A directed network graph visualizes relationships between entities using nodes connected by edges with arrows, indicating the direction of relationships or flow. Unlike undirected graphs, directed graphs reveal asymmetric relationships such as dependencies, hierarchies, or information flow. The arrows clearly communicate which entity points to which, making cause-and-effect relationships and directional dependencies immediately visible.
# anyplot.ai
# network-directed: Directed Network Graph
# Library: makie 0.21.9 | Julia 1.11.9
# Quality: 91/100 | Created: 2026-09-05
using CairoMakie
using Colors
using Random
Random.seed!(42)
# --- Theme tokens ------------------------------------------------------------
THEME = get(ENV, "ANYPLOT_THEME", "light")
PAGE_BG = THEME == "light" ? colorant"#FAF8F1" : colorant"#1A1A17"
INK = THEME == "light" ? colorant"#1A1A17" : colorant"#F0EFE8"
INK_SOFT = THEME == "light" ? colorant"#4A4A44" : colorant"#B8B7B0"
EDGE_COLOR = RGBAf(INK_SOFT.r, INK_SOFT.g, INK_SOFT.b, 0.55)
IMPRINT_PALETTE = [
colorant"#009E73", colorant"#C475FD", colorant"#4467A3", colorant"#BD8233",
colorant"#AE3030", colorant"#2ABCCD", colorant"#954477", colorant"#99B314",
]
# --- Data: a software package dependency graph --------------------------------
# Arrows point from a consumer to what it depends on / imports, exactly the
# "import direction" application called out in the specification.
nodes = [
"webapp", "cli",
"api-client", "auth", "renderer",
"http", "config", "crypto", "cache", "svg-utils",
"json",
"logging",
]
edges = [
("webapp", "api-client"), ("webapp", "auth"), ("webapp", "renderer"),
("cli", "api-client"), ("cli", "auth"), ("cli", "logging"),
("api-client", "http"), ("api-client", "config"),
("auth", "crypto"), ("auth", "config"), ("auth", "cache"),
("renderer", "svg-utils"), ("renderer", "config"),
("http", "logging"), ("crypto", "logging"), ("cache", "logging"),
("svg-utils", "json"), ("config", "logging"), ("json", "logging"),
]
# --- Hierarchical layout ------------------------------------------------------
# NetworkLayout.jl is not part of this catalog's Julia environment, so the
# layer assignment is computed directly: each node's layer is the length of
# the longest dependency chain reaching it, found by relaxing edges to a
# fixpoint (a tiny Bellman-Ford variant — the dependency graph is a DAG, so
# this always converges). Nodes with no incoming edges anchor layer 0.
layer = Dict(n => 0 for n in nodes)
changed = true
while changed
global changed = false
for (src, dst) in edges
if layer[dst] < layer[src] + 1
layer[dst] = layer[src] + 1
global changed = true
end
end
end
n_layers = maximum(values(layer)) + 1
layer_nodes = [String[] for _ in 1:n_layers]
for n in nodes
push!(layer_nodes[layer[n]+1], n)
end
indegree = Dict(n => 0 for n in nodes)
