Compare commits

...
12 Commits
Author SHA1 Message Date
LeoVasanko 09ebc63690 Record response status per path; mark 404 trails red in the viewer
Document GETs now stash their status (200/404) in a pending table,
consumed by the matching ping: visits gain a per-path statuses map and
crawler hits a status field. Trail links with a 404 status render in
red with the status code in the tooltip, alongside the read time.
2026-08-24 19:43:23 +00:00
LeoVasanko 2868843028 Fix inverted client filter in admin hide ping
The hide=1 branch kept the admin's own pending crawler hits (==) instead
of discarding them (!=), so an admin's document GETs flushed to the
crawler list 10s later while every other client's pending hits were
wrongly dropped. This is why ordinary admin browsers showed up as
crawlers.
2026-08-24 19:40:12 +00:00
LeoVasanko 6fcebaea3f Transition map: svg-scaled fonts, border-clipped pill text, tight crop
- Fonts scale with the svg instead of the --u constant-screen-size
  compensation (ResizeObserver machinery removed)
- Larger node text (slug 19px, count 15px)
- Pill labels no longer ellipsis-truncated: text is clipped at the pill
  border via per-node clipPaths; captions center when they fit and anchor
  left on overflow so the title's beginning survives; count lines stay
  centered
- Bounding box crops to pill half extents plus the ribbon halo instead of
  the diagonal radius, removing the large top/bottom margins
2026-08-24 18:35:36 +00:00
LeoVasanko 4a48d08a19 Analytics layout: larger charts with natural-width cap, one-line totals
- Charts grow to a larger intrinsic size (1052x174) and never upscale
  past it; centered with equal side margins above the cap, full width
  below, svg always within page bounds; overflow visible so wider fonts
  don't clip at the viewBox edge
- Totals row aligns its left edge with the charts and shrinks (gap first,
  then font) via container units to always stay on one line
2026-08-24 18:21:58 +00:00
LeoVasanko e3e29251ca Transition map: cull invisible connectors, stable beads, lane labels
- Cull connections whose thin middle would render below ~0.8px
  (MIN_WMID); drop external source/exit nodes whose connectors are
  all culled, while site page nodes always stay
- Bead simulation persists across data reloads: emitters keyed per edge
  direction, beads tracked by progress, so unrelated count changes no
  longer reshuffle bead positions
- Bead speed relative to span length: constant 1.5s traversal per edge
- Top lane labeled with a house icon; all lane labels left-aligned just
  past the source pill (half height on near-vertical branch lanes), with
  guides running to the lane end so long slugs are never truncated
2026-08-24 17:58:02 +00:00
LeoVasanko 8affc41289 Tighter analytics chart chrome
- Unified chart text at 11px system-ui; fixed size independent of theme font
- Day view y axis reads "visits / 5 min" / "views / 5 min"
- Left margin and y tick spacing tightened (MARGIN_L 56 -> 40)
- Gap between visits and views charts removed
- Legend repositioned for the larger font
2026-08-24 17:26:30 +00:00
LeoVasanko 2de4717230 Fix week overlay alignment, in-plot ISO week legend, shorter charts
- weeklySeries shifts overlaid weeks onto the current week's time axis so
  they overlay inside the plot instead of overflowing left; oldest weeks
  paint first, current week on top
- Legend moved inside the visits chart's top right: current ISO week in
  accent, past weeks as a single muted "Week M" / "Week M–N" specimen
- Past week curves use the muted color instead of faded accent
- Chart height reduced ~30% (180 -> 126)
2026-08-24 17:03:49 +00:00
LeoVasanko 563e8fcaf2 Rework analytics chart scaling; self-contained SVG charts
- Charts render as single SVGs with axis labels inside the viewBox,
  replacing the stretched plot + HTML overlay labels
- Rolling ranges end at now, t0 aligned to UTC day; bucket size follows
  the window (6h up to 31 days) so "all" at its 30-day minimum renders
  identically to "month"
- X labels always centered on their true position; no edge-align shifting
- rangeWindow simplified to rolling spans ending at now
- TransitionGraph "all" visual scale floored at the 30-day plot minimum
2026-08-24 12:31:25 +00:00
LeoVasanko 921a5484a2 Fixed-sigma smoothing of traffic history plots. 2026-08-24 05:54:43 +00:00
LeoVasanko 0e2e52fa45 Use last 24h/7d/30d/365d/all analytics data. Previously some fields were unfiltered and weekly view was based on calendar weeks. 2026-08-24 05:42:32 +00:00
LeoVasanko 075848f782 Crawlers should include all sorts of spiders along with bots and googleother. 2026-08-24 05:28:05 +00:00
LeoVasanko 7b8899af92 Transition map: bounded node scaling, concentric branch lanes, exit row at bottom 2026-08-24 04:47:06 +00:00
11 changed files with 799 additions and 625 deletions
+56 -22
View File
@@ -147,6 +147,8 @@ Each `Visit` record:
does not append.
