The Visual Studio Code editor has built-in debugging support for the Node.js runtime and can debug JavaScript, TypeScript, and many other languages that are transpiled into JavaScript. Setting up a project for Node.js debugging is usually straightforward with VS Code providing appropriate launch configuration defaults and snippets.
This Node.js debugging document goes into more detail about configurations and features for more advanced debugging scenarios. You’ll find instruction for debugging with source maps, stepping over external code, doing remote debugging, and much more.
Debugging configurations are stored in a launch.json
file located in your workspace’s .vscode
folder.
Below is a reference on launch.json
attributes specific to the Node.js debugger.
The following attributes are supported in launch configurations of type launch
and attach
:
protocol
- debug protocol to use.
port
- debug port to use.
address
- TCP/IP address of the debug port.
sourceMaps
- enable source maps by setting this to true
.
outFiles
- array of glob patterns for locating generated JavaScript files.
restart
- restart session on termination.
timeout
- when restarting a session, give up after this number of milliseconds.
stopOnEntry
- break immediately when the program launches.
localRoot
- VS Code’s root directory.
remoteRoot
- Node’s root directory.
smartStep
- try to automatically step over code that doesn’t map to source files.
skipFiles
- automatically skip files covered by these glob patterns.
trace
- enable diagnostic output.
These attributes are only available for launch configurations of request type launch
:
program
- an absolute path to the Node.js program to debug.args
- arguments passed to the program to debug. This attribute is of type array and expects individual arguments as array elements.cwd
- launch the program to debug in this directory.runtimeExecutable
- absolute path to the runtime executable to be used. Default is node
.runtimeArgs
- optional arguments passed to the runtime executable.runtimeVersion
- if “nvm” (or “nvm-windows”) or “nvs” is used for managing Node.js versions, this attribute can be used to select a specific version of Node.js.env
- optional environment variables. This attribute expects environment variables as a list of string typed key/value pairs.envFile
- optional path to a file containing environment variable definitions.console
- kind of console to launch the program (internalConsole
, integratedTerminal
, externalTerminal
).outputCapture
- if set to std
, output from the process stdout/stderr will be shown in the Debug Console, instead of listening to output over the debug port. This is useful for programs or log libraries that write directly to the stdout/stderr streams instead of using console.*
APIs.autoAttachChildProcesses
- track all subprocesses of debuggee and automatically attach to those that are launched in debug mode.This attribute is only available for launch configurations of request type attach
:
processId
- the debugger tries to attach to this process after having sent a USR1 signal. With this setting, the debugger can attach to an already running process that was not started in debug mode. When using the processId
attribute the debug port is determined automatically based on the Node.js version (and the used protocol) and cannot be configured explicitly. So don’t specify a port
attribute.You can trigger IntelliSense (Ctrl+Space) in your launch.json
file to see launch configuration snippets for commonly used Node.js debugging scenarios.
You can also bring up the snippets with the Add Configuration… button in the lower right of the launch.json
editor window.
Here is the list of all snippets:
address
attribute. Make sure that the Node.js program to debug has been started in debug mode and the debug port used is the same as the one specified in the snippet. To help VS Code mapping source files between your workspace and the filesystem of the remote host, make sure to specify correct paths for the localRoot
and remoteRoot
attributes.Ctrl+C
in the Integrated Terminal.test
folder of your project. Make sure that your project has ‘mocha’ installed in its node_modules
folder.node_modules
folder and that your generated project has been installed for debugging by running npm link
in the project folder.node_modules
folder.node_modules/.bin
directory of the workspace.By default, Node.js debug sessions launch the target in the internal VS Code Debug Console. Since the Debug Console does not support programs that need to read input from the console, you can enable either an external terminal or use the VS Code Integrated Terminal by setting the console
attribute in your launch configuration to externalTerminal
or integratedTerminal
respectively. The default is internalConsole
.
If an external terminal is used, you can configure which terminal program to use via the terminal.external.windowsExec
, terminal.external.osxExec
, and terminal.external.linuxExec
settings.
