I am doing development that gets pushed/built to multiple systems. To push/build to each system, I am using the same command but have to change an environment variable.
I am using tasks.json to push to multiple systems with one task and I am using task options to set the environment variable.
As you can see it, it is kinda messy.
{
"version": "2.0.0",
// global options
"options": {
"env": {
"SOME_ENV_VAR": "~/globalFile"
}
},
// global options for windows
"windows": {
"options": {
"env": {
"SOME_ENV_VAR": "${env:userprofile}\\globalFile",
}
}
},
"tasks": [
{
"label": "push to system A",
"type": "shell",
"command": "pushCommand"
},
{
"label": "push to system B",
"type": "shell",
"command": "pushCommand",
// special options for pushing to sytem B
"options": {
"env": {
"SOME_ENV_VAR": "~/systemBFile"
}
},
// special windows options for pushing to system B
"windows": {
"options": {
"env": {
"SOME_ENV_VAR": "${env:userprofile}\\systemBFile",
}
}
}
},
// main task to run both
{
"label": "push",
"type": "shell",
"dependsOrder": "parallel",
"dependsOn": [
"push to system A",
"push to system B"
]
},
{
"label": "build to system A",
"type": "shell",
"command": "buildCommand"
},
{
"label": "build to system B",
"type": "shell",
"command": "buildCommand",
// special options for pushing to sytem B
"options": {
"env": {
"SOME_ENV_VAR": "~/systemBFile"
}
},
// special windows options for pushing to system B
"windows": {
"options": {
"env": {
"SOME_ENV_VAR": "${env:userprofile}\\systemBFile",
}
}
}
},
// main task to run both
{
"label": "build",
"type": "shell",
"group": {
"kind": "build",
"isDefault": true
},
"dependsOrder": "parallel",
"dependsOn": [
"build to system A",
"build to system B"
]
}
]
}
So I am wondering if I can create different groups, one for each system: system A, system B, system C, etc... and then assign options and windows to those groups to make it easier to manage. Something like this:
{
"label": "build to system A",
"type": "shell",
"command": "buildCommand",
"group": "system A"
},
{
"label": "build to system B",
"type": "shell",
"command": "buildCommand",
"group": "system B"
}
I can use input variables from launch.json in launch.json.
"configurations": [
{
...
"args": [${input:file_no}]
"preLanuchTask": "runPreTasks"
...
}
],
"inputs": [
{
"id": "file_no",
"type": "promptString"
}
]
Now, I want to get access to the same variable without entering input a second time in tasks.json.
{
"version": "2.0.0",
"tasks":[
{
"label": "runPreTasks",
"type": "shell",
"command": sh,
"args": [
"/path2script/scriptName.sh",
"${input:file_no}" // This does not work, without defining input again
]
}
]
}
Is there a way to pass input variables from launch.json to tasks.json in vscode?
You can use the extension Command Variable v1.21.0
It has a command extension.commandvariable.promptStringRemember that behaves the same as an ${input:name} promptString variable. By adding a key property the result is saved under this key and you can retreive it with the extension.commandvariable.rememberPick command.
The extension.commandvariable.rememberPick command can be used in a different task/launch than the extension.commandvariable.promptStringRemember command
{
"version": "2.0.0",
"tasks": [
{
"label": "Task 1",
"type": "shell",
"command": "dostuff1",
"args": ["-p", "${input:promptPath}"]
},
{
"label": "Task 2",
"type": "shell",
"command": "dostuff2",
"args": ["-p", "${input:rememberPath}"]
},
{
"label": "Do Task 1 and 2",
"dependsOrder": "sequence",
"dependsOn": ["Task 1", "Task 2"],
"problemMatcher": []
}
],
"inputs": [
{
"id": "promptPath",
"type": "command",
"command": "extension.commandvariable.promptStringRemember",
"args": {
"key": "path",
"description": "Enter a path"
}
},
{
"id": "rememberPath",
"type": "command",
"command": "extension.commandvariable.rememberPick",
"args": { "key": "path" }
}
]
}
Following #rioV8 answer, I edited my json files as shown below:
launch.json:
"configurations": [
{
...