- `utm``utm_*` query parameters from the landing URL, as a dict.
- `read` — active reading time per path (seconds), keyed by path.
- `statuses` — HTTP status of the response when each path was first seen
(200 or 404), keyed by path.
Each `CrawlerHit` record:
@@ -154,7 +156,9 @@ Each `CrawlerHit` record:
- `entry` — page path requested,
- `client` — 6-byte blake3 hash referencing `Analytics.clients`,
- `referer` — external https origin of the request, `""` for direct/none,
- `query` — raw query string of the request.
- `query` — raw query string of the request,
- `status` — HTTP status of the served response (200 for a real page, 404
for a category placeholder or missing page).
Each `AbuseHit` record:
@@ -173,6 +177,10 @@ that triggered classification are lifted to the top, followed by other 404s
and then document GETs from the abuser. Within each category paths are
sorted by count descending, then by their earliest hit.
In the visitor and crawler tables, internal paths that returned a 404 status
are shown in red and the link title includes the status code, so it is easy
to tell misses from real pages at a glance.
## Aggregates
- `transitions`: time series of page transitions, sparse nested dict
@@ -216,12 +224,11 @@ spans less than 24 hours, otherwise `week`.
`AnalyticsView.vue` is no longer a full-screen overlay; the `body.analytics-open`
page-chrome hiding and `#/analytics/<range>` hash routing have been removed.
Charts are SVG curves (Catmull-Rom over an edge-aware adaptive Gaussian —
a change-point detector splits the series at traffic-level shifts, then
each segment is smoothed with a bandwidth that ramps with a broad pilot
estimate of the local rate: isolated events stay narrow (~0.4-unit sigma,
peaking at ~1 event/unit), busy traffic widens to a 1-unit sigma. The raw
series is drawn faint underneath). Values are
Charts are SVG curves (Catmull-Rom over an edge-aware Gaussian — a
change-point detector splits the series at traffic-level shifts, then each
segment is smoothed independently with a fixed sigma chosen so N events in
a single bucket peak at N events per unit. The raw series is drawn faint
underneath). Values are
**per-unit rates** — per hour on the week view (5-minute bucket counts × 12,
plotted at native 5-minute resolution), per day on the month+ ranges — and
the smoothing time scale follows the unit: the month+ sigmas are 24× the
@@ -232,31 +239,58 @@ labeled intervals, minor lines at fifths when integral; the minimum y-axis
range is 10 so tiny values such as a single visit are not stretched to a
fractional scale).
The week range is aligned to Monday 00:00 UTC and overlays up to 8 previous
weeks in the same accent color at decreasing opacity (the current week is
weeks in the muted color at decreasing opacity (the current week keeps the
accent color and is
truncated at the current bucket, never drawing fake zeroes for the future);
its x labels are weekday names centered at midday UTC, without vertical grid
a compact legend inside the top right of the visits chart marks the current
ISO week in accent and the overlaid past weeks as "Week M" or "Week MN" on
a muted specimen. Its x labels are weekday names centered at midday UTC, without
vertical grid
lines (day boundaries would be misleading in the viewer's timezone). The
month view labels days the same lineless way — day numbers at noon UTC,
with the month name substituted for the 1st. Year is a rolling 365-day window ending at now, re-bucketed to daily points,
with boundary lines at months/years. All uses the full data reach, but keeps
at least the past 30 days so the chart never collapses to a tiny sliver when
the site is young. Below the charts: a radial **transition map** (all pages from
`/_api/pages` — front page at the center, each slug level on its own ring,
siblings clockwise in navigation order from the top, radial gap equal to
the arc spacing — opposite transition directions joined into organic
with the month name substituted for the 1st. Month, year and all are
rolling windows ending at now, aligned to UTC day boundaries at the start
so the labels span the whole range; the bucket size follows the window —
6 hours up to 31 days, daily beyond — with boundary lines at months/years
on the longer ranges. All uses the full data reach, but keeps
at least the past 30 days (identical to the month view when the site is
younger than that, bucket size included) so the chart never collapses to a
tiny sliver when the site is young. Below the charts: a **transition map** (all pages from
`/_api/pages` — top-level menu items on a large-radius circular arc whose
bottom point is the last item (each earlier item a bit higher), connected
by a top lane labeled 🏠︎ beside the home pill (50% thicker than