Instead of launching the Node.js program directly with node, you can use ‘npm’ scripts or other task runner tools directly from a launch configuration:
runtimeExecutable
attribute and arguments can be passed via runtimeArgs
.program
attribute if your npm script or other tool implicitly specifies the program to launch.port
attribute, VS Code will not automatically add the --inspect-brk=nnnn
attribute because the debug port is typically specified by the npm script or other tool as well.Let’s look at an ‘npm’ example. If your package.json
has a ‘debug’ script, for example:
"scripts": {
"debug": "node --nolazy --inspect-brk=9229 myProgram.js"
},
the corresponding launch configuration would look like this:
{
"name": "Launch via npm",
"type": "node",
"request": "launch",
"cwd": "${workspaceFolder}",
"runtimeExecutable": "npm",
"runtimeArgs": ["run-script", "debug"],
"port": 9229
}
If you are using ‘nvm’ (or ‘nvm-windows’) to manage your Node.js versions, it is possible to specify a runtimeVersion
attribute in a launch configuration for selecting a specific version of Node.js:
{
"type": "node",
"request": "launch",
"name": "Launch test",
"runtimeVersion": "7.10.1",
"program": "${workspaceFolder}/test.js"
}
If you are using ‘nvs’ to manage your Node.js versions, it is possible to use runtimeVersion
attribute to select a specific version of Node.js:
{
"type": "node",
"request": "launch",
"name": "Launch test",
"runtimeVersion": "chackracore/8.9.4/x64",
"program": "${workspaceFolder}/test.js"
}
Make sure to have those Node.js versions installed that you want to use with the runtimeVersion
attribute as the feature will not download and install the version itself. So you will have to run something like nvm install 7.10.1
or nvs add 7.10.1
from the integrated terminal if you plan to add "runtimeVersion": "7.10.1"
to your launch configuration.
Note: If VS Code detects that “nvs” is installed, it does not fall back to “nvm” if a specific Node.js version cannot be found in “nvs”. Using both “nvs” and “nvm” at the same time is not supported.
The VS Code Node debugger supports loading environment variables from a file and passing them to the Node.js runtime. To use this feature, add an attribute envFile
to your launch configuration and specify the absolute path to the file containing the environment variables:
//...
"envFile": "${workspaceFolder}/.env",
"env": { "USER": "john doe" }
//...
Any environment variable specified in the env
dictionary will override variables loaded from the file.
Here is an example of an .env
file:
USER=doe
PASSWORD=abc123
# a comment
# an empty value:
empty=
# new lines expanded in quoted strings:
lines="foo\nbar"
If you want to attach the VS Code debugger to a Node.js program, launch Node.js in VS Code’s integrated terminal as follows:
node --inspect program.js
or if the program should not start running but must wait for the debugger to attach:
node --inspect-brk program.js
Now you have three options for attaching the debugger to your program:
Let’s go through these options in detail:
If the Auto Attach feature is enabled, the Node debugger automatically attaches to Node.js processes that have been launched in debug mode from VS Code’s Integrated Terminal.
To enable the feature, either use the Toggle Auto Attach action or, if the Node debugger is already activated, use the Auto Attach Status Bar item.
After enabling Auto Attach the debugger should attach to your program within a second:
Whether or not a process is in “debug mode” is determined by analyzing the program arguments. Currently, we detect the patterns --inspect
, --inspect-brk
, --inspect-port
, --debug
, --debug-brk
, --debug-port
(all optionally followed by a ‘=’ and a port number).
Note: this feature does not (yet) work for terminal multiplexers like tmux (where launched processes are not children of VS Code’s integrated terminal).
The Attach to Node Process action opens a Quick Pick menu that lists all potential processes that are available to the Node.js debugger:
The individual processes listed in the picker show the debug port and the detected protocol in addition to the process ID. You should find your Node.js process in that list and after selecting it, the Node.js debugger tries to attach to it.
In addition to Node.js processes, the picker also shows other programs that were launched with one of the various forms of --debug
or --inspect
arguments. This makes it possible to attach to Electron’s or VS Code’s helper processes.
This option requires more work but in contrast to the previous two options it allows you to configure various debug configuration options explicitly.
The simplest “attach” configuration looks like this:
{
"name": "Attach to Process",
"type": "node",
"request": "attach",
"port": 9229
}
The port 9229
is the default debug port of the --inspect
and --inspect-brk
options. To use a different port (for example 12345
), add it to the options like this: --inspect=12345
and --inspect-brk=12345
and change the port
attribute in the launch configuration accordingly.