"args": [${input:file_no}]
"preLanuchTask": "runPreTasks"
...
}
],
"inputs": [
{
"id": "file_no",
"type": "command",
"command": "extension.commandvariable.promptStringRemember",
"args": {
"key": "lastnumber",
"description": "Enter the number"
}
}
]
tasks.json:
{
"version": "2.0.0",
"tasks":[
{
"label": "runPreTasks",
"type": "shell",
"command": sh,
"args": [
"/path2script/scriptName.sh",
"${input:file_no}"
]
}
]
"inputs": [
{
"id": "file_no",
"type": "command",
"command": "extension.commandvariable.rememberPick",
"args": { "key": "lastnumber" }
}
]
}
I configured tasks.json to build and run the application.
{
"version": "2.0.0",
"tasks": [
{
"label": "build",
"type": "shell",
"command": "make"
},
{
"label": "close the file",
"type": "shell",
"command": "somecommand"
},
{
"label": "Run build",
"dependsOn": [
"build",
"close the file"
],
"dependsOrder": "sequence",
"group": {
"kind": "build",
"isDefault": true
}
}
]
}
The flow is first make command will execute and after that somecommand will be executed. The problem is some times make command returns exit code other than zero, Because of that somecommand is not executing. Is there any way to ignore the previous build status and execute the somecommand always?
To ignore error code in task of type shell, simply add some nop command at the end using || operator. Valid for both bash and batch.
Like:
{
"version": "2.0.0",
"tasks": [
{
"label": "build",
"type": "shell",
"command": "make || echo 0"
},
{
"label": "close the file",
"type": "shell",
"command": "somecommand"
},
{
"label": "Run build",
"dependsOn": [
"build",
"close the file"
],
"dependsOrder": "sequence",
"group": {
"kind": "build",
"isDefault": true
}
}
]
}
In my case a simple Power Shell:
-ErrorAction Ignore
Does the job, but it depends on what you're doing.
Here's the reference for the shell.
~400 times by year 1 :
open VS Code
open one terminal
divide the terminal into 2 parts
run "npm run hot" (first split)
leave the other split terminal empty
open a second terminal (with "+" button)
split it into 3 parts
run "php artisan websockets:serve" (first part)
run "php artisan queue:words" (second part)
run "maidev --ip=localhost" (third part)
and finally I can start to work
I'm sure I am not alone with this problem.
Would anyone have a method to optimize that with one command or configuration ?
Thanks for your help !
Use this setting:
"terminal.integrated.windowsEnableConpty": false // true is the default
In tasks.json:
{
"label": "Run 2 terminals from tasks",
"dependsOrder": "sequence", // or parallel
"dependsOn": [
"OpenTerminal1",
"RunInTerminal1",
"RenameTerminal1", // if you care to rename each terminal
"SplitTerminal1",
"OpenTerminal2",
"RunInTerminal2a",
"RenameTerminal2a", // if you care to rename each terminal
"SplitTerminal2",
"RunInTerminal2b",
"RenameTerminal2b", // if you care to rename each terminal
"SplitTerminal2",
"RunInTerminal2c",
"RenameTerminal2c", // if you care to rename each terminal
],
// "runOptions": { "runOn": "folderOpen" } // or trigger with keybinding
},
That is the "master" task which calls all the constituent tasks, which are:
{
"label": "OpenTerminal1",
"command": "${command:workbench.action.terminal.new}",
"type": "shell",
"problemMatcher": []
},
{
"label": "RunInTerminal1",
"command": "${input:runTerminal1}",
"type": "shell",
"problemMatcher": []
},
{
"label": "RenameTerminal1",
"command": "${input:renameTerminal1}",
"type": "shell",
"problemMatcher": []
},
{
"label": "SplitTerminal1",
"command": "${command:workbench.action.terminal.split}",
"type": "shell",
"problemMatcher": []
},
