Source Code: lib/vm.js
\nThe vm module enables compiling and running code within V8 Virtual\nMachine contexts. The vm module is not a security mechanism. Do\nnot use it to run untrusted code.
JavaScript code can be compiled and run immediately or\ncompiled, saved, and run later.
\nA common use case is to run the code in a different V8 Context. This means\ninvoked code has a different global object than the invoking code.
\nOne can provide the context by contextifying an\nobject. The invoked code treats any property in the context like a\nglobal variable. Any changes to global variables caused by the invoked\ncode are reflected in the context object.
\nconst vm = require('vm');\n\nconst x = 1;\n\nconst context = { x: 2 };\nvm.createContext(context); // Contextify the object.\n\nconst code = 'x += 40; var y = 17;';\n// `x` and `y` are global variables in the context.\n// Initially, x has the value 2 because that is the value of context.x.\nvm.runInContext(code, context);\n\nconsole.log(context.x); // 42\nconsole.log(context.y); // 17\n\nconsole.log(x); // 1; y is not defined.\nInstances of the vm.Script class contain precompiled scripts that can be\nexecuted in specific contexts.
Creates a code cache that can be used with the Script constructor's\ncachedData option. Returns a Buffer. This method may be called at any\ntime and any number of times.
const script = new vm.Script(`\nfunction add(a, b) {\n return a + b;\n}\n\nconst x = add(1, 2);\n`);\n\nconst cacheWithoutX = script.createCachedData();\n\nscript.runInThisContext();\n\nconst cacheWithX = script.createCachedData();\nRuns the compiled code contained by the vm.Script object within the given\ncontextifiedObject and returns the result. Running code does not have access\nto local scope.
The following example compiles code that increments a global variable, sets\nthe value of another global variable, then execute the code multiple times.\nThe globals are contained in the context object.
const vm = require('vm');\n\nconst context = {\n animal: 'cat',\n count: 2\n};\n\nconst script = new vm.Script('count += 1; name = \"kitty\";');\n\nvm.createContext(context);\nfor (let i = 0; i < 10; ++i) {\n script.runInContext(context);\n}\n\nconsole.log(context);\n// Prints: { animal: 'cat', count: 12, name: 'kitty' }\nUsing the timeout or breakOnSigint options will result in new event loops\nand corresponding threads being started, which have a non-zero performance\noverhead.
First contextifies the given contextObject, runs the compiled code contained\nby the vm.Script object within the created context, and returns the result.\nRunning code does not have access to local scope.
The following example compiles code that sets a global variable, then executes\nthe code multiple times in different contexts. The globals are set on and\ncontained within each individual context.
const vm = require('vm');\n\nconst script = new vm.Script('globalVar = \"set\"');\n\nconst contexts = [{}, {}, {}];\ncontexts.forEach((context) => {\n script.runInNewContext(context);\n});\n\nconsole.log(contexts);\n// Prints: [{ globalVar: 'set' }, { globalVar: 'set' }, { globalVar: 'set' }]\nRuns the compiled code contained by the vm.Script within the context of the\ncurrent global object. Running code does not have access to local scope, but\ndoes have access to the current global object.
The following example compiles code that increments a global variable then\nexecutes that code multiple times:
const vm = require('vm');\n\nglobal.globalVar = 0;\n\nconst script = new vm.Script('globalVar += 1', { filename: 'myfile.vm' });\n\nfor (let i = 0; i < 1000; ++i) {\n script.runInThisContext();\n}\n\nconsole.log(globalVar);\n\n// 1000\nIf options is a string, then it specifies the filename.
Creating a new vm.Script object compiles code but does not run it. The\ncompiled vm.Script can be run later multiple times. The code is not bound to\nany global object; rather, it is bound before each run, just for that run.
This feature is only available with the --experimental-vm-modules command\nflag enabled.
The vm.Module class provides a low-level interface for using\nECMAScript modules in VM contexts. It is the counterpart of the vm.Script\nclass that closely mirrors Module Records as defined in the ECMAScript\nspecification.
