SNAP8 configured with Python 3.6

development python
#1

Good morning,
is there anybody who has successfully configured SNAP8 with Python 3.6? I am experiencing several issues and would like to have some exchanges.
Thank you!
Mariapina

#2

I use Python 3.6 (Windows and linux) because it is the most recent version for which SNAP has been tested extensively. Please describe your platform (OS, Python distribution you used) and the issues you encountered. If you did not follow Configure Python to use the SNAP-Python snappy interface you might find it helpful to try the installation using those steps (let us know if any of the steps are not clear so the document can be improved).

#3

Thank you, the manual configuration worked properly (Ubuntu 18.04.2, Python 3.6.13). Good to know that you are also using it.
Best,
Mariapina

#4

It would be great if you could prepare some detail document on how to configure snappy with mac. I have tried several times with no success.

#5

I retired from a job where we used Macs, but don’t have one at present. Recent macOS versions have some changes (I assume made for improved security) that may be causing problems. I’m confident that the issues can be solved, but it will need someone who is comfortable working with command-line tools and compilers. What we know so far:

  1. Users should avoid using Apple’s Python as there is a risk that changes will affect system tasks that use Apple’s Python

  2. A user-installed Python-3.6 needs to support the same compiler-run time libraries as the supplied binary wheels. These wheels have been well tested and are widely used on other platforms, but we haven’t identified a suitable Python 3.6 package (there are dozens, but most are too old). Macports does provide binary packages – users can check for their macOS version at [https://ports.macports.org/port/python36/stats?days=30&days_ago=0](https://Macports Python-3.6 builds). There is a learning curve with macports and it is based on command-line tools.

  3. Without a suitable Python-3.6 binary package, the alternatives are either install a newer Python-3.x binary package and use that to build a jpy wheel, or build Python-3.6 locally. There are several systems that allow you to build python relatively easily: Brew, Macports, and Fink. Either approach should work, but both require some level of comfort with the command-line and software development.