#30DayMapChallenge Day 7: Accessibility Today's tip: calculate access from a location in one line of code with `mb_isochrone()` in the mapboxapi R package. Then, visualize creatively with a comparison slider in mapgl. Learn how: walker-data.com/posts/time-a...

Interactive heatmaps that fade away to reveal points on zoom - a useful technique for handling dense spatial data. Learn how to do it: walker-data.com/posts/mapgl-...

My election night R code worked last night! 🎉! Inspired by @andrew.heiss.phd & @gosterhout.bsky.social 's #PositConf2025 talk re Idaho's website built with R, Quarto & the {targets} #RStats 📦 My results page: www.district2framingham.com/posts/2025-1... Their repo: github.com/andrewheiss/... 1/2

Today's mapgl tip for Day 6 of #30DayMapChallenge: Dimensions Visualize your spatial data in 3D with `add_fill_extrusion_layer()`! The key is `fill_extrusion_height`; map a column of heights here or use a helper like `interpolate()` to control how you extrude your polygons.

Next semester I'm teaching an advanced #GIS course, Urban & Business GIS, for the first time in several years. My question to you: what are the *most important* skills / traits you are looking for in new geospatial hires in 2026? I have my thoughts, but I'd love your insights!

I got a lot of interest in yesterday's geolocation / Overture Maps places proof-of-concept. I've published the app here for you to try out: walkerke-restaurant-finder-app.share.connect.posit.cloud For #rstats devs, I've also published the source code here: gist.github.com/walkerke/4a1...

#30DayMapChallenge Day 5: Earth Today's tip: in #rstats mapgl, use `mapboxgl(style = mapbox_style("standard-satellite"))` to get @mapbox.com high-quality imagery on an interactive globe in one line of code. Even better - you get 3D terrain out of the box!

Proof-of-concept for an in-app geolocation control: - Find your location, draw a buffer (or isochrone, etc.) around it; - Query a massive tileset (64 million places from Overture Maps) to find nearby places; - Deliver that information to your app users.

Today's mapgl tip: use `add_geolocate_control()` to add a geolocation control to your map. Click to fly to your location; in Shiny, use `input$MAPID_geolocate` to get the coordinates. Great for building apps designed for use in the field! #30DayMapChallenge Day 4: My Data

1.7 million Texas oil & gas wells at your fingertips. Lasso wells on the map, download the data instantly. From selection to export in seconds. (By the way, this is powered by Shiny and #rstats)

311,000+ Texas Land Survey System areas. Explored over satellite imagery in a single line of code. Modern geospatial technology can take your land & ownership mapping to the next level. #30DayMapChallenge Day 3: Polygons

For Day 2 of #30DayMapChallenge (Lines): did you know that mapgl has a built-in measurement tool in its draw control? Use `show_measurements = TRUE` to interactively measure line distances (and polygon areas) on your map. #rstats

440,000+ oil and gas wells in Oklahoma. Intelligently clustered and smoothly displayed on a map with click-to-zoom. mapgl tip: use `cluster_options()` on a circle or symbol layer, no extra configuration required! #30DayMapChallenge Day 1: Points

The #30DayMapChallenge starts tomorrow across social media platforms! Pay attention to this hashtag if you love maps - there will be tons of creative examples from around the world.

I've been posting about PMTiles for geospatial visualization for good reason. However - you can push mapgl quite a ways without tiles! In this post, I cover how to map all 85,000+ Census tracts in the US without pre-tiling, along with performance tips. walker-data.com/post...

Two years later, {mapgl} has grown into the R package I always wanted. And I'll freely admit that it would not exist without the help of LLMs. I'm speaking tomorrow about my experiences with LLM-assisted R development. Register here: docs.google.com/forms/d/e/1F...

A lot of you are asking about PMTiles - one of the most exciting innovations in geospatial in the past decade. My most recent blog post helps you get started. Use this reproducible workflow to learn how to create, explore, and visualize PMTiles. walker-data.com/post...

On the blog: building an AI-powered location explorer, updated to Claude Sonnet 4.5! Learn how to use the ellmer and shinychat R packages to integrate chat into your Shiny apps. Check it out: walker-data.com/posts/ai-loc...

The 1.4 release of @duckdb.org supports using a DuckDB database to serve vector tiles! Of course, I had to try this out in R. Check it out: all 242,000 US Census block groups dynamically served as vector tiles from a DuckDB database, displayed on a MapLibre map from R in Positron.

My AI dev team right now: - Codex (Pro plan) for high-level planning and architecture; - Claude Code (Pro plan) for execution; - Codex coming off the bench for execution when I hit Claude's usage limits. Still trying to find the right cost / service fit, but this works well!

8.1 million Census blocks. 1.7 million Texas wells. Fast visualization, instant queries. The magic tech: MapLibre + PMTiles. The glue: R + Shiny. Seriously.

All 8.1 million US Census blocks. Visualized smoothly in 3D. Instant population and housing totals from a lasso selection. All running seamlessly in the browser, no traditional backend. While everyone’s talking about AI, it’s an incredible time for geospatial tech.

Charted: post-COVID income growth for America's largest metro areas. Since 2021, Miami's median household income grew over 26 percent. Orlando and Tampa weren't far behind. Among the US's largest metros, Seattle (10.5 percent) and Minneapolis (12 percent) have lagged.

Most accessibility analyses use "average" travel times. But rush hour isn't average. Your "15-minute service area" might be 25 minutes when it actually matters—when patients need care, employees commute, emergencies happen. Learn how to handle this: walker-data.com/posts/time-a...