Unlike vm.Script however, every vm.Module object is bound to a context from\nits creation. Operations on vm.Module objects are intrinsically asynchronous,\nin contrast with the synchronous nature of vm.Script objects. The use of\n'async' functions can help with manipulating vm.Module objects.
Using a vm.Module object requires three distinct steps: creation/parsing,\nlinking, and evaluation. These three steps are illustrated in the following\nexample.
This implementation lies at a lower level than the ECMAScript Module\nloader. There is also no way to interact with the Loader yet, though\nsupport is planned.
\nconst vm = require('vm');\n\nconst contextifiedObject = vm.createContext({\n secret: 42,\n print: console.log,\n});\n\n(async () => {\n // Step 1\n //\n // Create a Module by constructing a new `vm.SourceTextModule` object. This\n // parses the provided source text, throwing a `SyntaxError` if anything goes\n // wrong. By default, a Module is created in the top context. But here, we\n // specify `contextifiedObject` as the context this Module belongs to.\n //\n // Here, we attempt to obtain the default export from the module \"foo\", and\n // put it into local binding \"secret\".\n\n const bar = new vm.SourceTextModule(`\n import s from 'foo';\n s;\n print(s);\n `, { context: contextifiedObject });\n\n // Step 2\n //\n // \"Link\" the imported dependencies of this Module to it.\n //\n // The provided linking callback (the \"linker\") accepts two arguments: the\n // parent module (`bar` in this case) and the string that is the specifier of\n // the imported module. The callback is expected to return a Module that\n // corresponds to the provided specifier, with certain requirements documented\n // in `module.link()`.\n //\n // If linking has not started for the returned Module, the same linker\n // callback will be called on the returned Module.\n //\n // Even top-level Modules without dependencies must be explicitly linked. The\n // callback provided would never be called, however.\n //\n // The link() method returns a Promise that will be resolved when all the\n // Promises returned by the linker resolve.\n //\n // Note: This is a contrived example in that the linker function creates a new\n // \"foo\" module every time it is called. In a full-fledged module system, a\n // cache would probably be used to avoid duplicated modules.\n\n async function linker(specifier, referencingModule) {\n if (specifier === 'foo') {\n return new vm.SourceTextModule(`\n // The \"secret\" variable refers to the global variable we added to\n // \"contextifiedObject\" when creating the context.\n export default secret;\n `, { context: referencingModule.context });\n\n // Using `contextifiedObject` instead of `referencingModule.context`\n // here would work as well.\n }\n throw new Error(`Unable to resolve dependency: ${specifier}`);\n }\n await bar.link(linker);\n\n // Step 3\n //\n // Evaluate the Module. The evaluate() method returns a promise which will\n // resolve after the module has finished evaluating.\n\n // Prints 42.\n await bar.evaluate();\n})();\nThe specifiers of all dependencies of this module. The returned array is frozen\nto disallow any changes to it.
\nCorresponds to the [[RequestedModules]] field of Cyclic Module Records in\nthe ECMAScript specification.
If the module.status is 'errored', this property contains the exception\nthrown by the module during evaluation. If the status is anything else,\naccessing this property will result in a thrown exception.
The value undefined cannot be used for cases where there is not a thrown\nexception due to possible ambiguity with throw undefined;.
Corresponds to the [[EvaluationError]] field of Cyclic Module Records\nin the ECMAScript specification.
The namespace object of the module. This is only available after linking\n(module.link()) has completed.
Corresponds to the GetModuleNamespace abstract operation in the ECMAScript\nspecification.
" }, { "textRaw": "`status` {string}", "type": "string", "name": "status", "desc": "The current status of the module. Will be one of:
\n'unlinked': module.link() has not yet been called.
'linking': module.link() has been called, but not all Promises returned\nby the linker function have been resolved yet.
'linked': The module has been linked successfully, and all of its\ndependencies are linked, but module.evaluate() has not yet been called.
'evaluating': The module is being evaluated through a module.evaluate() on\nitself or a parent module.
'evaluated': The module has been successfully evaluated.
'errored': The module has been evaluated, but an exception was thrown.
Other than 'errored', this status string corresponds to the specification's\nCyclic Module Record's [[Status]] field. 'errored' corresponds to\n'evaluated' in the specification, but with [[EvaluationError]] set to a\nvalue that is not undefined.
