Build your current project, directly from Atom







This package consumes the following services:

Atom Build package

Release notes

Plugin installs Package version Shield AppVeyor Shield

Gitter chat Slack Badge

Automatically build your project inside your new favorite editor, Atom.

Builds your project - configure it your way

work work

Automatically extract targets - here with build-make.


Match errors and go directly to offending code - with Atom Linter.

error matching

(You can also use keyboard shortcuts to go to errors if you don't like Atom Linter, or want to keep package dependencies to a minimum).

Quick start

Install the build package using apm (apm can be installed using the install shell commands tool in Atom)(Linux/Mac):

$ apm install build

Create a file called .atom-build.yml (note the inital dot):

cmd: echo Hello world

Save it, and press Cmd Alt B (OS X) / Ctrl Alt B (Linux/Windows) and you should see the output of echo Hello world, which should be Hello world if all is correct.

Build providers

Instead of specifying commands manually, you can use a build provider. They often include functionality such as parsing targets (for instance all tasks from gulpfile.js or Makefile).

Full list of build providers

Specify a custom build command

If no build provider is enough to suit your needs, you can configure the custom build command extensively.

Supported formats and the name of the configuration file is

Pick your favorite format, save that file in your project root, and specify exactly how your project is built (example in yml)

cmd: "<command to execute>"
name: "<name of target>"
  - <argument1>
  - <argument2>
sh: true,
cwd: <current working directory for `cmd`>
  - ^regexp1$
  - ^regexp2$
  - ^regexp1$
  - ^regexp2$
keymap: <keymap string>
atomCommandName: namespace:command
      cmd: "<command to execute>"
      # (any previous options are viable here except `targets` itself)

Note that if sh is false cmd must only be the executable - no arguments here. If the executable is not in your path, either fully qualify it or specify the path in you environment (e.g. by setting the PATH var appropriately on UNIX-like systems).

If sh is true, it will use a shell (e.g. /bin/sh -c) on unix/linux, and command (cmd /S /C) on windows.

Programmatic Build commands (Javascript)

Using a JavaScript (JS) file gives you the additional benefit of being able to specify preBuild and postBuild and being able to run arbitrary match functions instead of regex-matching. The javascript function needs to return an array of matches. The fields of the matches must be the same as those that the regex can set.

Keep in mind that the JavaScript file must export the configuration

module.exports = {
  cmd: "myCommand",
  preBuild: function () {
    console.log('This is run **before** the build command');
  postBuild: function () {
    console.log('This is run **after** the build command');
  functionMatch: function (terminal_output) {
    // this is the array of matches that we create
    var matches = [];
    terminal_output.split(/\n/).forEach(function (line, line_number, terminal_output) {
      // all lines starting with a slash
      if line[0] == '/' {
          file: 'x.txt',
          line: line_number.toString(),
          message: line
    return matches;

A major advantage of the functionMatch method is that you can keep state while parsing the output. For example, if you have a Makefile output like this:

make[1]: Entering directory 'foo'
make[2]: Entering directory 'foo/src'
testmake.c: In function 'main':
testmake.c:3:5: error: unknown type name 'error'

then you can't use a regex to match the filename, because the regex doesn't have the information about the directory changes. The following functionMatch can handle this case. Explanations are in the comments:

module.exports = {
  cmd: 'make',
  name: 'Makefile',
  sh: true,
  functionMatch: function (output) {
    const enterDir = /^make\[\d+\]: Entering directory '([^']+)'$/;
    const error = /^([^:]+):(\d+):(\d+): error: (.+)$/;
    // this is the list of error matches that atom-build will process
    const array = [];
    // stores the current directory
    var dir = null;
    // iterate over the output by lines
    output.split(/\r?\n/).forEach(line => {
      // update the current directory on lines with `Entering directory`
      const dir_match = enterDir.exec(line);
      if (dir_match) {
        dir = dir_match[1];
      } else {
        // process possible error messages
        const error_match = error.exec(line);
        if (error_match) {
          // map the regex match to the error object that atom-build expects
            file: dir ? dir + '/' + error_match[1] : error_match[1],
            line: error_match[2],
            col: error_match[3],
            message: error_match[4]
    return array;

Another feature of functionMatch is that you can attach informational messages to the error messages:

pic of traces and custom error types

You can add these additional messages by setting the trace field of the error object. It needs to be an array of objects with the same fields as the error. Instead of adding squiggly lines at the location given by the file, line and col fields, a link is added to the popup message, so you can conveniently jump to the location given in the trace.