the branch lanes, its label font and guide offset scaled along), each item's
subtree fanning out below it in menu order along a large-radius circular
arc that leaves heading
straight down and gradually bends right, index pages without views omitted
and their children promoted in their place. The submenu structure is drawn
as wide branch lanes: one per path prefix with at least two visible
nodes, running behind the branch's node pills as circle arcs concentric
with the fan (parent levels one radius step outward, so all lanes of a
group share exactly one form), each labeled with its branch slug
left-aligned just past the first pill and allowed to run along the lane to
its end, disappearing under later pills when long — so the lanes reflect
the path
structure even where index pages are omitted — opposite transition
directions joined into organic
tapered connections whose middle width grows logarithmically with the
count (a single count renders as a ~1 px line, uncapped), connections
count (uncapped), connections
carrying less than 1% of the total traffic
pruned; beads are simulated one by one in JS (requestAnimationFrame) and
flow along each edge, emitted at a rate linearly proportional
pruned, as are those whose thin middle would render below ~0.8 px —
fainter strands are invisible and only their wide end flares would show; beads are simulated one by one in JS (requestAnimationFrame) and
flow along each edge, persisting across data reloads (emitters are keyed
per edge direction and beads tracked by progress, so an unrelated count
change never reshuffles them), emitted at a rate linearly proportional
to the directional count with no in-flight limit, opposing directions
offset onto parallel lanes. External sources show as a node row above the
offset onto parallel lanes. External sources and exits whose connectors are
all culled by the width threshold are dropped from their rows themselves
(the site's own page nodes always stay, connected or not). External sources show as a node row above the
map: each visit is attributed to `utm_campaign`, then `utm_source`, then the
referer origin, then any other `utm_*` tag, so UTM-tagged visits are grouped
under their campaign/source value rather than the referer domain. A UTM
source node only links to its referer when every visit carrying that tag
came from the same origin. External exits are small nodes fanned outwards
from their source page), per-page view
came from the same origin. External exits are full-size nodes in a matching
row centered below the map, so the site itself stays in the middle), per-page view
counts, the top transitions and the 50 most recent visit trails. Data is
streamed live over `WebSocket /_api/ws/analytics`, which pushes the latest
JSON snapshot on connect and again whenever the analytics file is updated
+49 -11
View File
@@ -5,7 +5,13 @@
// visit/crawler tables. Read-only.
// See docs/analytics.md for the data format.
import { computed, onMounted, onUnmounted, ref, watch } from 'vue'
import { RANGES } from './analytics/time.js'
import {
RANGES,
rangeWindow,
filterRecordsByRange,
filterTransitionsByRange,
filterViewsByRange,
} from './analytics/time.js'
import {
calcReadStats,
calcTotalViews,
@@ -20,6 +26,11 @@ import TrailLink from './TrailLink.vue'
import VisitorCell from './VisitorCell.vue'
import TransitionGraph from './TransitionGraph.vue'
import VisitorCharts from './VisitorCharts.vue'
import { VIEW_W } from './analytics/chart.js'
// Same centering margin as the charts, so the totals row's left edge
// aligns with the chart svg above the natural width.
const CHART_MARGIN = `max(0px, calc(50% - ${VIEW_W / 2}px))`
const ABUSE_MAX_LINES = 5
@@ -90,8 +101,25 @@ onUnmounted(() => {
}
})
const visits = computed(() => data.value?.visits || [])
const totalViews = computed(() => calcTotalViews(data.value?.views))
const window = computed(() => rangeWindow(range.value))
// All non-chart stats follow the selected range; the charts keep their own
// range-specific x windows (week overlays previous weeks aligned to Monday).
const rangeData = computed(() => {
if (!data.value) return null
const { t0, t1 } = window.value
return {
...data.value,
transitions: filterTransitionsByRange(data.value.transitions, t0, t1),
views: filterViewsByRange(data.value.views, t0, t1),
visits: filterRecordsByRange(data.value.visits, t0, t1),
crawlers: filterRecordsByRange(data.value.crawlers, t0, t1),
abuse: filterRecordsByRange(data.value.abuse, t0, t1),
}
})
const visits = computed(() => rangeData.value?.visits || [])
const totalViews = computed(() => calcTotalViews(rangeData.value?.views))
const readStats = computed(() => calcReadStats(visits.value))
// Keep the URL shareable when the range changes.
