(Incomplete) guide to Python + NML here
viewtopic.php?f=68&t=58390
The principal benefit of using python is separation of content (data) from presentation (nml code). Any other similar language would work just as well.
For me this provides three useful benefits:
1. Removes repetition
Instead of writing out all the nml for every vehicle (or industry), including all the switch chains, the nml is written once in a template.
This means that if nml needs maintained / changed, it's changed once in a template and applied everywhere.
The downside is that it's possible to construct very complex conditional templates (I do), and these can require a lot of design and debugging themselves (I make this mistake).
Similarly, global values can be used for things like vehicle speeds, refittable cargos etc. Changing these in one place changes them everywhere they are used. Pure nml would require careful find and replace, or manual changes.
2. Offload calculation and value checking
Defining properties (for e.g. vehicles) in python means they can be checked for valid values trivially, before they reach nml (this means errors are found faster).
Properties in python can also be used in calculations, which are then passed to the nml template, for example I calculate capacity for vehicles based on length and intro date.
NML can do really quite complex calculations, but it's not as flexible as python and it makes for hard-to-read nml code.
3. Re-use the values elsewhere.p
Because the vehicle / industry / cargo properties aren't locked up in nml files, they can be re-used by other programs. So for example
* I auto-generate html docs + images for all my newgrfs
* I auto-generate cargo sprites for vehicles
* I use a shared library of cargo labels, refits and other constants between all my newgrfs
None of this is needed for a simple grf, but for making multiple large sets, it would be awful to not have the data separate from the nml files.
Doesn't have to be python. Other approaches are available.