One more feature provided by functionMatch is the ability to use HTML in the message text by setting html_message instead of message. If both html_message and message are set, the latter takes priority.

Configuration options

Option Required Description
cmd [required] The executable command
name [optional] The name of the target. Viewed in the targets list (toggled by build:select-active-target).
args [optional] An array of arguments for the command
sh [optional] If true, the combined command and arguments will be passed to /bin/sh. Default true.
cwd [optional] The working directory for the command. E.g. what . resolves to.
env [optional] An object of environment variables and their values to set
errorMatch [optional] A (list of) regular expressions to match output to a file, row and col. See Error matching for details.
warningMatch [optional] Like errorMatch, but is reported as just a warning
functionMatch [optional] A (list of) javascript functions that return a list of match objects
keymap [optional] A keymap string as defined by Atom. Pressing this key combination will trigger the target. Examples: ctrl-alt-k or cmd-U.
killSignals [optional] An array of signals. The signals will be sent, one after each time Escape is pressed until the process has been terminated. The default value is SIGINT -> SIGTERM -> SIGKILL. The only signal which is guaranteed to terminate the process is SIGKILL so it is recommended to include that in the list.
atomCommandName [optional] Command name to register which should be on the form of namespace:command. Read more about Atom CommandRegistry. The command will be available in the command palette and can be trigger from there. If this is returned by a build provider, the command can programatically be triggered by dispatching.
targets [optional] Additional targets which can be used to build variations of your project.
preBuild [optional] JS only. A function which will be called before executing cmd. No arguments. this will be the build configuration.
postBuild [optional] JS only. A function which will be called after executing cmd. It will be passed 3 arguments: bool buildOutcome indicating outcome of the running cmd, string stdout containing the contents of stdout, and string stderr containing the contents of stderr. this will be the build configuration.


The following parameters will be replaced in cmd, any entry in args, cwd and values of env. They should all be enclosed in curly brackets {}

Creating a build provider

Creating a build provider require very little code in the easiest case, and can be as complicated as necessary to achieve the correct functionality. Read more about building your own provider in the create provider documentation.

Error matching

Error matching lets you specify a single regular expression or a list of regular expressions, which capture the output of your build command and open the correct file, row and column of the error. For instance:

../foo/bar/a.c:4:26: error: expected ';' after expression
  printf("hello world\n")
1 error generated.

Would be matched with the regular expression: (?<file>[\\/0-9a-zA-Z\\._]+):(?<line>\\d+):(?<col>\\d+):\\s+(?<message>.+). After the build has failed, pressing Cmd Alt G (OS X) or Ctrl Alt G (Linux/Windows) (or F4 on either platform), a.c would be opened and the cursor would be placed at row 4, column 26.

Note the syntax for match groups. This is from the XRegExp package and has the syntax for named groups: (?<name> RE ) where name would be the name of the group matched by the regular expression RE.

The following named groups can be matched from the output:

The file should be relative the cwd specified. If no cwd has been specified, then the file should be relative the project root (e.g. the top most directory shown in the Atom Editor).

If your build outputs multiple errors, all will be matched. Press Cmd Alt G (OS X) or Ctrl Alt G (Linux/Windows) to cycle through the errors (in the order they appear, first on stderr then on stdout), or you can use the Atom Linter integration discussed in the next section.

Often, the first error is the most interesting since other errors tend to be secondary faults caused by that first one. To jump to the first error you can use Cmd Alt H (OS X) or Shift F4 (Linux/Windows) at any point to go to the first error.

Error matching and Atom Linter

Install Atom Linter and all your matched errors will listed in a neat panel.

Linter integration


The atom-build package uses google analytics to keep track of which features are in use and at what frequency. This gives the maintainers a sense of what parts of the package is most important and what parts can be removed.

The data is fully anonymous and can not be tracked back to you in any way. This is what is collected

If you really do not want to share this information, you can opt out by disabling the metrics package. This will disable all analytics collection, including the one from atom-build.