Just listened to a really thought-provoking conversation about LLMs and open source on Posit PBC's "The Test Set" podcast. Given that LLMs recommend solutions they *know about* from their training data, does this discourage innovation in open source?

Pleased to contribute data analysis and graphics to this Nicholas Kusnetz story for @insideclimatenews.org on the coal-fired power plant fleet it has helped build and finance beyond its own borders — notably in Indonesia. insideclimatenews.org/news/1910202...

And a shout-out to @kylewalker.bsky.social for the #mapgl #rstats package, which I used for the first time to make this interactive globe visualization walker-data.com/mapgl/

An anecdote about why LLMs matter for open source software developers: I maintain the tigris R package, which makes it easy to get geographic datasets from the US Census Bureau. In April, an issue on the Census website started blocking download requests via tigris.

Starting projects is easy. Finishing them? Hard. Then, maintaining them? That's the hardest part. Just look at the number of open source projects that haven't been touched in 2+ years. I've found LLMs to be essential for #3, tackling issues I wouldn't have otherwise.

Of my last 2 #rstats packages: One would have been far too complex and time-consuming for me to write without using LLMs; The other took me 2 days to get working with an LLM; it would have taken weeks (if done at all) otherwise I get LLM skepticism, but this matters a ton for open source devs.

On the blog: how to map hundreds of thousands of US Census blocks performantly with R and PMTiles. This is the first in a three-post series on production mapping workflows for big spatial datasets with PMTiles. Read the post: walker-data.com/post...

Super easy to use the mapgl package and @oldinsurancemaps.net to compare 1886 Richmond, VA to today in #rstats. The site of Gallego Mills, the ruins of which became an iconic photo after the fall of Richmond in the American Civil War, is now an office building and freeway.

One thing I'm noticing with Claude Sonnet 4.5 compared with older models - it's much more willing to tell me I'm wrong. "You're absolutely right!" has become a meme, but it's not what you want when you're workshopping ideas. This is the best feature of 4.5 for me so far.

Feeling stuck with data.census.gov unavailable during the government shutdown? Now's a great time to try out tidycensus. The decennial Census and American Community Survey are at your fingertips - and get pre-joined geographies for mapping.

The @maprva.org surveillance map, ported to #Rstats using mapgl, osmdata, sf, and dplyr. gist.github.com/mhpob/17782b... Not 1:1 in terms of Ultra/Mapbox GL JS -> R, but pretty close! Original query: overpass-ultra.us#query=url%3A... cc @mackaszechno.bsky.social @kylewalker.bsky.social

New post: mapping 650,000+ Census blocks in Texas in 3D with PMTiles You've heard me talk about @protomaps.com's PMTiles for a while and how transformative this tech can be for geospatial visualization. In my latest post, you'll learn how to use PMTiles in R.

Your company's affordable production mapping stack: - R + Shiny (seriously) - PMTiles for massive data visualization; - Cloudflare R2 Workers for serving data securely; - MapLibre for your maps; - Docker + a cloud service (I like Fly.io)

In this week's upcoming blog post - learn how to make this visualization! Follow here, coming soon -

On the blog: enable lasso selection in your R / Shiny web mapping applications. Learn how: walker-data.com/post...

Real-time spatial analysis in your R Shiny apps? The {mapgl} package - and its TurfJS integration - make it possible. Watch as Voronoi polygons instantly update on marker drag. Pick a metro and try it out: gist.github.com/walkerke/f4d...

redistio now uses @kylewalker.bsky.social's mapgl for interactive redistricting map editing! redistio is an #rstats package that uses Shiny to draw redistricting maps locally. Unlike other district editors, this provides access to our simulation algorithms, interactively! Short demo:

Preview the 2.5 billion buildings from @overturemaps.bsky.social in a single line of R code? Why not? Check out the pmtiles package: github.com/walkerke/pmt...

R-Universe by @rOpenSci is an extremely useful project. My latest R package, {pmtiles}, is not a good fit for CRAN as it bundles a pre-compiled Go library. With R-Universe, I was able to get installable cross-platform binaries built in hours! #rstats walkerke.r-universe....

Introducing {pmtiles}: an #rstats package to create, view, and interact with PMTiles. PMTiles is one of the most exciting geospatial innovations in recent years. Map tiles are stored & served from a single file, making massive geospatial data visualization widely accessible.

The US Census website / data.census.gov are down. R users, however, can still access data: - Geographies (via FTP): tigris github.com/walkerke/... - Decennial / ACS with geographies: tidycensus walker-data.com/tidy... - All datasets: censusapi www.hrecht.com/censu...

I regret to inform you that Claude Sonnet 4.5 cannot one-shot developing an R package of moderate complexity. We now return you to your regularly scheduled AI hypefest programming.

I've been having to build "user state" a lot more into my Shiny apps, where the app remembers where the user's last session left off. A couple #rstats packages that caught my eye: {polished}: auth and admin for Shiny {shinystate}: state preservation for Shiny apps

MapGL Lab - a SaaS that makes the power of the R mapgl package available to everyone - is coming together nicely. Visualize and re-order layers on drag-and-drop, change your basemap, quickly style your data and explore it. The goal: easy mapping, minimal overhead.

Despite what people tell you - You CANNOT one-shot your data analysis with an LLM. People who say you can are either: Lying for clout, or: Doing something unrealistically simple. Beware the hype and be careful with shortcuts. (This is advice to myself, by the way)

I've been as loyal a Claude Code user as there is, but I'm really having a good experience with Codex. Codex feels like much less of a "yes-man" and more of a development partner who finds pain points and makes suggestions. Maybe using both on $20 plans is the sweet spot.