essentially runs two files build.js and buildTs.js in the script folder:
"scripts": {
"build": "yarn build:js && yarn build:ts",
"build:js": "node ./scripts/build.js",
"build:ts": "node ./scripts/buildTs.js",
}
build.js essentially uses the babel library, create a new build folder in the package/xxx package, and then use transformFileSync to generate the file into the build folder:
const transformed = babel.transformFileSync(file, options).code;
And buildTs.js essentially uses the tsc command to compile the ts file into the build folder, and use the execa library to execute the command:
const args = ["tsc", "-b", ...packagesWithTs, ...process.argv.slice(2)];
await execa("yarn", args, { stdio: "inherit" });
Successful execution will display as follows, it will help you compile all files js files and ts files in the packages folder to the build folder of the directory where you are:
Next we can start the jest command:
npm run jest
# Equivalent to
# node ./packages/jest-cli/bin/jest.js
Here you can do analysis processing according to the different parameters passed in, such as:
npm run jest -h
node ./packages/jest-cli/bin/jest.js /path/test.spec.js
It will execute the jest.js
file, and then enter the run method in the build/cli
file. The run method will parse various parameters in the command. The specific principle is that the yargs library cooperates with process.argv to achieve
const importLocal = require("import-local");
if (!importLocal(__filename)) {
if (process.env.NODE_ENV == null) {
process.env.NODE_ENV = "test";
}
require("../build/cli").run();
}
When various command parameters are obtained, the core method of runCLI
will be executed, which is the core method of the @jest/core -> packages/jest-core/src/cli/index.ts
library.
import { runCLI } from "@jest/core";
const outputStream = argv.json || argv.useStderr ? process.stderr : process.stdout;
const { results, globalConfig } = await runCLI(argv, projects);
The runCLI
method will use the input parameter argv parsed in the command just now to read the configuration file information with the readConfigs
method. readConfigs
comes from packages/jest-config/src/index.ts
, here There will be normalize to fill in and initialize some default configured parameters. Its default parameters are recorded in the packages/jest-config/src/Defaults.ts
file. For example, if you only run js single test, the default setting of require. resolve('jest-runner')
is a runner that runs a single test, and it also cooperates with the chalk library to generate an outputStream to output the content to the console.
By the way, let me mention the principle of introducing jest into the module. First, require.resolve(moduleName)
will find the path of the module, and save the path in the configuration, and then use the tool library packages/jest-util/src/requireOrImportModule The
requireOrImportModulemethod of .ts
calls the encapsulated native import/reqiure
method to match the path in the configuration file to take out the module.
- globalConfig configuration from argv
- configs are from the configuration of jest.config.js
const { globalConfig, configs, hasDeprecationWarnings } = await readConfigs(
argv,
projects
);
if (argv.debug) {
/*code*/
}
if (argv.showConfig) {
/*code*/
}
if (argv.clearCache) {
/*code*/
}
if (argv.selectProjects) {
/*code*/
}
jest-haste-map is used to get all the files in the project and the dependencies between them. It achieves this by looking at the import/require
calls, extracting them from each file and constructing a map containing each A file and its dependencies. Here Haste is the module system used by Facebook. It also has something called HasteContext, because it has HasteFS (Haste File System). HasteFS is just a list of files in the system and all dependencies associated with it. Item, it is a map data structure, where the key is the path and the value is the metadata. The contexts
generated here will be used until the onRunComplete
stage.
const { contexts, hasteMapInstances } = await buildContextsAndHasteMaps(
configs,
globalConfig,
outputStream
);
The _run10000
method will obtain contexts
according to the configuration information globalConfig
and configs
. contexts
will store the configuration information and path of each local file, etc., and then will bring the callback function onComplete
, the global configuration globalConfig
and scope contexts
enter the runWithoutWatch
method.
Next, you will enter the runJest
method of the packages/jest-core/src/runJest.ts
file, where the passed contexts
will be used to traverse all unit tests and save them in an array.
let allTests: Array<Test> = [];
contexts.map(async (context, index) => {
const searchSource = searchSources[index];
const matches = await getTestPaths(
globalConfig,
searchSource,
outputStream,
changedFilesPromise && (await changedFilesPromise),
jestHooks,
filter
);
allTests = allTests.concat(matches.tests);
return { context, matches };
});
And use the Sequencer
method to sort the single tests
const Sequencer: typeof TestSequencer = await requireOrImportModule(
globalConfig.testSequencer
);
const sequencer = new Sequencer();
allTests = await sequencer.sort(allTests);
The runJest
method calls a key method packages/jest-core/src/TestScheduler.ts
's scheduleTests
method.
