Python Interactive passes subprocess arguments to node instead of the provided executable

[Clocktown]

Environment data

• VS Code version: 1.29.1
• Extension version: 2018.10.1
• OS and version: Windows 10 64 Bit
• Python version: Anaconda 3, Python 3.7, 64 Bit
• Type of virtual environment used (N/A | venv | virtualenv | conda | ...): conda
• Relevant/affected Python packages and their versions: subprocess

Actual behavior

When the following code is run, which is supposed to start orca (Electron application from Plotly), the arguments are instead passed to node e.g. node --help output is printed, but only if the code is run in the Python Interactive Pane (using Run Cell). This also happens with other arguments than --help, but not with other executables, e.g. code, conda and others work as expected.

import subprocess
res = subprocess.run(
 args=[<path_to_orca_executable>, '--help'],
 stdout=subprocess.PIPE,
 stderr=subprocess.PIPE)

print(res.stdout)

The problem also happens when I connect to the Jupyter Notebook instance launched by VS Code in my Browser, but it does not happen with a jupyter notebook instance launched in the Anaconda Terminal, nor does it happend in a REPL, when launched as a file, or in a qtconsole.

Expected behavior

Orca is run with the provided arguments.

Steps to reproduce:

1. Install orca conda install -c plotly plotly-orca
2. Execute above code inside a Python Interactive Pane in Code, using Run Cell

Logs

Output for Python in the Output panel (View→Output, change the drop-down the upper-right of the Output panel to Python)

None.

Output from Console under the Developer Tools panel (toggle Developer Tools on under Help)

console.ts:134 [Extension Host] Python Extension: Error: ENOENT: no such file or directory, open 'C:\Users\<username>\.vscode\extensions\ms-python.python-2018.10.1\out\client\datascience\defaultTheme.json'
t.log @ console.ts:134
t._logExtensionHostMessage @ extensionHost.ts:437
(anonymous) @ extensionHost.ts:244
emitTwo @ events.js:126
emit @ events.js:214
emit @ internal/child_process.js:772
_combinedTickCallback @ internal/process/next_tick.js:141
_tickCallback @ internal/process/next_tick.js:180
console.ts:134 [Extension Host] Python Extension: [I 16:40:00.286 NotebookApp] JupyterLab extension loaded from C:\Users\<username>\Anaconda3\lib\site-packages\jupyterlab

console.ts:134 [Extension Host] Python Extension: [I 16:40:00.286 NotebookApp] JupyterLab application directory is C:\Users\<username>\Anaconda3\share\jupyter\lab

console.ts:134 [Extension Host] Python Extension: [I 16:40:00.290 NotebookApp] Serving notebooks from local directory: C:\Users\<username>\AppData\Local\Temp

console.ts:134 [Extension Host] Python Extension: [I 16:40:00.290 NotebookApp] The Jupyter Notebook is running at:
[I 16:40:00.290 NotebookApp] http://localhost:8888/?token=3d93585235d8531034b0571a9462f8dd19708771c2d1c4bd
[I 16:40:00.290 NotebookApp] Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).

console.ts:134 [Extension Host] Python Extension: [C 16:40:00.299 NotebookApp] 
 
 Copy/paste this URL into your browser when you connect for the first time,
 to login with a token:
 http://localhost:8888/?token=3d93585235d8531034b0571a9462f8dd19708771c2d1c4bd

console.ts:134 [Extension Host] Python Extension: [I 16:40:00.617 NotebookApp] Kernel started: 20c2be11-5288-4dce-806f-dd44b35f1ace

console.ts:134 [Extension Host] Starting WebSocket: ws://localhost:8888/api/kernels/20c2be11-5288-4dce-806f-dd44b35f1ace
console.ts:134 [Extension Host] Python Extension: [I 16:40:01.650 NotebookApp] Adapting to protocol v5.1 for kernel 20c2be11-5288-4dce-806f-dd44b35f1ace

console.ts:134 [Extension Host] Kernel: connected (20c2be11-5288-4dce-806f-dd44b35f1ace)
console.ts:134 [Extension Host] Python Extension: Execute for import pandas as pd
import numpy
%matplotlib inline
import matplotlib.pyplot as plt silently finished.
console.ts:134 [Extension Host] Python Extension: Execute for %cd "c:\Users\nille\Documents\Python Scripts\Pyplot" silently finished.
console.ts:134 [Extension Host] Python Extension: [I 16:40:53.197 NotebookApp] 302 GET /?token=3d93585235d8531034b0571a9462f8dd19708771c2d1c4bd (::1) 0.00ms

console.ts:134 [Extension Host] Python Extension: [I 16:41:10.171 NotebookApp] Creating new notebook in 

console.ts:134 [Extension Host] Python Extension: [I 16:41:11.718 NotebookApp] Kernel started: d3a5e5ef-ed29-4f99-9426-b157731b8fdd

console.ts:134 [Extension Host] Python Extension: [I 16:41:12.537 NotebookApp] Adapting to protocol v5.1 for kernel d3a5e5ef-ed29-4f99-9426-b157731b8fdd

I have also posted an issue at orca's GitHub because I am unsure what is at fault here: plotly/orca#149

[rchiodo]

This sounds like potentially like another conda activation issue.

Does this work correctly if you install jupyter in a non conda environment?

[Clocktown]

I am going to try it later when I get home and update my comment. However, I made sure to run VS Code from a command prompt with activated environment. Without that Jupyter won't even start, but that's a known issue.

Edit:
Tested it with "regular" Python 3.6.6 I had installed just now and the same problem happens.
My guess is that this is somehow Node or Electron related. It could be orca's fault e.g. orca could check whether it was invoked by something that was a subprocess of a Node/Electron application. In that case I might have luck with my issue over there. But it could also be one of these problems that we can't do anything about in an extension because of limitations in VS Code itself. I really don't know.

Was someone able to reproduce my issue?

[jonmmease]

It looks like the issue here is the ELECTRON_RUN_AS_NODE environment variable that is set by VSCode (See plotly/orca#149 for more info). Do you know why this variable is set, and whether it would cause VSCode any problem if plotly.py clears it from os.environ?

[rchiodo]

I believe this is by design. We don't have a list of environment variables we'd set/unset when running python interactive. That list would be impossible to know. The environment variables is coming from VS Code.

Seems like plotly should handle this. They're the ones with the incorrect behavior.

[DonJayamanne]

Seems like plotly should handle this. They're the ones with the incorrect behavior.

Rich Chiodo (@rchiodo), shouldn't we at a minimum clear this environment variable now that we know the cause, when spinning up the jupter process, this way we get things working for users using plotly,

Else people using this extension have no choice but to do this out side of VSCode!! I.e. move out of the extension!

[rchiodo]

And what if something else needs this environment variable to know they were launched from electron?

[rchiodo]

I don't like guessing what other packages might or might not do.

[Clocktown]

I agree with Rich Chiodo (@rchiodo) - side effects should be kept minimal. This issue is now documented here, and a workaround does exist by clearing it yourself. Python packages that rely on launching external processes might want to handle this themselves if the application they want to launch does not work with that environment variable.
This extension is not at fault here. I assume VScode itself has a reason for setting this environment variable.
Don Jayamanne (@DonJayamanne) you could open an issue at VScode to discuss this further.