Lists and map for tags, bursts, years. Do Base R + dplyr first.
An R list holds heterogeneous objects — data frames, vectors, metadata — in one named container.
by_site <- split(reef, reef$site) # base split to named list
class(by_site) # should be 'list'[1] "list"names(by_site) # site names as list names[1] "East" "North" "South"lapply(by_site, nrow) # rows per site$East
[1] 20
$North
[1] 20
$South
[1] 20map() applies a function to each list element and returns a list of results — always, unlike the sapply() simplification surprises in base R.
species_lists <- map(by_site, ~ unique(.x$species)) # .x is each site df
species_lists # list of character vectors$East
[1] "Parrotfish" "Surgeonfish" "Grouper" "Snapper" "Damselfish"
$North
[1] "Parrotfish" "Surgeonfish" "Grouper" "Snapper" "Damselfish"
$South
[1] "Parrotfish" "Surgeonfish" "Grouper" "Snapper" "Damselfish" lengths(species_lists) # how many species per site East North South
5 5 5 Also:
by_port <- split(fishery, fishery$port)
map_int(by_port, nrow) Broome Cairns Mooloolaba
9 9 6 Also:
by_zone <- split(acoustic, acoustic$reef_zone)
map_chr(by_zone, ~ .x$receiver[1]) Crest Lagoon Slope
"R2" "R9" "R11" Typed maps map_dbl(), map_int(), map_chr() assert return types — the purrr analogue of vapply() from Level 1.
mean_counts <- map_dbl(by_site, ~ mean(.x$count)) # numeric vector out
mean_counts # named by site East North South
10.95 8.80 9.80 sum(mean_counts) # sanity check total[1] 29.55map2() walks two parallel vectors element-wise — year paired with species, depth paired with pitch.
labels <- map2_chr(fishery$year, fishery$species, ~ paste(.x, .y, sep = "-"))
head(labels) # first few composite ids[1] "2018-Snapper" "2018-Grouper" "2018-Tuna" "2018-Shark" "2019-Snapper"
[6] "2019-Grouper"length(labels) == nrow(fishery) # lengths must match[1] TRUEAlso:
map2_dbl(reef$count, reef$depth_m, function(c, d) if (d > 0) c / d else NA_real_)[1:5][1] 0.6451613 1.9354839 0.1612903 1.1290323 3.3870968Also:
b1 <- biolog[biolog$burst_id == 1, ]
map2_dbl(b1$depth_m[-length(b1$depth_m)], b1$depth_m[-1], ~ .y - .x)[1:5][1] 0.26751027 -0.31068411 -0.02501225 -0.06485909 0.20001095pmap() maps over rows of a data frame — row-wise logic without an explicit index i.
ac5 <- acoustic |> dplyr::slice_head(n = 5) # tiny slice
summaries <- pmap_chr(ac5, function(animal_id, receiver, detections, reef_zone) {
paste0(animal_id, " @ ", receiver, ": ", detections, " (", reef_zone, ")")
})
summaries # one string per row[1] "A101 @ R11: 5 (Slope)" "A101 @ R2: 2 (Crest)" "A101 @ R6: 7 (Slope)"
[4] "A101 @ R1: 4 (Crest)" "A101 @ R5: 5 (Slope)" Also:
r3 <- reef |> slice_head(n = 3)
pmap_chr(r3, function(transect_id, site, species, count, depth_m) {
paste0(transect_id, " ", species, ": ", count, " @ ", site)
})[1] "T01 Parrotfish: 4 @ North" "T01 Surgeonfish: 12 @ North"
[3] "T01 Grouper: 1 @ North" Also:
f3 <- fishery |> slice_head(n = 3)
pmap_chr(f3, function(year, species, tonnes, port) paste0(year, " ", species, " ", tonnes, "t @ ", port))[1] "2018 Snapper 75.7t @ Cairns" "2018 Grouper 79.6t @ Mooloolaba"
[3] "2018 Tuna 23.5t @ Broome" map_dfr() row-binds data frame results — ideal for simulation replicates or per-group outputs.