for (_, dst) in edges
indegree[dst] += 1
end
# Barycenter crossing-minimization: repeatedly reorder each layer by the mean
# position of its neighbors, alternating downward/upward sweeps (Sugiyama-style).
# This is what pulls "auth"/"renderer" and their fan-out into straighter columns
# instead of the crossing tangle the review flagged.
neighbors = Dict(n => String[] for n in nodes)
for (src, dst) in edges
push!(neighbors[src], dst)
push!(neighbors[dst], src)
end
order_x = Dict{String,Float64}(n => Float64(j) for ns in layer_nodes for (j, n) in enumerate(ns))
for iter in 1:6
layer_order = isodd(iter) ? (1:n_layers) : reverse(1:n_layers)
for li in layer_order
ns = layer_nodes[li]
length(ns) <= 1 && continue
bary = Dict(n => begin
xs = [order_x[m] for m in neighbors[n]]
isempty(xs) ? order_x[n] : sum(xs) / length(xs)
end for n in ns)
sort!(ns, by=n -> bary[n])
for (j, n) in enumerate(ns)
order_x[n] = Float64(j)
end
end
end
layer_spacing = 2.4
node_spacing = 2.2
pos = Dict{String,Point2f}()
for (li, ns) in enumerate(layer_nodes)
k = length(ns)
y = (n_layers - li) * layer_spacing
for (j, n) in enumerate(ns)
x = (j - (k + 1) / 2) * node_spacing
pos[n] = Point2f(x, y)
end
end
marker_size(n) = 26.0f0 + 6.0f0 * indegree[n]
node_radius(n) = 0.16 + 0.006 * marker_size(n) # data-space clearance so arrows stop at the node edge
tier_labels = ["Applications", "Services", "Infrastructure/utilities", "Data format", "Core"]
node_color(n) = IMPRINT_PALETTE[min(layer[n] + 1, length(IMPRINT_PALETTE))]
# --- Plot ----------------------------------------------------------------------
fig = Figure(
size=(1200, 1200),
fontsize=14,
backgroundcolor=PAGE_BG,
)
ax = Axis(
fig[1, 1];
title="network-directed · julia · makie · anyplot.ai",
titlesize=20,
titlecolor=INK,
backgroundcolor=PAGE_BG,
aspect=DataAspect(),
)
hidedecorations!(ax)
hidespines!(ax)
limits!(ax, -5.2, 5.2, -1.0, 10.2)
# Move `from` toward `to` by clearance `r` (data units) — keeps arrow shafts
# and heads from disappearing under the node markers they connect.
function pull_in(from::Point2f, to::Point2f, r)
d = to - from
u = d / hypot(d[1], d[2])
from + u * Float32(r)
end
function draw_arrow!(ax, src::String, dst::String; waypoint::Union{Point2f,Nothing}=nothing)
if waypoint === nothing
tail = pull_in(pos[src], pos[dst], node_radius(src))
else
tail = waypoint
start = pull_in(pos[src], waypoint, node_radius(src))
lines!(ax, [start, waypoint]; color=EDGE_COLOR, linewidth=2.0)
end
head = pull_in(pos[dst], tail, node_radius(dst))
arrows!(ax, [tail], [head - tail]; color=EDGE_COLOR, linewidth=2.0, arrowsize=15)
end
# The one long-range dependency (cli → logging) is routed around the middle
# tiers with a dog-leg instead of a straight line, so it doesn't cut through
# unrelated nodes — the "curved edges to avoid overlap" case from the spec.
skip_edge = ("cli", "logging")
waypoint = Point2f(5.0, 4.8)
for (src, dst) in edges
if (src, dst) == skip_edge
draw_arrow!(ax, src, dst; waypoint=waypoint)
else
draw_arrow!(ax, src, dst)
end
end
for n in nodes
scatter!(ax, [pos[n]]; color=node_color(n), markersize=marker_size(n), strokewidth=0)
end
# Labels sit below each node rather than inside it — several ids ("api-client",
# "svg-utils") are wider than even the largest marker and would get clipped.
for n in nodes
label_pos = pos[n] - Point2f(0, node_radius(n) + 0.16)
text!(ax, label_pos; text=n, color=INK, fontsize=13, align=(:center, :top))
end
legend_elements = [MarkerElement(color=IMPRINT_PALETTE[i], marker=:circle, markersize=14) for i in 1:n_layers]
Legend(fig[2, 1], legend_elements, tier_labels[1:n_layers];
orientation=:horizontal, framevisible=false, labelcolor=INK, nbanks=2)
rowgap!(fig.layout, 0)
rowsize!(fig.layout, 2, Auto(0.05))
# --- Save ------------------------------------------------------------------
save("plot-$(THEME).png", fig; px_per_unit=2)
Runnable source as JSON, for any HTTP client: https://api.anyplot.ai/specs/network-directed/makie/code. Any spec id and library id listed in llms-full.txt fit the same URL shape; every URL below is complete and callable.
{
"spec_id": "network-directed",
"language": "julia",
"library": "makie",
"page": "https://anyplot.ai/network-directed/julia/makie",
"hub": "https://anyplot.ai/network-directed",
"code_json": "https://api.anyplot.ai/specs/network-directed/makie/code",
"spec_json": "https://api.anyplot.ai/specs/network-directed",
"render_light_png": "https://storage.googleapis.com/anyplot-images/plots/network-directed/julia/makie/plot-light.png",
"render_dark_png": "https://storage.googleapis.com/anyplot-images/plots/network-directed/julia/makie/plot-dark.png",
"quality_score": 91.0,
"license": "MIT",
"guide": "https://anyplot.ai/llms.txt"
}Part of Directed Network Graph on anyplot.ai.