version: 5.1.1
session: 67f00511acbd9ccac373edf7
steps:
  - prompt: launch chrome
    commands:
      - command: exec
        lang: shell
        linux: |
          jumpapp google-chrome --disable-fre --no-default-browser-check --no-first-run "${TD_WEBSITE}" &
          exit
        mac: |
          open -na "Google Chrome" --args --disable-fre --no-default-browser-check --no-first-run --disable-features=PasswordManagerEnabled "${TD_WEBSITE}" &
          exit
        windows: |
          Start-Process "C:/Program Files/Google/Chrome/Application/chrome.exe" -ArgumentList "--start-maximized", "--disable-fre", "--no-default-browser-check", "--no-first-run", "--guest", "${TD_WEBSITE}"
      - command: wait-for-text
        text: ${TD_WEBSITE}
        timeout: 30000
      - command: assert
        expect: Google Chrome is running

​
Description

The exec command allows you to execute custom Node.js scripts within your TestDriver tests. This is useful for tasks like generating dynamic data, interacting with APIs, or performing custom logic during a test. The output of the script can be stored in a variable for use in subsequent steps. It’s important to note that the output from exec must be a string.

​
Arguments

ArgumentTypeDescription
langstringThe language of the script to execute. Supported values are shell and js.
outputstringThe variable name to store the result of the script. This variable can be accessed as ${OUTPUT.<var>} in future steps.
silentstringDefaults to true. This is useful for suppressing unnecessary or private output in the test logs. If set to false, the command will print the output of the script. This is useful for debugging.
linuxstringThe script to execute on Linux systems. For js, the script must define the output as result.
windowsstringThe script to execute on Windows systems. For js, the script must define the output as result.
macstringThe script to execute on macOS systems. For js, the script must define the output as result.

​
Example usage

This example demonstrates how to use the exec command to generate a TOTP (Time-based One-Time Password) using the totp-generator library.

version: 5.3.8
steps:
  - commands:
    - command: exec
      lang: shell
      linux: |
        npm install totp-generator
    - command: exec
      lang: js
      output: totp
      linux: |
        const { TOTP } = require("totp-generator");
        let otp = TOTP.generate("JBSWY3DPEB3W64TMMQQQ").otp;
        console.log(otp);
        result = otp;
    - command: type
      text: ${OUTPUT.totp}

​
Additional details

  • The exec command now takes a lang argument and supports different operating systems (linux, mac, and windows).
  • Supported lang values are js or shell:
    • js code is executed in a Node.js VM on the host machine (for example the machine where your CI/CD runs, or your computer if using the local agent).
    • shell code is executed in the shell on target runner (which can be the cloud runner, local sandbox, or local machine, depending on where you run your tests).
      • Note: You can also use shell in prerun scripts to install npm packages if you need them.
      • Otherwise, the shell code can be used within test steps to launch applications or perform simple commands (like writing text to a file on the machine to perform a simple file upload).
  • Code specified in linux, mac, and windows is executed based on the platform of the runner machine.
  • The outputargument is assigned automatically by setting result = somestringvalue in the script you run.

​
Protips

  • The result variable is already available in your script, overwrite it to store the output as shown in the examples.
  • Node.js code executes in the same context as the calling process on the host machine, using the VM module internally.
  • Do any handling of arrays or nested objects within your js script:
    • ✅ result = users[1].profile.firstName
    • ✅ result = data.length > 0 ? data[0].userEmail : 'no user found' if no data is found the value of output will be null
    • ✅ result = someTestUserEmail
    • ✅ result = someTextToAssert
    • ✅ result = someDescriptionOfAnImageToScrollTo
  • Don’t try to pass any non-string values to output:
    • ❌ result = [...users, ...values]
    • ❌ result = {name: "Dale Earnhardt", starts: 676, wins: 76}
    • ❌ result = [{user1: ...}, {user2: ...}]

​
Ways to use exec

Here is an example using both shell and js contexts within a prerun.yaml script:

./lifecycle/prerun.yaml
version: 5.3.8
steps:
  - commands:
      - command: exec
        lang: shell
        silent: false
        mac: |
          npm install @sendgrid/mail
      - command: exec
        lang: js
        output: accountData
        silent: false
        mac: |
          const Mailjs = require("@cemalgnlts/mailjs");
          const mailjs = new Mailjs();
          let account = await mailjs.createOneAccount()
          console.log("Account created:", account);
          result = JSON.stringify(account.data)
      - command: exec
        lang: js
        output: emailAddress
        silent: false
        mac: |
          const accountData = ${OUTPUT.accountData};
          result = accountData.username
  - prompt: Enter the generated email into the email field
    commands:
      - command: hover-text
        text: standard_user
        description: email input field label
        action: click
      - command: type
        text: ${OUTPUT.emailAddress}
  - prompt: Wait for an email, extract links, and open each link
    commands:
      - command: exec
        lang: js
        silent: false
        mac: |
          const Mailjs = require("@cemalgnlts/mailjs");
          const { JSDOM } = require('jsdom'); // To parse HTML and extract links

          const accountData = ${OUTPUT.accountData};

          const getLatestEmailAndClickLinks = async () => {
            try {
              // Initialize the Mailjs client
              const mailjs = new Mailjs();