If you want to attach to a Node.js process that hasn’t been started in debug mode, you can do this by specifying the process ID of the Node.js process as a string:
{
"name": "Attach to Process",
"type": "node",
"request": "attach",
"processId": "53426"
}
Since it is a bit laborious to repeatedly find the process ID and enter it in the launch configuration, Node debug supports a command variable PickProcess
that binds to the process picker (from above) and that lets you conveniently pick the process from a list of Node.js processes.
By using the PickProcess
variable the launch configuration looks like this:
{
"name": "Attach to Process",
"type": "node",
"request": "attach",
"processId": "${command:PickProcess}"
}
Using the Debug: Stop action (available in the Debug toolbar or via the Command Palette) stops the debug session.
If the debug session was started in “attach” mode (and the red terminate button in the Debug toolbar shows a superimposed “plug”), pressing Stop disconnects the Node.js debugger from the debuggee that then continues execution.
If the debug session is in “launch” mode, pressing Stop does the following:
When pressing Stop for the first time, the debuggee is requested to shutdown gracefully by sending a SIGINT
signal. The debuggee is free to intercept this signal and clean up anything as necessary and then shut down. If there are no breakpoints (or problems) in that shutdown code, the debuggee and the debug session will terminate.
However if the debugger hits a breakpoint in the shutdown code or if the debuggee does not terminate properly by itself, then the debug session will not end. In this case, pressing Stop again will force terminate the debuggee and its child processes (SIGKILL
).
So if you see that a debug session doesn’t end when you press the red Stop button, then press the button again to force a shutdown of the debuggee.
Note that on the Windows operating system, pressing Stop always forcibly kills the debuggee and its child processes.
The Node.js debugger of VS Code supports JavaScript source maps that help debugging of transpiled languages, for example, TypeScript or minified/uglified JavaScript. With source maps, it is possible to single step through or set breakpoints in the original source. If no source map exists for the original source or if the source map is broken and cannot successfully map between the source and the generated JavaScript, then breakpoints show up as unverified (gray hollow circles).
Source maps can be generated with two kinds of inlining:
VS Code supports both the inlined source maps and the inlined source.
The source map feature is controlled by the sourceMaps
attribute that defaults to true
. This means that node debugging always tries to use source maps (if it can find any) and as a consequence, you can even specify a source file (for example, app.ts) with the program
attribute.
If you need to disable source maps for some reason, you can set the sourceMaps
attribute to false
.
If the generated (transpiled) JavaScript files do not live next to their source but in a separate directory, you must help the VS Code debugger locating them by setting the outFiles
attribute. This attribute takes multiple glob patterns for including and excluding files from the set of generated JavaScript files. Whenever you set a breakpoint in the original source, VS Code tries to find the generated JavaScript code in the files specified by outFiles
.
Since source maps are not automatically created, you must configure the transpiler you are using to create them. For TypeScript this can be done in the following way:
tsc --sourceMap --outDir bin app.ts
This is the corresponding launch configuration for a TypeScript program:
{
"version": "0.2.0",
"configurations": [
{
"name": "Launch TypeScript",
"type": "node",
"request": "launch",
"program": "app.ts",
"outFiles": ["${workspaceFolder}/bin/**/*.js"]
}
]
}
With the smartStep
attribute set to true
in a launch configuration, VS Code will automatically skip ‘uninteresting code’ when stepping through code in the debugger. ‘Uninteresting code’ is code that is generated by a transpiling process but is not covered by a source map so it does not map back to the original source. This code gets in your way when stepping through source code in the debugger because it makes the debugger switch between the original source code and generated code that you are not really interested in. smartStep
will automatically step through code not covered by a source map until it reaches a location that is covered by a source map again.
This is especially useful for cases like async/await downcompilation in TypeScript, where the compiler injects helper code that is not covered by a source map.
Please note that the smartStep
feature only applies to JavaScript code that was generated from source and therefore has a source map. For JavaScript without sources, the smart stepping option has no effect.
A common issue when debugging with source maps is that you’ll set a breakpoint, and it will turn gray. If you hover the cursor over it, you’ll see the message, "Breakpoint ignored because generated code not found (source map problem?)"
. What now? There are a range of issues that can lead to this. First, a quick explanation of how the Node debug adapter handles source maps.