set.seed(7) # reproducible draws
sp <- unique(reef$species) # species labels
sim <- map_dfr(sp, function(s) { # one df per species
tibble(species = s, draw = rpois(3, lambda = 5)) # three random counts
})
sim # stacked draws# A tibble: 15 × 2
species draw
<chr> <int>
1 Parrotfish 11
2 Parrotfish 4
3 Parrotfish 2
4 Surgeonfish 2
5 Surgeonfish 3
6 Surgeonfish 7
7 Grouper 4
8 Grouper 10
9 Grouper 3
10 Snapper 5
11 Snapper 3
12 Snapper 3
13 Damselfish 7
14 Damselfish 2
15 Damselfish 5Also:
ports <- unique(fishery$port)
map_dfr(ports, function(p) {
tibble(port = p, tonnes = runif(3, 20, 90))
})# A tibble: 9 × 2
port tonnes
<chr> <dbl>
1 Cairns 25.9
2 Cairns 59.2
3 Cairns 20.6
4 Mooloolaba 89.0
5 Mooloolaba 42.2
6 Mooloolaba 64.8
7 Broome 40.7
8 Broome 89.8
9 Broome 83.4Also:
rx <- unique(acoustic$receiver)[1:4]
map_dfr(rx, function(r) tibble(receiver = r, sim_det = rpois(3, lambda = 4)))# A tibble: 12 × 2
receiver sim_det
<chr> <int>
1 R11 9
2 R11 1
3 R11 4
4 R2 4
5 R2 8
6 R2 3
7 R6 5
8 R6 3
9 R6 2
10 R1 2
11 R1 3
12 R1 6safely() wraps a function to return result or error instead of stopping — essential on messy field spreadsheets.
raw <- c("12", "4.5", "NA", "oops") # mixed strings
parsed <- map(raw, safely(as.numeric)) # list of result/error
ok <- map_lgl(parsed, ~ is.null(.x$error)) # which succeeded
map_dbl(parsed[ok], "result") # extract numeric successes[1] 12.0 4.5 NA NAAlso:
raw_counts <- c("8", "12", "12a", "3")
map(raw_counts, safely(as.numeric))[[1]]
[[1]]$result
[1] 8
[[1]]$error
NULL
[[2]]
[[2]]$result
[1] 12
[[2]]$error
NULL
[[3]]
[[3]]$result
[1] NA
[[3]]$error
NULL
[[4]]
[[4]]$result
[1] 3
[[4]]$error
NULLAlso:
raw_t <- c("45.2", "", "88")
map(raw_t, safely(as.numeric))[[1]]
[[1]]$result
[1] 45.2
[[1]]$error
NULL
[[2]]
[[2]]$result
[1] NA
[[2]]$error
NULL
[[3]]
[[3]]$result
[1] 88
[[3]]$error
NULLwalk() maps for side effects such as messages or file writes and discards return values.
walk(names(by_site), function(nm) { # iterate site names
message("Site: ", nm, " — rows: ", nrow(by_site[[nm]])) # logging side effect
})
invisible(NULL) # walk returns input invisiblyreduce() folds a list with a binary function — combine vectors or merge tables stepwise.
by_an <- split(acoustic$detections, acoustic$animal_id) # list of nums
per_an <- map_dbl(by_an, sum) # one total per animal tag
total_det <- reduce(per_an, `+`) # fold scalar sums together
total_det # grand total detections[1] 91Tibbles can store list-columns — each cell holds a vector or mini data frame before unnest().
nested_bio <- biolog |> # biologging table
group_by(burst_id) |> # one row per burst soon
summarise(times = list(seconds), depths = list(depth_m), .groups = "drop")
nested_bio # list-columns visible# A tibble: 10 × 3
burst_id times depths
<int> <list> <list>
1 1 <int [20]> <dbl [20]>
2 2 <int [20]> <dbl [20]>
3 3 <int [20]> <dbl [20]>
4 4 <int [20]> <dbl [20]>
5 5 <int [20]> <dbl [20]>
6 6 <int [20]> <dbl [20]>
7 7 <int [20]> <dbl [20]>
8 8 <int [20]> <dbl [20]>
9 9 <int [20]> <dbl [20]>
10 10 <int [20]> <dbl [20]>nested_bio$depths[[1]] # first burst depth vector [1] 4.798260 5.065770 4.755086 4.730074 4.665214 4.865225 5.037666 5.419779
[9] 5.008206 5.065953 5.231723 5.179684 5.222335 5.230477 4.876622 5.201536
[17] 4.637366 4.143637 3.940643 4.300171Split acoustic data by animal_id, map summaries — the animal-centric pattern in movement ecology.
