Hello Bruno Le 08/05/2018 à 11:08, Bruno P. Kinoshita a écrit :
> Without looking at the Python module with more calm, isn't it common > to use Jython for creating a bridge from Python to access Java > objects? Any reason for using JPY instead? (hadn't heard about it). > In my understanding, JPython is not really a Java-Python bridge but rather a Python implementation running on the JVM. Executing Python scripts in JPython make easy to interact with Java since they are running on the same JVM. However (if I understood correctly), invoking native C/C++ code from JPython is complicated. The Python developers I spoke to told me that they like Python because it is easy to invoke C/C++ code form Python. They use Python with GDAL and other C/C++ libraries, together with Python modules (NumPy, SciPy) which use C/C++ code for fast calculations on arrays. For example the Open Data Cube project (https://www.opendatacube.org/) relies on Python with those C/C++ libraries. Interoperability between Open Data Cube and Apache SIS is actually the requirement that trigged this Java-Python bridge work in GeoAPI. So if I understood correctly, it would not be possible to execute Open Data Cube on JPython (but I did not tried). The JPY project allow us to have Open Data Cube executed on the native Python interpreter, and still communicate with the JVM. > At work (research company) most scientists are using R and Python. So > perhaps if it's simple enough, there could be something similar with > rJava? > I was not familiar with R-Java, thanks for the tip. However my goal here is not to create a generic Python-Java or R-Java bridge, but only bridges for the specific case of GeoAPI interfaces in those two languages. Since we have no GeoAPI interfaces in R, the question of R-Java bridge does not yet apply in this context (unless someone volunteer, of course). > (… snip …) And most sections contain a short example, similar to what > you would see in a Jupyter notebook. When I click on "Spatial > representation", I see the description of the types, but not how to > use it. > Yes, documentation needs to be improved. GeoAPI interfaces should be about the same (for most parts) in Java and Python, so a documentation would be common to both of them. I do not know yet what would be the best way to share examples in a cross-languages fashion. > If there are some issues for this API, I'd be happy to spend some > spare time reading the docs, and trying to suggest a few paragraphs > for installation, requirements, and some examples. That way I will > learn about it as well. > Thanks! Martin