When you set a breakpoint in app.ts
, the debug adapter has to figure out the path to app.js
, the transpiled version of your TypeScript file, which is what is actually running in Node. But, there is not a straightforward way to figure this out starting from the .ts
file. Instead, the debug adapter uses the outFiles
attribute in the launch.json
to find all the transpiled .js
files, and parses them for a source map, which contains the locations of its associated .ts
files.
When you build your app.ts
file in TypeScript with source maps enabled, it either produces an app.js.map
file, or a source map inlined as a base64-encoded string in a comment at the bottom of the app.js
file. To find the .ts
files associated with this map, the debug adapter looks at two properties in the source map, sources
, and sourceRoot
. sourceRoot
is optional - if present, it is prepended to each path in sources
, which is an array of paths. The result is an array of absolute or relative paths to .ts
files. Relative paths are resolved relative to the source map.
Finally, the debug adapter searches for the full path of app.ts
in this resulting list of .ts
files. If there’s a match, it has found the source map file to use when mapping app.ts
to app.js
. If there is no match, then it can’t bind the breakpoint, and it will turn gray.
Here are some things to try when your breakpoints turn gray:
"sourceMaps": false
in your launch.json
?.js.map
files, or inlined source maps in your .js
files?outFiles
property in your launch.json
? It should be a glob pattern for an absolute path that matches your .js
files.sourceRoot
and sources
properties in your source map correct? Can they be combined to get the correct path to the .ts
file?webpack:///
prefix, which the debug adapter can’t resolve. You can change this in your Webpack configuration with the devtoolModuleFilenameTemplate
option, or try using the ‘inspector’ protocol, which provides some extra options for resolving these paths.foo/
from the command line like code FOO
in which case source maps may not be resolved correctly.debugger
statement. If it breaks into the .ts
file there, but breakpoints at that spot don’t bind, that is useful information to include with a GitHub issue.The Node.js debugger supports remote debugging for versions of Node.js >= 4.x. Specify a remote host via the address
attribute. Here is an example:
{
"type": "node",
"request": "attach",
"name": "Attach to remote",
"address": "TCP/IP address of process to be debugged",
"port": "9229"
}
By default, VS Code will stream the debugged source from the remote Node.js folder to the local VS Code and show it in a read-only editor. You can step through this code, but cannot modify it. If you want VS Code to open the editable source from your workspace instead, you can setup a mapping between the remote and local locations. A localRoot
and a remoteRoot
attribute can be used to map paths between a local VS Code project and a (remote) Node.js folder. This works even locally on the same system or across different operating systems. Whenever a code path needs to be converted from the remote Node.js folder to a local VS Code path, the remoteRoot
path is stripped off the path and replaced by localRoot
. For the reverse conversion, the localRoot
path is replaced by the remoteRoot
.
{
"type": "node",
"request": "attach",
"name": "Attach to remote",
"address": "TCP/IP address of process to be debugged",
"port": "9229",
"localRoot": "${workspaceFolder}",
"remoteRoot": "C:\\Users\\username\\project\\server"
}
Two frequently used applications of remote debugging are:
If you are running Node.js inside a Docker container, you can use the approach from above to debug Node.js inside the Docker container and map back the remote source to files in your workspace. We have created a “recipe” on GitHub that walks you through on how to set this up Node.js in Docker with TypeScript.
If you want to run Node.js in the Windows Subsystem for Linux (WSL), you can use the approach from above as well. However, to make this even simpler, we’ve introduced a useWSL
flag to automatically configure everything so that Node.js runs in the Linux subsystem and source is mapped to files in your workspace.
Here is the simplest debug configuration for debugging hello.js
in WSL:
```
{
"type": "node",
"request": "launch",
"name": "Launch in WSL",
"useWSL": true,
"program": "${workspaceFolder}/hello.js"
}
```
Please note: With the arrival of the Remote - WSL extension, VS Code got universal support for Windows Subsystem for Linux (WSL). Consequently, the useWSL
debug configuration attribute has been deprecated and support for it will be dropped soon. For more details, please see our Developing in WSL documentation.
If you need to set a breakpoint in a script that is not part of your workspace and therefore cannot be easily located and opened through normal VS Code file browsing, you can access the loaded scripts via the LOADED SCRIPTS view in the Run view:
Alternatively, you can use the Debug: Open Loaded Script action (Ctrl+F4) which opens a Quick Pick, where you can filter and select the script to open.