by_animal <- group_split(acoustic, animal_id) # list of tibbles
max_det <- map_dbl(by_animal, ~ max(.x$detections)) # peak ping count
names(max_det) <- map_chr(by_animal, ~ unique(.x$animal_id)) # name elements
max_det # per-tag maxA101 A102 A103
7 5 7 Split fishery by year and map total tonnes — functional annual report tables.
by_year <- group_split(fishery, year) # one df per year
year_tot <- map_dbl(by_year, ~ sum(.x$tonnes)) # annual totals
setNames(year_tot, map_int(by_year, ~ unique(.x$year))) # names = years 2018 2019 2020 2021 2022 2023
246.5 277.1 301.1 241.6 202.5 179.8 year_tot # numeric summary[1] 246.5 277.1 301.1 241.6 202.5 179.8map2() on paired vectors supports weighted calculations — tonnes times price, detections times effort.
x <- biolog$depth_m[1:10] # depths subsample
w <- rep(1, 10) # equal weights demo
weighted <- map2_dbl(x, w, ~ .x * .y) # element products
sum(weighted) / sum(w) # weighted mean depth[1] 4.941123possibly() replaces errors with a default value — softer than safely() when you want NA back.
mult <- c(1, 2, 0, 4) # includes zero divisor
safe_div <- possibly(function(a, b) a / b, otherwise = NA_real_)
map2_dbl(acoustic$detections[1:4], mult, safe_div) # NA where divide fails[1] 5 1 Inf 1quietly() captures messages while returning results — useful when mapped functions are chatty.
chatty_mean <- function(x) { # function with message
message("computing mean on length ", length(x))
mean(x)
}
q <- quietly(chatty_mean) # wrap to capture noise
res <- map(by_site, ~ q(.x$count)) # list of quiet results
map_dbl(res, "result") # extract means only East North South
10.95 8.80 9.80 furrr future_map() mirrors map() on parallel workers — worth it on heavy simulations, not tiny reef tables.
has_furrr <- requireNamespace("furrr", quietly = TRUE) # check package
if (has_furrr) {
suppressPackageStartupMessages({ library(furrr); library(future) })
plan(multisession, workers = 2) # local worker pool
future_map_dbl(by_site, ~ mean(.x$count)) # parallel map
} else {
message("Install furrr for parallel demo; using map_dbl")
map_dbl(by_site, ~ mean(.x$count)) # serial fallback
} East North South
10.95 8.80 9.80 The future plan() selects sequential vs multisession execution — set once before furrr maps.
time_serial <- system.time({ # time serial map
map_dbl(by_site, ~ mean(.x$count))
})
time_serial # print timing object user system elapsed
0.001 0.000 0.000 Map over burst splits to build hierarchical summaries — ‘deep’ mapping across nested structure.
burst_stats <- biolog |> # biologging data
group_split(burst_id) |> # list of burst tibbles
map(~ summarise(.x, mean_depth = mean(depth_m), n_sec = n()))
burst_tbl <- map_dfr(burst_stats, identity) # row-bind summaries
head(burst_tbl) # one row per burst# A tibble: 6 × 2
mean_depth n_sec
<dbl> <int>
1 4.87 20
2 4.43 20
3 4.76 20
4 5.57 20
5 5.26 20
6 4.98 20compose() builds function pipelines — conceptual kin to |> but for function values passed to map.
scale01 <- function(x) (x - min(x)) / (max(x) - min(x)) # min-max scale
round2 <- function(x) round(x, 2) # two decimals
fmt <- compose(round2, scale01) # combined function
fmt(fishery$tonnes) # formatted vector [1] 0.47 0.50 0.07 0.41 0.34 0.51 0.30 0.55 1.00 0.07 0.23 0.58 0.64 0.46 0.00
[16] 0.32 0.27 0.10 0.20 0.56 0.34 0.26 0.12 0.24Integrate split, map, and map_dfr() on acoustic data — the workflow to verify before batch-processing receiver exports with AI.
zone_list <- group_split(acoustic, reef_zone) # list by habitat zone
zone_summary <- map_dfr(zone_list, function(df) {
tibble(
reef_zone = unique(df$reef_zone),
total_det = sum(df$detections),
n_rows = nrow(df)
)
})
ggplot(zone_summary, aes(x = reef_zone, y = total_det, fill = reef_zone)) +
geom_col(show.legend = FALSE) + # bar chart by zone
labs(title = "Detections by reef zone", y = "Sum of detections")