NodeJS，Windows，Wmic标准输入

Question

I've having a problem with the Windows program WMIC.exe (used for querying Windows Management Instrumentation), from within NodeJS.

I can spawn a wmic.exe process, but it'll not output anything or accept any input. However, if I set the stdin to null it'll run correctly and I'll receive output.

var spawn = require('child_process').spawn;
var wmic = spawn('wmic', [], {stdio: ['ignore']});

wmic.stdout.on('data', function(data) {
  console.log('Data received:' + data);
});

wmic.on('close', function(code) {
  console.log('Spawned process ended with code: ' + code);
});

I want to use wmic.exe interactively, which it supports, keeping it open so I don't have to repeatedly start it for each WMI query. Others have had similar problems with wmic.exe, but they only wanted to capture output for a single query, so having a null stdin doesn't matter to them.

Update

If I do the following...

var spawn = require('child_process').spawn;
var wmic = spawn('wmic', []);

wmic.stdout.on('data', function(data) {
  console.log('Data received:' + data);
});

wmic.stderr.on('data', function(data) {
  console.log('Error! - ' + data);
});

wmic.on('close', function(code) {
  console.log('Spawned process ended with code: ' + code);
});

wmic.stdin.end('cpu get caption /format:csv\n');

Then it'll actually respond with a result, prompt back ready for the next query but next ends the process and obviously I can't use .write() after .end(). Changing it to a .write() instead, process doesn't respond at all and I don't even get the prompt from stdout "wmic:root\\cli>" that I receive when using .end().

Or

wmic.stdin.push(null);
wmic.stdin.write('cpu/n'); // Ctrl-Z aka Windows EOF control code

If I use the above, instead of the .end() call, it'll also work. But it seems that if I don't keep throwing constant data at it, the process ends.

Or even

wmic.stdin.write('cpu');
wmic.stdin.write('\x1a');

This works too, but yet again, the wmic.exe process decides to exit after the cpu results are returned. :/

Almost

var wmic = spawn('wmic.exe', [], {stdio: [process.stdin, 'pipe', 'pipe']});

This works. The process starts up correctly so I receive the prompt from wmic.exe through wmic.stdout.on('data', fn) and it stays open. However, I'm unable to send input to it via code but can type directly into the command prompt window. Anything I type is correctly executed and output via the node script. Again, weird that it works with this pipe but not any I set and useless to me as I want to send commands via code, not typing them.

Answer 1

You want to use wmic interactively, but driven by input from an application rather than input typed at the console.

The normal way to achieve this for any application is to feed it input from a pipe. Conveniently, NodeJS appears to create a pipe by default if you don't specify any other stdio option.

The problem is that wmic behaves badly when it is receiving input from a pipe. It refuses to do anything until the input is complete; it then processes all the input and stops looking for any more. wmic thinks that the input is complete when the pipe is closed or it receives a CTRL-Z.

Hence you cannot issue one command, read the output and then issue another command. wmic doesn't write output until it thinks it has received all the commands, and then it won't accept any more.

You can illustrate this problem at the console. type con: | wmic type con: | wmic puts keyboard input through a pipe so wmic goes into its broken mode. In the example below this is followed by a couple of commands. Note that wmic does nothing until CTRL-Z is entered.

C:\>type con: | wmic
cpu get name
useraccount list brief
^Z
wmic:root\cli>cpu get name
Name
Pentium(R) Dual-Core CPU       T4500  @ 2.30GHz

wmic:root\cli>
"/?" for help, QUIT to Exit.
wmic:root\cli>useraccount list brief
AccountType  Caption                    Domain     FullName         Name
512          frog\Administrator         frog                        Administrator

wmic:root\cli>
"/?" for help, QUIT to Exit.
wmic:root\cli>
C:\>

You might think (looking at this example) that wmic just has a large output buffer. However, if you turn on tracing ( /TRACE:ON as the first command) wmic produces reams of output, but nothing before the input is complete.

A solution is to not redirect the standard input, like your final example, so the input is coming from the keyboard. (NB when a Windows console application receives input from the keyboard it is not coming through a pipe. Hence your final example is not using a pipe, which is why is behaves differently.)

You can then supply input to wmic by simulating keystrokes. Answers to this question suggest a couple of methods: either using SendKeys or by sending WM_KEYUP messages. This is not a great solution.

Answer 2

For posterity...

This is now possible by using the ms-wmic package:

• https://www.npmjs.com/package/ms-wmic

This completely bypasses having to directly interact with the CLI, and allows you to deal with WMIC through a preset API.

---

But the specific thing you tried to use WMIC for in your example above, is also handled by the windows-cpu package:

• https://www.npmjs.com/package/windows-cpu

Answer 3

That's how it worked for me:

var spawn = require('child_process').spawn;
var wmic = spawn('wmic', []);

wmic.stdout.on('data', function(data) {
  console.log('Data received: [[' + data + ']]');
});

wmic.stderr.on('data', function(data) {
  console.log('Error! - ' + data);
});

wmic.on('close', function(code) {
  console.log('Spawned process ended with code: ' + code);
});

wmic.stdin.end( 'cpu get caption /format:csv' );