Scripts are loaded into a read-only editor where you can set breakpoints. These breakpoints are remembered across debug sessions but you only have access to the script content while a debug session is running.
The restart
attribute of a launch configuration controls whether the Node.js debugger automatically restarts after the debug session has ended. This feature is useful if you use nodemon to restart Node.js on file changes. Setting the launch configuration attribute restart
to true
makes the node debugger automatically try to reattach to Node.js after Node.js has terminated.
If you have started your program server.js
via nodemon on the command line like this:
nodemon --inspect server.js
you can attach the VS Code debugger to it with the following launch configuration:
{
"name": "Attach to node",
"type": "node",
"request": "attach",
"restart": true,
"port": 9229
}
Alternatively you can start your program server.js
via nodemon directly with a launch config and attach the VS Code debugger:
{
"name": "Launch server.js via nodemon",
"type": "node",
"request": "launch",
"runtimeExecutable": "nodemon",
"program": "${workspaceFolder}/server.js",
"restart": true,
"console": "integratedTerminal",
"internalConsoleOptions": "neverOpen"
}
Tip: Pressing the Stop button stops the debug session and disconnects from Node.js, but nodemon (and Node.js) will continue to run. To stop nodemon, you will have to kill it from the command line (which is easily possible if you use the
integratedTerminal
as shown above).
Tip: In case of syntax errors, nodemon will not be able to start Node.js successfully until the error has been fixed. In this case, VS Code will continue trying to attach to Node.js but eventually give up (after 10 seconds). To avoid this, you can increase the timeout by adding a
timeout
attribute with a larger value (in milliseconds).
The Node debugger has a mechanism that tracks all subprocesses of a debuggee and tries to automatically attach to those processes that are launched in debug mode. This feature simplifies debugging of programs that fork or spawn Node.js processes like programs based on the “cluster” node module:
The feature is enabled by setting the launch config attribute autoAttachChildProcesses
to true:
{
"type": "node",
"request": "launch",
"name": "Cluster",
"program": "${workspaceFolder}/cluster.js",
"autoAttachChildProcesses": true
}
Tip: In order to be able to track the subprocesses, we need the process ID of the parent. For this we require that the main debuggee launched from the launch config is a Node.js process and we use an “evaluate” to find its process ID.
Tip: Whether a process is in debug mode is guessed by analyzing the program arguments. Currently we detect the patterns
--inspect
,--inspect-brk
,--inspect-port
,--debug
,--debug-brk
,--debug-port
(all optionally followed by a=
and a port number).
The Node debugger supports restarting execution at a stack frame. This can be useful in situations where you have found a problem in your source code and you want to rerun a small portion of the code with modified input values. Stopping and then restarting the full debug session can be very time-consuming. The Restart Frame action allows you to reenter the current function after you have changed variables with the Set Value action:
Note that Restart Frame won’t unroll any state changes, so it may not always work as expected.
Make sure to use a Node.js version >= 5.11 since earlier versions do not work in all situations.
The Node.js debugger only supports function breakpoints when the “legacy” protocol is used (that is when targeting Node.js < 8.0 versions). In addition, be aware of the following limitations when using function breakpoints:
The ‘hit count condition’ controls how many times a breakpoint needs to be hit before it will ‘break’ execution. The hit count syntax supported by the Node.js debugger is either an integer or one of the operators <
, <=
, ==
, >
, >=
, %
followed by an integer.
Some examples:
>10
break always after 10 hits<3
break on the first two hits only10
same as >=10
%2
break on every other hitFor performance reasons, Node.js parses the functions inside JavaScript files lazily on first access. As a consequence, breakpoints don’t work in source code areas that haven’t been seen (parsed) by Node.js.
Since this behavior is not ideal for debugging, VS Code passes the --nolazy
option to Node.js automatically. This prevents the delayed parsing and ensures that breakpoints can be validated before running the code (so they no longer “jump”).
Since the --nolazy
option might increase the start-up time of the debug target significantly, you can easily opt out by passing a --lazy
as a runtimeArgs
attribute.
