Hacker News Reader

Comment by reaperducer | Hacker News Reader

Making maps without an external service on macOS is an exercise in dependency hell.

About once a month, I have to run a script I wrote that generates a list of lat/lon values from company data to generate maps in python. I have to keep a separate machine on an old version of macOS just for this task because python3 broke everything.

ALL of the tutorials online are out of date, broken, or simply don't work. I even contacted the people at Geofabrik and offered to pay them for a solution, but they wouldn't touch the problem with a ten meter cattle prod.

I guess I'll spin up an Ubuntu VM and see if this solution works for me. But the state of mapping is abysmal. Everything is an online service or a GUI. And none of the online services can do batch processing.

View on HN

R and Python both have state of the art mapping tools and libraries that you can either script or use interactively in a notebook to accomplish literally any conceivable task using vector or raster data, from a simple choropleth map to geographically weighted regression to analysing satellite imagery using pre-trained models. Making publication-quality maps using point data is absolutely trivial, and there are a number of high-quality learning resources available (https://geo-python.github.io/site/, https://automating-gis-processes.github.io/site/)

I started working with spatial data on the Mac platform before Python 3 had gained much traction, and I can assure you that Python 3 really didn’t break much to speak of. Every major library and tool works as well as ever, and in most cases is vastly improved.

I think we have very different definitions of trivial

In the context of Hacker News and the parent comment’s lack of interest in GUI-based tools, I’m pretty confident in my definition.

I never trust a “trivial” on HN.

Another option for you to try: I recently used https://pypi.org/project/py-staticmaps/ to generate maps in Python 3 on my Mac, and even got the Cairo option working. My script to generate maps for state and country capitals: https://github.com/pamelafox/chrome-cards/blob/master/genera...

Staticmaps is great! I've used it to generate a map of our area pointing out where the BBQ area is for guests, I've used it to generate maps of bike rides and hikes I go on, it's really fun. Definitely needs a bit of light GIS knowledge but nothing someone with decent Python skills and a free-ish weekend can't pick up.

I had absolutely zero trouble on Windows 10/Ubuntu 22.04 a couple weeks ago doing mapping. I downloaded the JSON location data from my Google account and this is what I ended up with: cartopy + matplotlib + shapely.

I believe the examples for cartopy mostly work out of the box.

Can't speak for Mac OS but I assume all those are conda-installable there as well.

What's your source data? I am currently working on standalone map renderer written in Nim. My source data is are vector tiles and it works reasonably well...

Weird. I've had no issues on several recent versions of Mac. I use Brew to manage the base, add gcc, and use virtual environments. No issues whatsoever.

Reply on news.ycombinator.com