const results = await new TestScheduler(
globalConfig,
{ startRun },
testSchedulerContext
).scheduleTests(allTests, testWatcher);
The scheduleTests
method will do a lot of things, it will collect the contexts
in the allTests
into the contexts
, collect the duration
into the timings
array, and subscribe to four life cycles before executing all single tests :
- test-file-start
- test-file-success
- test-file-failure
- test-case-result
Then traverse the contexts
and use a new empty object testRunners
to do some processing and save it, which will call the createScriptTransformer
method provided by @jest/transform
to process the imported modules.
import { createScriptTransformer } from "@jest/transform";
const transformer = await createScriptTransformer(config);
const Runner: typeof TestRunner = interopRequireDefault(
transformer.requireAndTranspileModule(config.runner)
).default;
const runner = new Runner(this._globalConfig, {
changedFiles: this._context?.changedFiles,
sourcesRelatedToTestsInChangedFiles: this._context?.sourcesRelatedToTestsInChangedFiles,
});
testRunners[config.runner] = runner;
The scheduleTests
method will call the runTests
method of packages/jest-runner/src/index.ts
.
async runTests(tests, watcher, onStart, onResult, onFailure, options) {
return await (options.serial
? this._createInBandTestRun(tests, watcher, onStart, onResult, onFailure)
: this._createParallelTestRun(
tests,
watcher,
onStart,
onResult,
onFailure
));
}
In the final _createParallelTestRun
or _createInBandTestRun
method:
There will be a runTestInWorker
method, which, as the name suggests, is to perform a single test in the worker.
_createInBandTestRun
will execute a core method runTest
in packages/jest-runner/src/runTest.ts
, and execute a method runTestInternal
in runJest
, which will prepare a lot of preparations before executing a single test The thing involves global method rewriting and hijacking of import and export methods.
await this.eventEmitter.emit("test-file-start", [test]);
return runTest(
test.path,
this._globalConfig,
test.context.config,
test.context.resolver,
this._context,
sendMessageToJest
);
In the runTestInternal
method, the fs
module will be used to read the content of the file and put it into cacheFS
, which can be cached for quick reading later. For example, if the content of the file is json later, it can be read directly in cacheFS
. Also use Date.now
time difference to calculate time-consuming.
const testSource = fs().readFileSync(path, "utf8");
const cacheFS = new Map([[path, testSource]]);
In the runTestInternal
method, packages/jest-runtime/src/index.ts
will be introduced, which will help you cache and read modules and trigger execution.
const runtime = new Runtime(
config,
environment,
resolver,
transformer,
cacheFS,
{
changedFiles: context?.changedFiles,
collectCoverage: globalConfig.collectCoverage,
collectCoverageFrom: globalConfig.collectCoverageFrom,
collectCoverageOnlyFrom: globalConfig.collectCoverageOnlyFrom,
coverageProvider: globalConfig.coverageProvider,
sourcesRelatedToTestsInChangedFiles: context?.sourcesRelatedToTestsInChangedFiles,
},
path
);
Here, the @jest/console
package is used to rewrite the global console. In order for the console of the single-tested file code block to print the results on the node terminal smoothly, in conjunction with the jest-environment-node
package, set the global environment.global
all Rewritten to facilitate subsequent methods to get these scopes in vm.
// Essentially it is rewritten using node's console to facilitate subsequent overwriting of the console method in the vm scope
testConsole = new BufferedConsole();
const environment = new TestEnvironment(config, {
console: testConsole, // Suspected useless code
docblockPragmas,
testPath: path,
});
// Really rewrite the console method
setGlobal(environment.global, "console", testConsole);
runtime
mainly uses these two methods to load the module, first judge whether it is an ESM module, if it is, use runtime.unstable_importModule
to load the module and run the module, if not, use runtime.requireModule
to load the module and run the module .
