sp:spTransform detecting a "non finite transformation"

[Blevy2]

Hello,

I have been running VAST on my personal windows PC over the last few weeks without any problems. I tried to scale up my work by shifting to a Linux NOAA server and receive an error from sp::spTransform when I use fit_model on the server using the exact same input and scripts. After the extrapolation grid completes VAST outputs "non finite transformation detected" before listing several hundred lines of the form "x, y, Inf, Inf", as shown in the image below:

The x, y locations are not lat/lon values, which is confusing. After listing 250 of 976 rows as shown above, the below error is shown:

After looking into it I believe the issue has something to do with the projection assumed by sp::spTransform. However, spTransform is called within the package so I am not sure what to do about that. Additionally, since the same script worked on my personal windows machine, I wondered whether there was an issue with an old version of the sp package. I updated sp on the server, but the problem persists.

I see similar errors in #280 and #287, but they result from different functions in VAST where users were able to pass arguments that fixed the problem. Any idea what could cause this problem or how to fix it?

thank you for your help,

Ben Levy

[James-Thorson-NOAA]

Thanks for posting the issue here where I can track it easily. Can you make a minimal example (https://github.com/nwfsc-assess/geostatistical_delta-GLMM/wiki/How-to-create-a-minimal-example) and email it to me?

[Blevy2]

I am updating this issue based on what Jim and I found through exploring a minimal example. In short, we were able to identify the version of the package 'sp' as being a compatibility issue. The linux server I am using had sp version 1.5.0 installed, whereas Jim and I were running sp 1.4.6 and 1.4.7 without any problems.

After rolling back the version of 'sp' to 1.4.6 on the server I was able to move past this issue.

As a quick side note I am now having additional compatibility issues even after moving past the sp-related problem. First I had a problem with INLA version 22.05.03 which was fixed by rolling back to version 20.03.29. Now I am having an issue with Matrix, which was documented in other issues. I just wanted to mention them here in case others come across similar problems.

Thank you very much for your help, Jim.

Ben