The identifier of the current module, as set in the constructor.
" } ], "methods": [ { "textRaw": "`module.evaluate([options])`", "type": "method", "name": "evaluate", "signatures": [ { "return": { "textRaw": "Returns: {Promise}", "name": "return", "type": "Promise" }, "params": [ { "textRaw": "`options` {Object}", "name": "options", "type": "Object", "options": [ { "textRaw": "`timeout` {integer} Specifies the number of milliseconds to evaluate before terminating execution. If execution is interrupted, an [`Error`][] will be thrown. This value must be a strictly positive integer.", "name": "timeout", "type": "integer", "desc": "Specifies the number of milliseconds to evaluate before terminating execution. If execution is interrupted, an [`Error`][] will be thrown. This value must be a strictly positive integer." }, { "textRaw": "`breakOnSigint` {boolean} If `true`, the execution will be terminated when `SIGINT` (Ctrl+C) is received. Existing handlers for the event that have been attached via `process.on('SIGINT')` will be disabled during script execution, but will continue to work after that. If execution is interrupted, an [`Error`][] will be thrown. **Default:** `false`.", "name": "breakOnSigint", "type": "boolean", "default": "`false`", "desc": "If `true`, the execution will be terminated when `SIGINT` (Ctrl+C) is received. Existing handlers for the event that have been attached via `process.on('SIGINT')` will be disabled during script execution, but will continue to work after that. If execution is interrupted, an [`Error`][] will be thrown." } ] } ] } ], "desc": "Evaluate the module.
\nThis must be called after the module has been linked; otherwise it will reject.\nIt could be called also when the module has already been evaluated, in which\ncase it will either do nothing if the initial evaluation ended in success\n(module.status is 'evaluated') or it will re-throw the exception that the\ninitial evaluation resulted in (module.status is 'errored').
This method cannot be called while the module is being evaluated\n(module.status is 'evaluating').
Corresponds to the Evaluate() concrete method field of Cyclic Module\nRecords in the ECMAScript specification.
" }, { "textRaw": "`module.link(linker)`", "type": "method", "name": "link", "signatures": [ { "return": { "textRaw": "Returns: {Promise}", "name": "return", "type": "Promise" }, "params": [ { "textRaw": "`linker` {Function}", "name": "linker", "type": "Function", "options": [ { "textRaw": "`specifier` {string} The specifier of the requested module:```js import foo from 'foo'; // ^^^^^ the module specifier ```", "name": "specifier", "type": "string", "desc": "The specifier of the requested module:```js import foo from 'foo'; // ^^^^^ the module specifier ```" }, { "textRaw": "`referencingModule` {vm.Module} The `Module` object `link()` is called on.", "name": "referencingModule", "type": "vm.Module", "desc": "The `Module` object `link()` is called on." }, { "textRaw": "Returns: {vm.Module|Promise}", "name": "return", "type": "vm.Module|Promise" } ] } ] } ], "desc": "Link module dependencies. This method must be called before evaluation, and\ncan only be called once per module.
\nThe function is expected to return a Module object or a Promise that\neventually resolves to a Module object. The returned Module must satisfy the\nfollowing two invariants:
Module.status must not be 'errored'.If the returned Module's status is 'unlinked', this method will be\nrecursively called on the returned Module with the same provided linker\nfunction.
link() returns a Promise that will either get resolved when all linking\ninstances resolve to a valid Module, or rejected if the linker function either\nthrows an exception or returns an invalid Module.
The linker function roughly corresponds to the implementation-defined\nHostResolveImportedModule abstract operation in the ECMAScript\nspecification, with a few key differences:
\nThe actual HostResolveImportedModule implementation used during module\nlinking is one that returns the modules linked during linking. Since at\nthat point all modules would have been fully linked already, the\nHostResolveImportedModule implementation is fully synchronous per\nspecification.
\nCorresponds to the Link() concrete method field of Cyclic Module\nRecords in the ECMAScript specification.
