Capture, separate and serialise all 6 PowerShell message streams at their source.
PowerShell has 6 message streams in addition to Stdout and Stderr. Piping these 6 streams all through stdout simultaneously removes each stream's semantics and is a nightmare to parse - unexpected/intermittent warning or error messages (very common when remoting) can easily break your application.
Full-PowerShell captures and serialises these streams before returning them, so output from one stream will not affect the other.
This library wraps your commands in an Invoke-Command
block that pipes their output into 6 individual streams. All 6 streams are captured while your command/script executes, once it has completed the streams are serialised as JSON, returned to parent Node.js process, and finally deserialised as individual streams.
Invocations can be subscribed to as an RxJS Observable, or as a Promise.
The source code is fairly concise, take a look at index.ts and wrapper.ts to see exactly how it works.
Spawns a PowerShell child process on instantiation and exposes methods to read/write to/from that process:
class PowerShell {
constructor(private options?: PowerShellOptions);
success$: Subject<any[]>();
error$: Subject<any[]>();
warning$: Subject<any[]>();
verbose$: Subject<any[]>();
debug$: Subject<any[]>();
info$: Subject<any[]>();
call(command: string, format: Format = 'json'): SubjectWithPromise<PowerShellStreams>;
destroy(): SubjectWithPromise<boolean>;
}
Note: The SubjectWithPromise
object returned by call()
will only emit the value returned by the command passed to call()
. Use the streams postfixed with $
to listen to output from all calls made to that PowerShell
instance.
Emittied by the <Observable|Promise>
returned from .call().subscribe()
or .call().promise()
.
interface PowerShellStreams {
success: any[];
error: any[];
warning: any[];
verbose: any[];
debug: any[];
info: any[];
}
The subjects exposed by the PowerShell
class, as well as the singleton observable/promise returned by PowerShell.call
all return arrays of strings or parsed JSON. It's important to note that these arrays reflect the output for each PowerShell command passed to PowerShell.call
. For example, if you were to call PowerShell.call('Get-Date; Get-Date;')
, you should expect to receive an Array containing two items in the next emission's success stream. However, there are exceptions to this - debug and verbose are newline delimited due to limitations of PowerShell redirection. While they will generally equate to one string per Write-Debug
or Write-Verbose
, it is up to you to ensure output has not been broken into multiple lines.
ES6
import { PowerShell } from 'full-powershell';
CommonJS
const { PowerShell } = require('full-powershell');
const shell = new PowerShell();
Options:
interface PowerShellOptions {
tmp_dir?: string
exe_path?: string
timeout?: number
verbose?: boolean
debug?: boolean
}
-
tmp_dir
- Default: current directory Change the path for ephemeral '.tmp' files. Must have a trailing slash. (Must be set to/tmp/
when executing on AWS Lambda). -
exe_path
- Default:powershell
for windows,pwsh
for nix. Explicitly set the path or command name for the PowerShell executable (example:'pwsh'
or'C:\\Program Files\\PowerShell\\7\\pwsh.exe'
) -
timeout
- Default: 10 minutes. Set number of milliseconds before each call to this shell will timeout. Warning: A timeout will result in the PowerShell child process being terminated and a new process created, any pending calls will be errored and PowerShell context will be lost. -
verbose
- Default: true Optionally disable verbose output (this may improve performance as capturing verbose requires writing to disk) -
debug
- Default: true Optionally disable debug output (this may improve performance as capturing debug requires writing to disk)
Example:
const options: PowerShellOptions = {
tmp_dir: '/tmp/'
exe_path: 'pwsh',
timeout: 60000
}
const shell = new PowerShell(options);
The call
method accepts a PowerShell command as a string, and an optional Format
parameter. Use the format parameter to change how command output is serialised before returing it from PowerShell.
type Format = 'string' | 'json' | null;
shell.call(command: string, format: Format = 'json')
The call
method returns an SubjectWithPromise
object that provides two methods for handling the response:
subscribe
- an RxJS Subjectpromise
- a standard JavaScript promise
Subscribe directly to a call by calling subscribe()
on the returned object (observable will complete after first emission):
shell.call('My-Command')
.subscribe(
(res: PowerShellStreams) => {
/* result handler */
},
(err: Error) => {
/* error handler */
}
);
The object returned by the call
method also exposes a function called promise()
which returns a promise instead of a subscription.
const { success, error, warning } = await shell.call('My-Command').promise();
In addition to subscribing/awaiting to individual calls, you can subscribe to the message streams individually. These Subscriptions will emit the results of all calls made to the instance. This may be useful for logging.
const shell = new PowerShell();
shell.success$
.subscribe(
(res: Array<any>) => {
/* result handler */
},
(err: Error) => {
/* error handler */
}
);
shell.error$.subscribe( /* same as success$ */);
shell.warning$.subscribe( /* same as success$ */);
shell.verbose$.subscribe( /* same as success$ */);
shell.debug$.subscribe( /* same as success$ */);
shell.info$.subscribe( /* same as success$ */);
It's important to signal when you are finished with the instance, otherwise your program might not exit gracefully. You should also wait for this task to complete.
shell.destroy().subscribe((destroyed: boolean) => { /*...*/ });
or
const destroyed: boolean = await shell.destroy().promise();