Josiah

@josiah.rs

2638 Followers 480 Following

putting R in production @ricochet.rs doing spatial stats @ Esri blog: josiah.rs

library(sf) library(arcgisutils) library(arcgisrouting) set_arc_token(auth_user()) stops <- st_sf( name = c("Start", "Middle", "End"), geometry = st_sfc( st_point(c(-122.4194, 37.7749)), # San Francisco st_point(c(-122.0312, 37.3318)), # Cupertino st_point(c(-121.8863, 37.3382)), # San Jose crs = 4326 ) ) resp <- find_routes(stops) # print the directions resp$directions |> tidyr::unnest(directions) |> dplyr::select(length, time, text) #> # A tibble: 22 × 3 #> length time text #> <dbl> <dbl> <chr> #> 1 0 0 Start at Start #> 2 0.148 0.828 Go northeast on Market St toward Van Ness Ave / S Van Ness Ave #> 3 0.603 2.44 At the traffic light, turn right on 10th St #> 4 34.9 36.4 Bear left to merge onto US-101 S toward San Jose #> 5 5.29 5.45 Take exit 398B on the right to merge onto CA-85 S toward Cupe… #> 6 0.152 0.152 Take the exit on the right to I-280 toward San Jose / San Fra… #> 7 1.48 1.48 Keep left at the fork to merge onto I-280 S toward San Jose #> 8 0.230 0.793 Take exit 11 on the right to De Anza Boulevard #> 9 0.240 0.766 At the traffic light, turn right on N De Anza Blvd #> 10 0.0996 0.209 At the traffic light, make a U-turn and go back on N De Anza … #> # ℹ 12 more rows
Error in `scale_fill_viridis_d()` at 2026-06-22/R/dev-block-grid.R:21:1: ! Continuous value supplied to a discrete scale. ℹ Example values: 5, 6, 3, 4, and 1. Show Traceback
[services] web = { image = "nginx:latest", environment = { DB_HOST = "postgres", DB_PORT = 5432 }, resources = { limits = { memory = "512M", cpu = "0.5" } } }
use anyhow::bail; use arrow_extendr::FromArrowRobj; use extendr_api::prelude::*; use geo_traits::{CoordTrait, PointTrait}; use geoarrow::array::{GeoArrowArrayAccessor, PointArray}; #[extendr] fn geoarrow_centroid(x: Robj) -> anyhow::Result<Doubles> { let res = PointArray::from_arrow_robj(&x)?; let (sum_x, sum_y, n) = res.iter_values().filter_map(|xi| xi.ok()).fold( (0.0_f64, 0.0_f64, 0usize), |(sx, sy, n), v| { let (x, y) = v.coord().unwrap().x_y(); (sx + x, sy + y, n + 1) }, ); if n == 0 { bail!("cannot compute centroid of empty array"); } let res = [sum_x / n as f64, sum_y / n as f64] .map(Rfloat::from) .into_iter() .collect::<Doubles>(); Ok(res) }
ggplot() + geom_sf(data = nodes[c(i, nb[[i]])]) + geom_sf(data = nodes[i, ])
Screenshot of a failed github action workflow "Check for non-API calls." Text output: Run RESPONSE=$(curl -sf "https://dev.vm.ubuntu.ricochet.rs/app/non-api-checks/check?pkg=$PKG_NAME") { "last_updated": "2026-04-10 17:52:44 CEST", "has_non_api_calls": true, "symbols": [ "R_UnboundValue" ] } Error: non-API calls detected in b64 Error: Process completed with exit code 1.
library(rlang) macro_fun <- function(name) { f <- new_function(alist(x=), quote(x^2)) nm <- sym(name) env_bind(caller_env(), !!nm := f) } macro_fun("square2") square2(5) #> [1] 25
Below is the output from claude ❯ help me plan and priortize my week. tomorrow is monday ⏺ readMcpResource(Read resource "objectives://all" from server "jebbysays") ⎿  { "contents": [ { … +24 lines (ctrl+o to expand) ⏺ Here's a proposed week plan based on your objectives, priorities, and deadlines: --- Monday - Ricochet: Review and rewrite PR #785 (user avatar support) — this is priority 1 for the launch objective - Health: BJJ (deadline is March 24, so Monday works) Tuesday - Ricochet: Continue/finish PR #785 if not done - jebbysays: Migrate to Leptos SSR — this unblocks all other frontend work, tackle it early in the week while you're fresh Wednesday - jebbysays: Add tower-sessions + use database sessions (these are related, do them together) - Health: Gym Thursday - jebbysays: Build navbar with user icon (depends on sessions being in place) - Open Source: Review ANIME paper Friday - jebbysays: Implement MCP resource for single task by ID + dark mode (smaller tasks, good for end of week) - Health: Gym or BJJ
Hand drawn meme top panel says "3 rules: no wishing for death. no falling in love. no bring back dead people" bottom let panel says "i wish to understand kubernetes" the bottom right says "there are 4 rules"
claude saying: "You're right, I'm contradicting myself and waffling. Let me be direct about what I actually don't know and what's causing the contradiction."
Astronaut meme template. The one looking at the earth "wait, i was the bug?" the one with the gun "always have been"
a shiny app with code on the right in the Positron IDE. Shiny app has an action bar on the left with "Upload" and "Analysis" with a file input and geometry input and a layer summary. The middle is an interactive map with puple and orange points.