" } ] }, { "textRaw": "Class: `vm.SourceTextModule`", "type": "class", "name": "vm.SourceTextModule", "meta": { "added": [ "v9.6.0" ], "changes": [] }, "stability": 1, "stabilityText": "Experimental", "desc": "This feature is only available with the --experimental-vm-modules command\nflag enabled.
The vm.SourceTextModule class provides the Source Text Module Record as\ndefined in the ECMAScript specification.
Creates a code cache that can be used with the SourceTextModule constructor's\ncachedData option. Returns a Buffer. This method may be called any number\nof times before the module has been evaluated.
// Create an initial module\nconst module = new vm.SourceTextModule('const a = 1;');\n\n// Create cached data from this module\nconst cachedData = module.createCachedData();\n\n// Create a new module using the cached data. The code must be the same.\nconst module2 = new vm.SourceTextModule('const a = 1;', { cachedData });\nCreates a new SourceTextModule instance.
Properties assigned to the import.meta object that are objects may\nallow the module to access information outside the specified context. Use\nvm.runInContext() to create objects in a specific context.
const vm = require('vm');\n\nconst contextifiedObject = vm.createContext({ secret: 42 });\n\n(async () => {\n const module = new vm.SourceTextModule(\n 'Object.getPrototypeOf(import.meta.prop).secret = secret;',\n {\n initializeImportMeta(meta) {\n // Note: this object is created in the top context. As such,\n // Object.getPrototypeOf(import.meta.prop) points to the\n // Object.prototype in the top context rather than that in\n // the contextified object.\n meta.prop = {};\n }\n });\n // Since module has no dependencies, the linker function will never be called.\n await module.link(() => {});\n await module.evaluate();\n\n // Now, Object.prototype.secret will be equal to 42.\n //\n // To fix this problem, replace\n // meta.prop = {};\n // above with\n // meta.prop = vm.runInContext('{}', contextifiedObject);\n})();\nThis feature is only available with the --experimental-vm-modules command\nflag enabled.
The vm.SyntheticModule class provides the Synthetic Module Record as\ndefined in the WebIDL specification. The purpose of synthetic modules is to\nprovide a generic interface for exposing non-JavaScript sources to ECMAScript\nmodule graphs.
const vm = require('vm');\n\nconst source = '{ \"a\": 1 }';\nconst module = new vm.SyntheticModule(['default'], function() {\n const obj = JSON.parse(source);\n this.setExport('default', obj);\n});\n\n// Use `module` in linking...\nThis method is used after the module is linked to set the values of exports. If\nit is called before the module is linked, an ERR_VM_MODULE_STATUS error\nwill be thrown.
const vm = require('vm');\n\n(async () => {\n const m = new vm.SyntheticModule(['x'], () => {\n m.setExport('x', 1);\n });\n\n await m.link(() => {});\n await m.evaluate();\n\n assert.strictEqual(m.namespace.x, 1);\n})();\nCreates a new SyntheticModule instance.
Objects assigned to the exports of this instance may allow importers of\nthe module to access information outside the specified context. Use\nvm.runInContext() to create objects in a specific context.
Measure the memory known to V8 and used by all contexts known to the\ncurrent V8 isolate, or the main context.
\noptions <Object> Optional.
mode <string> Either 'summary' or 'detailed'. In summary mode,\nonly the memory measured for the main context will be returned. In\ndetailed mode, the measure measured for all contexts known to the\ncurrent V8 isolate will be returned.\nDefault: 'summary'execution <string> Either 'default' or 'eager'. With default\nexecution, the promise will not resolve until after the next scheduled\ngarbage collection starts, which may take a while (or never if the program\nexits before the next GC). With eager execution, the GC will be started\nright away to measure the memory.\nDefault: 'default'The format of the object that the returned Promise may resolve with is\nspecific to the V8 engine and may change from one version of V8 to the next.
