posts/2024-06-06-designing-arcgisgeocode

[utterances-bot]

Josiah Parry - Making a Ridiculously Fast™ API Client

https://josiahparry.com/posts/2024-06-06-designing-arcgisgeocode

[rafapereirabr]

This is really fantastic work!!! thanks for sharing it, Josiah. Is there a chance you might also develop an R package to send calls do ArcGIS Network Analysit? I'm specifically thinking of calculating many-to-many travel time estimates by car using historical speed profile of road segments

[JosiahParry]

Thanks @rafapereirabr ! It is definitely something we want to make and have been planning on. I've been toying with prioritizing mapping capabilities but no one has asked for it.

Looking at what is possible for us is based on the documentation here: https://developers.arcgis.com/documentation/mapping-apis-and-services/routing/

Which of these features would be most useful for you? That way I can prioritize development based on an actual use case

[rafapereirabr]

Thanks, Josiah! What would be most useful for us is the 'Travel cost matrix' capability.
We currently use this modified Python script (the original version was developed by Chris Higgins at Toronto Uni), but it takes a really long time to run (say, compared to {r5r} for example). A simple change we could make is to save the output to .parquet files, which would certainly improve overall performance, but I guess the main bottleneck is in the routing itself. Anyway, we just renew our ArcGIS license, so if you want to, I'd be happy to be the guinea pig testing your package that does this travel cost matrix calculation

[JosiahParry]

Thank you! I think that is probably the simplest of the endpoints so it could be quite possible. Out of curiosity, do you have a ballpark estimate of computation times for O-D pairs?

How many OD pairs are you typically working with?

[rafapereirabr]

On average, the number of OD pairs is around 100 million, but in the largest area the number of pairs gets over 1.8 Billion [meme austin powers] :)