Templating NML with Python - the non-comprehensive guide...

[2TallTyler]

For those trying this tutorial with a more recent version of Python (I'm using 3.8 ), the only change required is the Print syntax, which now needs parenthesis around the string to be printed:

Code: Select all

print("Running script...")

Andy, any plans to dust off this tutorial with an explanation of how to import data from a separate file?

[Eddi]

not much of a tutorial, but i use this to import a "tab-separated-values" (.tsv) table.
https://dev.openttdcoop.org/projects/ce ... ts/read.py
(note that this is a python2 file, and i've made no attempts at converting it to python3)

[andythenorth]

Andy, any plans to dust off this tutorial with an explanation of how to import data from a separate file?

Not so much

The general trend for templating newgrf has been to import something like a .csv file, or similar. I think the idea is that it's easier for non-technical contributors.

I tried that for FISH, in a few different ways using .cfg files.

Now I just declare everything directly in python files and import those.

Using non-python file formats inevitably requires dealing with escapes, and parsing data types (e.g. strings vs integers, strings vs. lists) etc.

It's not hard to open and parse a .csv file. Something like yaml or json might also be a valid route.

But I'd be inclined to put values in python dicts or classes and just import them.

The FIRS approach is probably too complex as a starting point, but the cargos are the simplest example https://github.com/andythenorth/firs/bl ... os/acid.py

That uses a Cargo class which provides many benefits, but could have just been a dict in a simpler world.

If you're doing houses, you could probably put them all in one python file, houses.py or so, with house_1 = {keys, values}, house_2 = {keys, values} etc. Then you just import houses.py to your main script, and you use it as houses.house_1[key] or houses.house_1.items() or whatever you need for the case you have.

Hope that helps, happy to expand.

[2TallTyler]

Thanks for the informative reply, and for suggesting the cargos as a simple example. After some experimentation of my own I came to a similar conclusion that my original plan is probably more trouble than it's worth.

I'm moving forward with splitting up the set into smaller NML files which are then combined with a Python script based on how your tutorial added the header. Dead simple and all in NML so no translation problems or even templating, but it eliminates the problem of scrolling through a wall of code trying to locate a certain piece.

[andythenorth]

I'm moving forward with splitting up the set into smaller NML files which are then combined with a Python script based on how your tutorial added the header.

That makes sense

I'm not sure that houses would have the same pay off from templating that e.g. industries have. FIRS has many callbacks, complex industry layouts etc which all benefit from templating.

The major reasons I force all the industry or vehicle data into python are:

• templating
• can auto-generate all the docs, which possibly helps players
• can then also use the docs to QA / check things, much faster for some things than play-testing in game
• automated validation, python can find bad values and stop compiling much much faster than nmlc finds them

Also a question...how long does your house grf take to compile currently?

[2TallTyler]

Also a question...how long does your house grf take to compile currently?

Less than 10 seconds. It’s only about 1,850 lines of code and I’d imagine the lack of sprites helps.

[Zardoz89]

Good thread. I got the idea, and go a step beyond. Here it's a prototype of the idea : https://github.com/Zardoz89/spainset-en ... ild_nml.py

Simply, reads a CSV file generated from LibreOffice (or Excel), and generates multiple PNML files for each row. Also, uses "Airspeed" (a port of Velocity template engine to Python) that allows to apply logical conditions (if,else, foreach...). The real utility it's for creating a vehicle set. or something similar that have a lot of boring repeating NML code with a few exceptions. I have intention of using it to try to make a Train set GRF.