\nThe returned result is different from the statistics returned by\nv8.getHeapSpaceStatistics() in that vm.measureMemory() measure the\nmemory reachable by each V8 specific contexts in the current instance of\nthe V8 engine, while the result of v8.getHeapSpaceStatistics() measure\nthe memory occupied by each heap space in the current V8 instance.
const vm = require('vm');\n// Measure the memory used by the main context.\nvm.measureMemory({ mode: 'summary' })\n // This is the same as vm.measureMemory()\n .then((result) => {\n // The current format is:\n // {\n // total: {\n // jsMemoryEstimate: 2418479, jsMemoryRange: [ 2418479, 2745799 ]\n // }\n // }\n console.log(result);\n });\n\nconst context = vm.createContext({ a: 1 });\nvm.measureMemory({ mode: 'detailed', execution: 'eager' })\n .then((result) => {\n // Reference the context here so that it won't be GC'ed\n // until the measurement is complete.\n console.log(context.a);\n // {\n // total: {\n // jsMemoryEstimate: 2574732,\n // jsMemoryRange: [ 2574732, 2904372 ]\n // },\n // current: {\n // jsMemoryEstimate: 2438996,\n // jsMemoryRange: [ 2438996, 2768636 ]\n // },\n // other: [\n // {\n // jsMemoryEstimate: 135736,\n // jsMemoryRange: [ 135736, 465376 ]\n // }\n // ]\n // }\n console.log(result);\n });\nCompiles the given code into the provided context (if no context is\nsupplied, the current context is used), and returns it wrapped inside a\nfunction with the given params.
If given a contextObject, the vm.createContext() method will prepare\nthat object so that it can be used in calls to\nvm.runInContext() or script.runInContext(). Inside such scripts,\nthe contextObject will be the global object, retaining all of its existing\nproperties but also having the built-in objects and functions any standard\nglobal object has. Outside of scripts run by the vm module, global variables\nwill remain unchanged.
const vm = require('vm');\n\nglobal.globalVar = 3;\n\nconst context = { globalVar: 1 };\nvm.createContext(context);\n\nvm.runInContext('globalVar *= 2;', context);\n\nconsole.log(context);\n// Prints: { globalVar: 2 }\n\nconsole.log(global.globalVar);\n// Prints: 3\nIf contextObject is omitted (or passed explicitly as undefined), a new,\nempty contextified object will be returned.
The vm.createContext() method is primarily useful for creating a single\ncontext that can be used to run multiple scripts. For instance, if emulating a\nweb browser, the method can be used to create a single context representing a\nwindow's global object, then run all <script> tags together within that\ncontext.
The provided name and origin of the context are made visible through the\nInspector API.
Returns true if the given oject object has been contextified using\nvm.createContext().
The vm.runInContext() method compiles code, runs it within the context of\nthe contextifiedObject, then returns the result. Running code does not have\naccess to the local scope. The contextifiedObject object must have been\npreviously contextified using the vm.createContext() method.
If options is a string, then it specifies the filename.
The following example compiles and executes different scripts using a single\ncontextified object:
\nconst vm = require('vm');\n\nconst contextObject = { globalVar: 1 };\nvm.createContext(contextObject);\n\nfor (let i = 0; i < 10; ++i) {\n vm.runInContext('globalVar *= 2;', contextObject);\n}\nconsole.log(contextObject);\n// Prints: { globalVar: 1024 }\nThe vm.runInNewContext() first contextifies the given contextObject (or\ncreates a new contextObject if passed as undefined), compiles the code,\nruns it within the created context, then returns the result. Running code\ndoes not have access to the local scope.
If options is a string, then it specifies the filename.
The following example compiles and executes code that increments a global\nvariable and sets a new one. These globals are contained in the contextObject.
const vm = require('vm');\n\nconst contextObject = {\n animal: 'cat',\n count: 2\n};\n\nvm.runInNewContext('count += 1; name = \"kitty\"', contextObject);\nconsole.log(contextObject);\n// Prints: { animal: 'cat', count: 3, name: 'kitty' }\nvm.runInThisContext() compiles code, runs it within the context of the\ncurrent global and returns the result. Running code does not have access to\nlocal scope, but does have access to the current global object.
If options is a string, then it specifies the filename.