              // Login to your account
              await mailjs.login(accountData.username, accountData.password);

              // Fetch list of messages
              const messages = await mailjs.getMessages();

              if (messages.length === 0) {
                console.log('No emails found.');
                return;
              }

              // Assuming the latest email is the first one in the list
              const latestMessage = messages[0];

              // Fetch the full details of the latest email
              const fullMessage = await mailjs.getMessage(latestMessage.id);

              console.log('Latest Email Details:', fullMessage);

              // Parse the HTML content to extract links
              const dom = new JSDOM(fullMessage.html);
              const links = Array.from(dom.window.document.querySelectorAll('a')).map(a => a.href);

              console.log('Found Links:', links);

              // Click (fetch) every link using native fetch
              for (const link of links) {
                try {
                  const response = await fetch(link);
                  console.log('Clicked ${link}: ${response.status}');
                } catch (linkError) {
                  console.error('Error clicking ${link}:', linkError);
                }
              }

            } catch (error) {
              console.error('Error fetching latest email or clicking links:', error);
            }
          };

          getLatestEmailAndClickLinks();

​
Using exec shell commands in a test file

In a test file, you can use the shell context directly:

testfile.yaml
version: 5.5.5
steps:
  - prompt: launch a calculator
    commands:
      - command: exec
        lang: shell
        linux: |
          linux: |
            jumpapp /usr/bin/galculator > /dev/null 2>&1 &
            exit
        mac: |
          open "/Applications/Calculator.app"
          wait
        windows: start /B calc.exe
          timeout /t 5
      - command: wait-for-text
        text: "calculator"
        timeout: 30000
  - prompt: /try performing the operation 2 + 2 = on the calculator that is opened
    commands:
      - command: focus-application
        name: galculator
      - command: hover-image
        description: button with number 2 on the calculator
        action: click
      - command: hover-image
        description: plus button on the calculator
        action: click
      - command: hover-image
        description: button with number 2 on the calculator
        action: click
      - command: hover-image
        description: equals button on the calculator
        action: click

​
One more option

You can also save reusable snippets (like launching the calculator) to be inserted into a script later with the run command. That version would look something like this:

launchcalculator.yaml
version: 5.5.5
steps:
  - prompt: launch a calculator
    commands:
      - command: exec
        lang: shell
        linux: |
          linux: |
            jumpapp /usr/bin/galculator > /dev/null 2>&1 &
            exit
        mac: |
          open "/Applications/Calculator.app"
          wait
        windows: start /B calc.exe
          timeout /t 5
      - command: wait-for-text
        text: "calculator"
        timeout: 30000

Then in the test:

testfile.yaml
version: 5.5.5
steps:
  - prompt: launch a calculator
    commands:
      - command: run
        file: ./launchcalculator.yaml
  - prompt: /try performing the operation 2 + 2 = on the calculator that is opened
    commands:
      - command: focus-application
        name: galculator
      - command: hover-image
        description: button with number 2 on the calculator
        action: click
      - command: hover-image
        description: plus button on the calculator
        action: click
      - command: hover-image
        description: button with number 2 on the calculator
        action: click
      - command: hover-image
        description: equals button on the calculator
        action: click

​
Don’t try to run js within a test field

This example will fail at runtime, so don’t try to execute js context directly in a test file. Remember - use this in prerun to setup your test!

badtestfile.yaml
version: 5.5.5
commands:
      - command: exec
        lang: shell
        linux: |
          npm install -g axios json2csv
        mac: |
          npm install -g axios json2csv
        windows: |
          npm install -g axios json2csv
  - prompt: fetch user data from API
    commands:
      - command: exec
        output: user
        lang: js
        mac: |
          const axios = require('axios');
          const { Parser } = require('json2csv');
          const fs = require('fs');

          const response = await axios.get('https://jsonplaceholder.typicode.com/users');
          const parser = new Parser();
          const csv = parser.parse(response.data);
          fs.writeFileSync('users.csv', csv);
          const user = response.data[0].name;
          result = user;
          console.log('username', user);
    ...

This example will produce errors in the TestDriver output or CLI since the runner won’t have access to the Node.js VM context.