Add numpy and scipy to PythonCaller

Related products: FME Form

Add numpy and scipy to FME's python and support for treating raster features as numpy arrays in PythonCaller. Numpy and Scipy include a whole host of tools that are really useful and powerful for handling mainly raster data and to a lesser extent point data. I would love if it was easy to use these tools to manipulate features in a PythonCaller
@ryanproulx You are right that you can always use a custom python interpreter, which is what I'm currently doing. The real problem is mainly that there no easy way to access and manipulate raster features as numpy arrays.
The FME distribution is already quite large, I'm not sure I think it's such a good idea to add tens of megabytes for a couple of libraries that less than 1% of FME users will use. It's not very difficult to set a custom interpreter and install them there.

While I agree that the number of FME users needing additional python packages is quite small, I do wonder what percentage of them need the same libraries (like scipy) and if it could be an install option.


Would it be the time to add a FME Package for stats/analytics (scipy, maybe panda..) on the hub? These Python packages and a class with some basic statistic tools in a PythonCaller would help a lot of people without impacting the FME distribution size. @david_r if well done, it could help more than 1%, especially if Safe adds one day output ports to PythonCaller.