The following example illustrates using both vm.runInThisContext() and\nthe JavaScript eval() function to run the same code:
const vm = require('vm');\nlet localVar = 'initial value';\n\nconst vmResult = vm.runInThisContext('localVar = \"vm\";');\nconsole.log(`vmResult: '${vmResult}', localVar: '${localVar}'`);\n// Prints: vmResult: 'vm', localVar: 'initial value'\n\nconst evalResult = eval('localVar = \"eval\";');\nconsole.log(`evalResult: '${evalResult}', localVar: '${localVar}'`);\n// Prints: evalResult: 'eval', localVar: 'eval'\nBecause vm.runInThisContext() does not have access to the local scope,\nlocalVar is unchanged. In contrast, eval() does have access to the\nlocal scope, so the value localVar is changed. In this way\nvm.runInThisContext() is much like an indirect eval() call, e.g.\n(0,eval)('code').
When using either script.runInThisContext() or\nvm.runInThisContext(), the code is executed within the current V8 global\ncontext. The code passed to this VM context will have its own isolated scope.
In order to run a simple web server using the http module the code passed to\nthe context must either call require('http') on its own, or have a reference\nto the http module passed to it. For instance:
'use strict';\nconst vm = require('vm');\n\nconst code = `\n((require) => {\n const http = require('http');\n\n http.createServer((request, response) => {\n response.writeHead(200, { 'Content-Type': 'text/plain' });\n response.end('Hello World\\\\n');\n }).listen(8124);\n\n console.log('Server running at http://127.0.0.1:8124/');\n})`;\n\nvm.runInThisContext(code)(require);\nThe require() in the above case shares the state with the context it is\npassed from. This may introduce risks when untrusted code is executed, e.g.\naltering objects in the context in unwanted ways.
All JavaScript executed within Node.js runs within the scope of a \"context\".\nAccording to the V8 Embedder's Guide:
\n\n\nIn V8, a context is an execution environment that allows separate, unrelated,\nJavaScript applications to run in a single instance of V8. You must explicitly\nspecify the context in which you want any JavaScript code to be run.
\n
When the method vm.createContext() is called, the contextObject argument\n(or a newly-created object if contextObject is undefined) is associated\ninternally with a new instance of a V8 Context. This V8 Context provides the\ncode run using the vm module's methods with an isolated global environment\nwithin which it can operate. The process of creating the V8 Context and\nassociating it with the contextObject is what this document refers to as\n\"contextifying\" the object.
Promises and async functions can schedule tasks run by the JavaScript\nengine asynchronously. By default, these tasks are run after all JavaScript\nfunctions on the current stack are done executing.\nThis allows escaping the functionality of the timeout and\nbreakOnSigint options.
For example, the following code executed by vm.runInNewContext() with a\ntimeout of 5 milliseconds schedules an infinite loop to run after a promise\nresolves. The scheduled loop is never interrupted by the timeout:
const vm = require('vm');\n\nfunction loop() {\n console.log('entering loop');\n while (1) console.log(Date.now());\n}\n\nvm.runInNewContext(\n 'Promise.resolve().then(() => loop());',\n { loop, console },\n { timeout: 5 }\n);\n// This prints *before* 'entering loop' (!)\nconsole.log('done executing');\nThis can be addressed by passing microtaskMode: 'afterEvaluate' to the code\nthat creates the Context:
const vm = require('vm');\n\nfunction loop() {\n while (1) console.log(Date.now());\n}\n\nvm.runInNewContext(\n 'Promise.resolve().then(() => loop());',\n { loop, console },\n { timeout: 5, microtaskMode: 'afterEvaluate' }\n);\nIn this case, the microtask scheduled through promise.then() will be run\nbefore returning from vm.runInNewContext(), and will be interrupted\nby the timeout functionality. This applies only to code running in a\nvm.Context, so e.g. vm.runInThisContext() does not take this option.
Promise callbacks are entered into the microtask queue of the context in which\nthey were created. For example, if () => loop() is replaced with just loop\nin the above example, then loop will be pushed into the global microtask\nqueue, because it is a function from the outer (main) context, and thus will\nalso be able to escape the timeout.
If asynchronous scheduling functions such as process.nextTick(),\nqueueMicrotask(), setTimeout(), setImmediate(), etc. are made available\ninside a vm.Context, functions passed to them will be added to global queues,\nwhich are shared by all contexts. Therefore, callbacks passed to those functions\nare not controllable through the timeout either.