Jon is a developer

Nim is a statically typed compiled systems language with C interoperability. In the sphere of game development in Nim, developers often have to roll their own physics engine if their game requires physics. This presents a great opportunity to learn but can be a daunting task if the developer isn't already familiar with the underlying math.

This binding solved a personal need to create a physics-based game in Nim and by creating complete documentation and testing I was able to effectively solve this problem for others.

Click to play demo

Wikimedia uploads routine data dumps from their projects (Wikipedia, Wikibooks, Wiktionary, etc.), and these are accessible by their data downloads site. To query for assets on this site, people often resort to webscraping which provides a short term solution but can be taxing for Wikimedia's servers and is prone to failure.

To solve this issue, I created an API that accesses the index of the site and wraps common query operations around idiomatic Python classes.

Below is a typical usage of the library.

from wiki_data_dump import WikiDump, File
import re
from typing import List

wiki = WikiDump()

xml_stubs_dump_job = wiki["enwiki", "xmlstubsdump"]

stub_history_files: List[File] = xml_stubs_dump_job.get_files(
    re.compile(r"stub-meta-history[0-9]+\.xml\.gz$")
)

for file in stub_history_files:
    wiki.download(file).join()
        

Online Python interpreters usually don't have operability with the user's actual file system, and you have to upload individual scripts and files to use them within the interpreter, which can be tedious. In addition to this, the code is often run within a cloud service (such as Google Colab) which may not always be desired.

Pyodide is a WASM-based Python interpreter and when provided a native file system directory handle as a virtual file system mount point you can use the interpreter very much like a native Python installation: file reads/writes are updated in real time.

This is a huge quality improvement over typical online Python interpreters, and with the addition of mypy static type checking (again, entirely in the browser) this nears even closer to a native Python installation.

For another project, I needed a way to navigate the Wikipedia category hierarchy quickly, and without the Wikipedia API. Originally using the above wiki_dump project, I constructed a pipeline to automatically extract category data from the Wikimedia assets, transform it into a graph structure, and upload it to GitHub Pages.

There were multiple limitations that needed to be properly handled: the GH runners have a fairly low space limitation, and a GH Pages artifact can only be so large. This necessitated using streaming to process the data as it was being given to the runner without collecting much memory over the course of the run, and creating a compact data format that can be easily parsed.