Tasks: Shell Input

This extension aims to extend the possibilities of input in task execution. Currently, VSCode supports 3 types of inputs for your tasks:

promptString
pickString
Command

None of them allows to get an input from a system command for example. This extension executes a shell command in your OS and each line of the output will be used as a possible input for your task.

Usage example:

{
  "version": "2.0.0",
  "tasks": [
    {
      "label": "Echo Project File",
      "type": "shell",
      "command": "echo ${input:inputTest}",
      "problemMatcher": []
    }
  ],
  "inputs": [
    {
      "id": "inputTest",
      "type": "command",
      "command": "shellCommand.execute",
      "args": {
          "command": "cat ${file}",
          "cwd": "${workspaceFolder}",
          "env": {
              "WORKSPACE": "${workspaceFolder[0]}",
              "FILE": "${file}",
              "PROJECT": "${workspaceFolderBasename}"
          }
      }
    }
  ]
}

By default the extension returns the exact string value that was produced by the shell command and then shown and selected in 'Quick Pick' dialog. However, sometimes it is useful to show more descriptive information than the internal string value that is returned. This can be done by specifying a fieldSeparator and making the shell command return lines containing multiple fields separated by that value. Supported fields are:

<value>|<label>|<description>|<detail>

Here, <value> is what is returned as input variable and is not shown in the UI. Instead, <label> and <description> are shown on a single line and <detail> is rendered on a separate line. These fields can also include icons such as $(git-merge). All fields except <value> are optional and can be omitted. fieldSeparator can be any string value.

Next example shows a process picker very similar to the built-in ${command:pickProcess}:

{
  "version": "2.0.0",
  "tasks": [
    {
      "label": "Echo Process ID",
      "type": "shell",
      "command": "echo ${input:processID}",
      "problemMatcher": []
    }
  ],
  "inputs": [
    {
      "id": "processID",
      "type": "command",
      "command": "shellCommand.execute",
      "args": {
          "command": "ps axww --no-headers k comm -o '%p|%c|%p|%a' | sed -e 's/^\\s*//' -e 's/\\s*|\\s*/|/g'",
          "fieldSeparator": "|",
          "description": "Select the process to attach to"
      }
    }
  ]
}

VSCode renders it like this:

Arguments for the extension:

command: the system command to be executed (must be in PATH)
cwd: the directory from within it will be executed
env: key-value pairs to use as environment variables (it won't append the variables to the current existing ones. It will replace instead)
useFirstResult: skip 'Quick Pick' dialog and use first result returned from the command
useSingleResult: skip 'Quick Pick' dialog and use the single result if only one returned from the command
fieldSeparator: the string that separates value, label, description and detail fields
description: shown as a placeholder in 'Quick Pick', provides context for the input
maxBuffer: largest amount of data in bytes allowed on stdout. Default is 1024 * 1024. If exceeded ENOBUFS error will be displayed

As of today, the extension supports variable substitution for:

a subset of predefined variables like file, fileDirName, workspaceFolder and workspaceFolderBasename, pattern: ${variable}
all config variables, pattern: ${config:variable}
all environment variables, pattern: ${env:variable}
input variables which have been defined with shellCommand.execute, pattern: ${input:variable} (limited supported see below for usage)
Support for ${command:...} pattern, for example to extract CMake's build directory using ${command:cmake.buildDirectory}.

For a complete vscode variables documentation please refer to vscode variables.

Dependent Input Variables Usage example:

{
    "tasks": {
        "version": "2.0.0",
        "tasks": [
            {
                "label": "Nested input",
                "command": "ls ${input:rootDir}/${input:childDir}",
                "type": "shell",
                "problemMatcher": []
            }
        ],
        "inputs": [
            {
                "id": "rootDir",
                "type": "command",
                "command": "shellCommand.execute",
                "args": {
                    "command": "ls -1a"
                }
            },
            {
                "id": "childDir",
                "type": "command",
                "command": "shellCommand.execute",
                "args": {
                    "command": "ls -1a ${input:rootDir}"
                }
            }
        ]
    }
}

There are a few limitations to be aware of:

in the main task command the input variables should appear (left to right) in order of dependence
- i.e ${input:childDir} must be to the right of it's dependent variable ${input:rootDir}
- this can be worked around by having a dummy dependsOn task (or preLaunchTask for launch configs) with a dummy echo task which has the proper variable order
within an input command arg you can only reference other inputs defined with shellCommand.execute
ensure you don't have another input with the same exact inputs.args.command in your tasks or launch configs as this may confuse the extension

Penguinang/vscode-shell-command

Tasks: Shell Input