const esm = runtime.unstable_shouldLoadAsEsm(path);
if (esm) {
await runtime.unstable_importModule(path);
} else {
runtime.requireModule(path);
}
Immediately after the testFramework
in runTestInternal
will accept the incoming runtime to call the single test file to run, the testFramework
method comes from a library with an interesting name packages/jest-circus/src/legacy-code-todo-rewrite /jestAdapter.ts
, where legacy-code-todo-rewrite
means legacy code todo rewrite, jest-circus
mainly rewrites some methods of global global
, involving These few:
- afterAll
- afterEach
- beforeAll
- beforeEach
- describe
- it
- test
Before calling the single test here, the jestAdapter
function, which is the above-mentioned runtime.requireModule
, will load the xxx.spec.js
file. The execution environment globals
has been preset using initialize
before execution. And
snapshotState, and rewrite
beforeEach. If
resetModules,
clearMocks,
resetMocks,
restoreMocksand
setupFilesAfterEnv` are configured, the following methods will be executed respectively:
- runtime.resetModules
- runtime.clearAllMocks
- runtime.resetAllMocks
- runtime.restoreAllMocks
- runtime.requireModule or runtime.unstable_importModule
After running the initialization of the initialize
method, because initialize
has rewritten the global describe
and test
methods, these methods are all rewritten here in /packages/jest-circus/src/index.ts
, here Note that there is a dispatchSync
method in the test
method. This is a key method. Here, a copy of state
will be maintained globally. dispatchSync
means to store the functions and other information in the test
code block in the state. In
dispatchSync uses
name in conjunction with the
eventHandler method to modify the
state`. This idea is very similar to the data flow in redux.
const test: Global.It = () => {
return (test = (testName, fn, timeout) => (testName, mode, fn, testFn, timeout) => {
return dispatchSync({
asyncError,
fn,
mode,
name: "add_test",
testName,
timeout,
});
});
};
The single test xxx.spec.js
, that is, the testPath file will be imported and executed after the initialize
. Note that this single test will be executed when imported here, because the single test xxx.spec.js
file is written according to the specifications , There will be code blocks such as test
and describe
, so at this time all callback functions accepted by test
and describe
will be stored in the global state
.
const esm = runtime.unstable_shouldLoadAsEsm(testPath);
if (esm) {
await runtime.unstable_importModule(testPath);
} else {
runtime.requireModule(testPath);
}
Here, it will first determine whether it is an esm module, if it is, use the method of unstable_importModule
to import it, otherwise use the method of requireModule
to import it, specifically will it enter the following function.
this._loadModule(localModule, from, moduleName, modulePath, options, moduleRegistry);
The logic of _loadModule has only three main parts
- Judge whether it is a json suffix file, execute readFile to read the text, and use transformJson and JSON.parse to transform the output content.
- Determine whether the node suffix file is, and execute the require native method to import the module.
- For files that do not meet the above two conditions, execute the _execModule execution module.
_execModule will use babel to transform the source code read by fs. This transformFile
is the transform
method of packages/jest-runtime/src/index.ts
.
const transformedCode = this.transformFile(filename, options);
_execModule will use the createScriptFromCode
method to call node's native vm module to actually execute js. The vm module accepts safe source code, and uses the V8 virtual machine with the incoming context to execute the code immediately or delay the execution of the code, here you can Accept different scopes to execute the same code to calculate different results, which is very suitable for the use of test frameworks. The injected vmContext here is the above global rewrite scope including afterAll, afterEach, beforeAll, beforeEach, describe, it, test, So our single test code will get these methods with injection scope when it runs.
const vm = require("vm");
const script = new vm().Script(scriptSourceCode, option);
const filename = module.filename;
const vmContext = this._environment.getVmContext();
script.runInContext(vmContext, {
filename,
});
When the global method is overwritten and the state
is saved above, it will enter the logic of the callback function that actually executes the describe
, in the run
method of packages/jest-circus/src/run.ts
, here Use the getState
method to take out the describe
code block, then use the _runTestsForDescribeBlock
to execute this function, then enter the _runTest
method, and then use the hook function before and after the execution of _callCircusHook
, and use the _callCircusTest
to execute.
const run = async (): Promise<Circus.RunResult> => {
const { rootDescribeBlock } = getState();
await dispatch({ name: "run_start" });
await _runTestsForDescribeBlock(rootDescribeBlock);
await dispatch({ name: "run_finish" });
return makeRunResult(getState().rootDescribeBlock, getState().unhandledErrors);
};
const _runTest = async (test, parentSkipped) => {
// beforeEach
// test function block, testContext scope
await _callCircusTest(test, testContext);
// afterEach
};
This is the core position of the hook function implementation and also the core element of the Jest function.
I hope this article can help you understand the core implementation and principles of the Jest testing framework. Thank you for reading patiently. If the articles and notes can bring you a hint of help or inspiration, please don’t be stingy with your Star and Fork. The articles are continuously updated synchronously, your affirmation Is my biggest motivation to move forward😁
38