// ------------------------------------------------------------------
{
"label": "OpenTerminal2",
"command": "${command:workbench.action.terminal.new}",
"type": "shell",
"problemMatcher": []
},
{
"label": "RunInTerminal2a",
"command": "${input:runTerminal2a}",
"type": "shell",
"problemMatcher": []
},
{
"label": "RenameTerminal2a",
"command": "${input:renameTerminal2a}",
"type": "shell",
"problemMatcher": []
},
// ------------------------------------------------------------------
{
"label": "SplitTerminal2",
"command": "${command:workbench.action.terminal.split}",
"type": "shell",
"problemMatcher": []
},
{
"label": "RunInTerminal2b",
"command": "${input:runTerminal2b}",
"type": "shell",
"problemMatcher": []
},
{
"label": "RenameTerminal2b",
"command": "${input:renameTerminal2b}",
"type": "shell",
"problemMatcher": []
},
// ------------------------------------------------------------------
{
"label": "RunInTerminal2c",
"command": "${input:runTerminal2c}",
"type": "shell",
"problemMatcher": []
},
{
"label": "RenameTerminal2c",
"command": "${input:renameTerminal2c}",
"type": "shell",
"problemMatcher": []
},
You might think that is enough but you can see that task require args (what to send to the terminals) so that I had to use this form:
"command": "${input:runTerminal2c}",
so that that input can provide the needed arguments and commands. Those go into the inputs section of your tasks file like so:
"inputs": [
{
"id": "openTerminal1",
"type": "command",
"command": "workbench.action.terminal.new",
},
{
"id": "runTerminal1",
"type": "command",
"command": "workbench.action.terminal.sendSequence",
"args": {
"text": "npm run test\u000D"
}
},
{
"id": "renameTerminal1",
"type": "command",
"command": "workbench.action.terminal.renameWithArg",
"args": {
"name": "npm run hot"
}
},
{
"id": "openTerminal2",
"type": "command",
"command": "workbench.action.terminal.new",
},
{
"id": "runTerminal2a",
"type": "command",
"command": "workbench.action.terminal.sendSequence",
"args": {
"text": "php artisan websockets:serve"
}
},
{
"id": "renameTerminal2a",
"type": "command",
"command": "workbench.action.terminal.renameWithArg",
"args": {
"name": "websockets:serve"
}
},
{
"id": "runTerminal2b",
"type": "command",
"command": "workbench.action.terminal.sendSequence",
"args": {
"text": "php artisan queue:words"
}
},
{
"id": "renameTerminal2b",
"type": "command",
"command": "workbench.action.terminal.renameWithArg",
"args": {
"name": "queue:words"
}
},
{
"id": "runTerminal2c",
"type": "command",
"command": "workbench.action.terminal.sendSequence",
"args": {
"text": "maidev"
}
},
{
"id": "renameTerminal2c",
"type": "command",
"command": "workbench.action.terminal.renameWithArg",
"args": {
"name": "maidev"
}
}
Again maybe you don't care about all the renaming the terminal commands and can eliminate those. Here, I just sent your command to the terminal:
{
"id": "runTerminal2c",
"type": "command",
"command": "workbench.action.terminal.sendSequence",
"args": {
"text": "maidev"
}
},
but did not automatically start them. To have them start immediately, add the return unicode to the text that is sent to the terminal like:
{
"id": "runTerminal2c",
"type": "command",
"command": "workbench.action.terminal.sendSequence",
"args": {
"text": "maidev --ip=localhost"
}
},
And finally a keybinding to trigger the master task (if you don't use the run on folder open option):
{
"key": "alt+z",
"command": "workbench.action.tasks.runTask",
"args": "Run 2 terminals from tasks"
},
I prefer the macro version - it is much shorter and less prone to set-up error. Both are very fast.