When doing so, you will find that some of your breakpoints don’t “stick” to the line requested but instead “jump” for the next possible line in already-parsed code. To avoid confusion, VS Code always shows breakpoints at the location where Node.js thinks the breakpoint is. In the BREAKPOINTS section, these breakpoints are shown with an arrow between requested and actual line number:
This breakpoint validation occurs when a session starts and the breakpoints are registered with Node.js, or when a session is already running and a new breakpoint is set. In this case, the breakpoint may “jump” to a different location. After Node.js has parsed all the code (for example, by running through it), breakpoints can be easily reapplied to the requested locations with the Reapply button in the BREAKPOINTS section header. This should make the breakpoints “jump back” to the requested location.
VS Code Node.js debugging has a feature to avoid source code that you don’t want to step through (AKA ‘Just My Code’). This feature can be enabled with the skipFiles
attribute in your launch configuration. skipFiles
is an array of glob patterns for script paths to skip.
For example, using:
"skipFiles": [
"${workspaceFolder}/node_modules/**/*.js",
"${workspaceFolder}/lib/**/*.js"
]
all code in the node_modules
and lib
folders in your project will be skipped.
Built-in core modules of Node.js can be referred to by the ‘magic name’ <node_internals>
in a glob pattern. The following example skips all internal modules:
"skipFiles": [
"<node_internals>/**/*.js"
]
The exact ‘skipping’ rules are as follows:
Skipped source is shown in a ‘dimmed’ style in the CALL STACK view:
Hovering over the dimmed entries explains why the stack frame is dimmed.
A context menu item on the call stack, Toggle skipping this file enables you to easily skip a file at runtime without adding it to your launch config. This option only persists for the current debugging session. You can also use it to stop skipping a file that is skipped by the skipFiles
option in your launch config.
Note: The
legacy
protocol debugger supports negative glob patterns, but they must follow a positive pattern: positive patterns add to the set of skipped files, while negative patterns subtract from that set.
In the following (legacy
protocol-only) example all but a ‘math’ module is skipped:
"skipFiles": [
"${workspaceFolder}/node_modules/**/*.js",
"!${workspaceFolder}/node_modules/math/**/*.js"
]
Note: The
legacy
protocol debugger has to emulate theskipFiles
feature because the V8 Debugger Protocol does not support it natively. This might result in slow stepping performance.
Since the VS Code Node.js debugger communicates to the Node.js runtimes through wire protocols, the set of supported runtimes is determined by all runtimes supporting the wire protocols.
Today two wire protocols exist:
--inspect
flag in Node.js versions >= 6.3. It addresses most of the limitations and scalability issues of the legacy protocol.Currently these protocols are supported by specific version ranges of the following runtimes:
Although it appears to be possible that the VS Code Node.js debugger picks the best protocol always automatically, we’ve decided for a ‘pessimistic approach’ with an explicit launch configuration attribute protocol
and the following values:
auto
: tries to automatically detect the protocol used by the targeted runtime. For configurations of request type launch
and if no runtimeExecutable
is specified, we try to determine the version by running node from the PATH with an --version
argument. If the node version is >= 8.0, the new ‘inspector’ protocol is used. For configurations of request type ‘attach’, we try to connect with the new protocol and if this works, we use the ‘inspector’ protocol. We only switch to the new ‘inspector’ protocol for versions >= 6.9 because of severe problems in earlier versions.inspector
: forces the node debugger to use the ‘inspector’ protocol-based implementation. This is supported by node versions >= 6.3 and Electron versions >= 1.7.4.legacy
: forces the node debugger to use the ‘legacy’ protocol-based implementation. This is supported by node versions < v8.0 and Electron versions < 1.7.4.Starting with VS Code 1.11, the default value for the protocol
attribute is auto
.
If your runtime supports both protocols, here are a few additional reasons for using the inspector
protocol over legacy
:
inspector
protocol. This issue is tracked in Microsoft/vscode#12749.inspector
protocol can handle some trickier source map setups. If you have trouble setting breakpoints in source mapped files, try using inspector
.We try to keep feature parity between both protocol implementations but this becomes more and more difficult because the technology underlying legacy
is deprecated whereas the new inspector
evolves quickly. For this reason, we specify the supported protocols if a feature is not supported by both legacy
and inspector
.
Debugging Node in Visual Studio Code
Debugging Node.js with Visual Studio Code
#node-js #vscode #web-development