These were basically my points:
- new feature for our software should be to classify sensor data with small neuronal nets → we need mature NN libraries, thus either Python or Julia
- there should be an active community → Python and Julia have, both also have their own conferences
- we supply scientists → Julia is focused on that, Python is also widely adopted in R&D, but before it is compiled to e.g. C its code runs slow → for rapid prototyping Julia seems to be better
- we need a UI → both offer to compile its code as library to use it for GUI programming of our choice
In effect, except of the potential faster development, Python and Julia were on a par, so I decided to use Julia.
Except of the last point, I was right. Meanwhile I struggle with a point I did not have in mind - the development policy. I work with Python for 15 years now, despite I never fall in love with it. But I never had regressions that blocked me, for known issues workaround were published. I also cannot remember a case where a regression was known but a new release was issued. For Julia this is not the case. Take the case of PackageCompiler. Since Julia 1.11 one cannot use more than one thread. This reduces the compilation speed a lot. The problem was already reported in October. But since it was not mentioned in release notes, I run into that. Now I know, but it caused me hours to find out and thus of course also unnecessary frustration. Now I see that Julia 1.11.5 was released and despite this ongoing discussion, the issue is still not mentioned in the release notes and a new release was made despite there is a known regression. However, for this issue I already created a separate discussion thread.