Okay, a day after answering this with other two answers, I saw this extension: Restore Terminals at this other question https://stackoverflow.com/a/62595681/836330
This extension seems to work quite well. Put this into your settings.json:
"restoreTerminals.runOnStartup": false, // true is the default
// set to false if using a keybinding or the command palette
"restoreTerminals.terminals": [
{
"splitTerminals": [
{
"name": "server",
"commands": [
"npm run test"
]
},
{
"name": "empty",
}
]
},
{
"splitTerminals": [
{
"name": "websockets",
"commands": [
"ls -l"
]
},
{
"name": "queue",
"commands": [
"cd zip",
"gulp sass"
]
},
{
"name": "maidev",
"commands": [
"cd zip-multiple",
"gulp"
]
}
]
}
],
And a keybinding:
{
"key": "shift+alt+t", // whatever keybinding if you wish
"command": "restore-terminals.restoreTerminals",
},
If on Windows, I would still use the setting:
"terminal.integrated.windowsEnableConpty": false
as the same issue occurs when using this extension as any other method to open and write to terminals quickly - if you then try to close the terminals vscode will hang and eventually timeout and need to be reopened.
I'll make this two answers as one version is particularly long. One answer uses a macro extension to help solve this and the other uses only tasks.
For both solutions, use this setting:
"terminal.integrated.windowsEnableConpty": false // true is the default
There is a nasty unsolved bug affecting Winpty and vscode when you try to delete a terminal opened by these methods.
Using the macro extension multi-command. This goes into your tasks.json:
{
"label": "Run 2 terminals with macro",
"dependsOrder": "sequence", // or parallel
"dependsOn": [
"terminal1",
"terminal2"
],
// "runOptions": { "runOn": "folderOpen" }
},
{
"label": "terminal1",
"command": "${command:multiCommand.startFirstTerminal}"
},
{
"label": "terminal2",
"command": "${command:multiCommand.startSecondTerminal}"
},
The two tasks are run by the macro. This goes into your settings.json:
"multiCommand.commands": [
{
"command": "multiCommand.startFirstTerminal",
"sequence": [
"workbench.action.terminal.newInActiveWorkspace",
{
"command": "workbench.action.terminal.sendSequence",
"args": {
"text": "npm run test\u000D"
}
},
{
"command": "workbench.action.terminal.renameWithArg",
"args": {
"name": "npm run hot"
}
},
"workbench.action.terminal.split",
]
},
{
"command": "multiCommand.startSecondTerminal",
"sequence": [
"workbench.action.terminal.newInActiveWorkspace",
{
"command": "workbench.action.terminal.renameWithArg",
"args": {
"name": "websockets:serve"
}
},
{
"command": "workbench.action.terminal.sendSequence",
"args": {
"text": "ls -lrt\u000D"
}
},
"workbench.action.terminal.split",
{
"command": "workbench.action.terminal.renameWithArg",
"args": {
"name": "queue:words"
}
},
{
"command": "workbench.action.terminal.sendSequence",
"args": {
"text": "ls -lrt\u000D"
}
},
"workbench.action.terminal.split",
{
"command": "workbench.action.terminal.renameWithArg",
"args": {
"name": "maidev"
}
},
{
"command": "workbench.action.terminal.sendSequence",
"args": {
"text": "ls -lrt\u000D"
}
},
]
}
]
I renamed the terminals to match your commands - perhaps you don't care about that and can eliminate the renameWithArg sections to shorten the whole thing.
I couldn't test with your php setup, so I substituted ls -lrt\u000D running in each terminal. Just substitute your
{
"command": "workbench.action.terminal.sendSequence",
"args": {
"text": "php artisan websockets:serve\u000D"
}
},
for example. The \u000D is a return so the command runs immediately. Now you can either use the "runOptions": { "runOn": "folderOpen" } option so the master task runs when the workspace opens or assign a keybinding to the master task like
{
"key": "alt+z", // or watever you want
"command": "workbench.action.tasks.runTask",
"args": "Run 2 terminals with macro"
},