scipy missing from linux installer

[clyne]

The scipy module is missing from the linux installer.

To reproduce:

1. Run vapor installed with the linux installer
2. Load any data set (e.g. Duku) and launch the python calculation engine
3. Create a simply python script that contains the line "import scipy" and hit test. The script will fail because scipy cannot be found.

Describe the bug

(Please click check boxes AFTER submitting ticket)

• [ ] Did VAPOR crash?
• [x] Did you get wrong results?

Impact

• [ ] High - User productivity significantly degraded
• [x] Medium - User productivity partially degraded
• [ ] Low - User productivity trivially degraded

Desktop

Linux

Additional context

[sgpearse]

This does not occur on the 3rdPartyLibs branch, which is pending review.

[clyne]

Let’s close it then and we’ll be sure to do our validation testing with the new libraries.

cheers - jc

On Nov 21, 2019, at 12:32 PM, sgpearse notifications@github.com wrote:

This does not occur on the 3rdPartyLibs branch, which is pending review.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

John Clyne National Center for Atmospheric Research 303.497.1236 (w), 303.809.1922 (c) clyne@ucar.edu

[vema1818]

This error occuring to me as well in VAPOR3.9. For example, I want to smooth the variable with scipy.ndimage module gaussian_filter, which is not loading properly. I spent lot time how to do it, but arrives to this discussion. All vapor_utils use numpy but no scipy packages. Hopefully someone suggest the solution.

[sgpearse]

@vema1818 Could you provide the script you're trying to execute? I'm guessing you're on Ubuntu20/22 but that's a guess. I'm able to import scipy on MacOS/arm64 with the script below:

[vema1818]

Thanks Pearse. Yes, I am on Ubuntu 20. This problem occurs only in VAPOR3.9 not in 3.8 and 3.4 as I checked. The script I use is

import numpy as np import os from scipy.ndimage import gaussian_filter

dx=0.1456 fourpi=12.57 bh = np.sqrt(bx2+by2) jx = (np.gradient(bz,dx, axis=1) - np.gradient(by,dx, axis=2) )/fourpi jy = (np.gradient(bx,dx, axis=2) - np.gradient(bz,dx, axis=0) )/fourpi jz = (np.gradient(by,dx, axis=0) - np.gradient(bx,dx, axis=1) )/fourpi jtot = np.sqrt(jx2+ jy2+ jz**2) jtot1 = gaussian_filter(jtot,sigma=1)

I think, 3.9 version has some issues as I found. So, I am using 3.8.1 version as of now.

[StasJ]

This is broken on the AppImage installer on Casper