pkg_size_ <- function(pkg) { # https://stackoverflow.com/questions/70991241/installed-r-packages-size pkg_dir <- system.file(package = pkg) if (pkg_dir == "") { return(fs::as_fs_bytes(NA)) } res <- system( paste("du -sh", system.file(package = pkg), "| awk '{print $1}'"), intern = TRUE ) fs::as_fs_bytes(res) } pkg_size <- function(pkg) { fs::as_fs_bytes(unlist(lapply(pkg, pkg_size_))) } pkg_size_recursive <- function(pkg) { deps <- c(pkg, unlist(tools::package_dependencies(pkg, recursive = TRUE, which = "Imports"))) size <- pkg_size(deps) total_size <- sum(size) data.frame( pkg = c(deps, "total"), size = structure(c(size, sum(size)), class = c("fs_bytes", "numeric")) ) |> dplyr::arrange(-as.numeric(size)) }
Wrong Approach and Repeated Misdirection Claude consistently proposes incorrect solutions that you have to interrupt and redirect, often multiple times per session. You could reduce this friction by front-loading more explicit constraints and architectural context in your prompts — for example, specifying which layer a fix should target or which patterns are off-limits before Claude starts working.
Screenshot of "cargo clean" showing that 110269 files were removed cleaning 261.9GiB of strorage
rendered reprex of the following: penguins |> subset( select = c("island", "bill_len"), subset = island == "Biscoe" & bill_len > 55 ) #> island bill_len #> 186 Biscoe 59.6 #> 254 Biscoe 55.9 #> 268 Biscoe 55.1 penguins |> dplyr::select("island", "bill_len") |> dplyr::filter(island == "Biscoe", bill_len > 55) #> island bill_len #> 1 Biscoe 59.6 #> 2 Biscoe 55.9 #> 3 Biscoe 55.1
Interactive data visualization dashboard showing Palmer Penguins dataset. Left sidebar contains collapsible controls for selecting X axis (Bill length), Y axis (Bill depth), and color grouping (Species) variables using dropdown menus. Main panel displays a scatter plot with points colored by penguin species (Adelie, Chinstrap, Gentoo), built with Esri's Calcite Design System components in R Shiny.
Table 5 from Tidy Data. The description of which is "Table 5: A simple example of melting. (a) is melted with one colvar, row, yielding the molten dataset (b). The information in each table is exactly the same, just stored in a different way."
The code below shows us setting the class os our list ``` #[extendr] fn decode_(x: Strings, alphabet: Option<&str>, padded: Option<bool>) -> List { let alphabet = get_alphabet(alphabet, padded); let mut res = x .into_iter() .map(|xi| match base32::decode(alphabet, xi) { Some(bytes) => Raw::from_bytes(&bytes), None => Raw::new(0), }) .collect::<List>(); let _ = res.set_class(["blob", "vctrs_list_of", "vctrs_vctr", "list"]); res } ```
# make a new counter counter <- extendrtesting:::Counter$new() # create the callback function that is always evaluated task_callback <- function(expr, value, ok, visible) { counter$increment() TRUE } # register the callback addTaskCallback(task_callback, name = "rust_counter") # do some stuff 1 + 1 rnorm(4) counter$format()
input_raw <- strsplit(readLines("day11.txt"), ": ") from_raw <- vapply(input_raw, `[`, character(1), 1) to_raw <- lapply(input_raw, \(.x) { strsplit(.x[2], " ")[[1]] }) from <- rep.int(from_raw, lengths(to_raw)) to <- unlist(to_raw) library(igraph) # create the graph g <- graph_from_data_frame(data.frame(from, to)) # find the node for you and out find_node <- function(g, name) { idx <- which(names(V(g)) == name) V(g)[idx] } origin <- find_node(g, "you") dest <- find_node(g, "out") # find all simple paths paths <- all_simple_paths(g, origin, dest) length(paths)
# create listw object nb <- spdep::cell2nb(4, 4) listw <- spdep::nb2listw(nb) format.swm <- function(x, ...) { nbs <- vctrs::field(x, "neighbours") card <- spdep::card(nbs) out <- sprintf("(%s)", vapply(nbs, toString, character(1), width = 15)) out[which(card == 0)] <- NA out } as_swm <- function(listw) { if (!inherits(listw, "listw")) { cli::cli_abort("{.arg listw} must be an {.cls listw} from {.pkg spdep}") } class(listw$neighbours) <- c("nb", "list") vctrs::new_rcrd(listw, class = "swm") } as_swm(listw)
gif of a shiny application using an ArcGIS Online login. The nav bar has "Login" then after pressing navigated to a login page. After logging in redirected to the shiny application with the navbar showing the user information.
Screenshot of a youtube comment which says "I miss your takes on Twitter and your content on Youtube (understand a lot of work goes into the latter). Are you active on BlueSky or elsewhere?"
clip of function body in the linked github url
reports_raw <- readLines("input/day2.txt") is_safe <- function(.x) { # parse the line .x <- strsplit(.x, " ")[[1]] |> as.integer() # check monotonicity monotonic_decreasing <- all(.x <= cummin(.x)) monotinic_increasing <- all(.x >= cummax(.x)) # check the first condition if (!(monotinic_increasing || monotonic_decreasing)) { return(FALSE) } # calculate diffs diffs <- abs(diff(.x)) # check diffs if (any(diffs > 3) || any(diffs == 0)) { return(FALSE) } TRUE } vapply( reports_raw, is_safe, logical(1), USE.NAMES = FALSE ) |> sum()
Bench mark output from R showing that the R solutions performs 2461 iterations a second whereas the rust+r solutions performs 17362 solutions per second.
sum(abs(do.call(`-`, lapply(read.table("input/day1.txt"), sort))))