@@ -103,9 +131,9 @@ watch(range, (r) => {
const clients = computed(() => data.value?.clients || {})
const visitRows = computed(() => formatVisitRows(visits.value, clients.value, pageTree.value, now.value))
const crawlers = computed(() => data.value?.crawlers || [])
const crawlers = computed(() => rangeData.value?.crawlers || [])
const crawlerRows = computed(() => formatCrawlerRows(crawlers.value, clients.value, pageTree.value, now.value))
const abuseRows = computed(() => formatAbuseRows(data.value?.abuse || [], clients.value, now.value))
const abuseRows = computed(() => formatAbuseRows(rangeData.value?.abuse || [], clients.value, now.value))
</script>
@@ -125,15 +153,15 @@ const abuseRows = computed(() => formatAbuseRows(data.value?.abuse || [], client
<p v-if="error" class="error"> {{ error }}</p>
<p v-else-if="!data" class="loading">loading</p>
<template v-else>
<section class="totals">
<section class="totals" :style="{ marginLeft: CHART_MARGIN }">
<div><strong :title="String(visits.length)">{{ formatCount(visits.length) }}</strong> visits</div>
<div><strong :title="String(totalViews)">{{ formatCount(totalViews) }}</strong> page views</div>
<div><strong>{{ readStats.avgMinPerVisit }}</strong> min/visit</div>
<div><strong>{{ readStats.avgArticleMedianMin }}</strong> min article read</div>
<div><strong>{{ readStats.avgArticleMedianMin }}</strong> min/read</div>
</section>
<VisitorCharts :data="data" :range="range" />
<TransitionGraph :data="data" :range="range" :page-tree="pageTree" />
<TransitionGraph :data="rangeData" :window="window" :page-tree="pageTree" />
<section>
<h2>Recent visits</h2>
@@ -265,6 +293,8 @@ const abuseRows = computed(() => formatAbuseRows(data.value?.abuse || [], client
width: 100%;
/* Same 1.25rem side spacing as main's article padding. */
padding: 1.5rem 1.25rem 4rem;
/* Container for cqw-based shrink-to-fit (see .totals). */
container-type: inline-size;
}
.analytics-panel header {
@@ -331,12 +361,15 @@ const abuseRows = computed(() => formatAbuseRows(data.value?.abuse || [], client
.analytics-view :deep(.muted) { color: var(--muted); }
.analytics-view :deep(.small) { font-size: 0.75em; }
/* One line at any width: the gap shrinks first, then the font (the number
scales along in em), both following the panel's container width. */
.totals {
display: flex;
gap: 2rem;
font-size: 1.1rem;
gap: clamp(0.5rem, 3cqw, 2rem);
font-size: clamp(0.6rem, 2.2cqw, 1.1rem);
white-space: nowrap;
}
.totals strong { font-size: 1.5rem; }
.totals strong { font-size: 1.36em; }
.visit-table-wrap {
overflow-x: auto;
@@ -392,6 +425,11 @@ const abuseRows = computed(() => formatAbuseRows(data.value?.abuse || [], client
margin-left: 0.5rem;
}
.analytics-view :deep(.trail-link.error),
.analytics-view :deep(.trail-link.error:hover) {
color: var(--error, #c00);
}
.visit-table .utm-tag {
display: inline-block;
max-width: 100%;
+18 -3
View File
@@ -1,18 +1,33 @@
<script setup>
import { formatCount } from './analytics/format.js'
import { computed } from 'vue'
import { formatCount, formatReadTime } from './analytics/format.js'
defineProps({
const props = defineProps({
step: { type: Object, required: true },
count: { type: Number, default: 0 },
})
defineEmits(['close'])
const hasError = computed(() => props.step.status >= 400)
const title = computed(() => {
const parts = [props.step.title]
if (props.step.readSeconds > 0) {
parts.push(formatReadTime(props.step.readSeconds))
}
if (hasError.value) {
parts.push(`${props.step.status}`)
}
return parts.filter(Boolean).join(' — ')
})
</script>
<template>
<a class="trail-link"
:class="{ error: hasError }"
:href="step.path"
:title="count > 1 ? `${step.title} (${count} hits)` : step.title"
:title="title"
:target="step.external ? '_blank' : undefined"
:rel="step.external ? 'noopener' : undefined"
@click="(e) => { if (!step.external) $emit('close') }">
+152 -128
View File
@@ -1,165 +1,168 @@
<script setup>
/**
* Radial transition map filtered to the selected time range.
* Radial transition map for a pre-filtered time range.
*
* Transitions are stored per 5-minute bucket (from -> to -> bucket ->
* count), so the graph sums the buckets falling inside the selected
* range, exactly like the charts and per-page views do.
* The parent filters transitions, views and visits to the selected range
* before passing them in; `window` carries the absolute [t0, t1) window
* so the visual scale can normalize against a one-week reference.
*/
import { computed, onBeforeUnmount, onMounted, ref, shallowRef, watch } from 'vue'
import { rangeWindow, WEEK } from './analytics/time.js'
import { computed, onBeforeUnmount, onMounted, shallowRef, watch } from 'vue'
import { DAY, WEEK } from './analytics/time.js'
import { formatCount } from './analytics/format.js'
import {
TNODE_W,
TNODE_H,
BEAD_R,
BEAD_SPEED,
buildTransitionGraph,
filterTransitionsByRange,
filterViewsByRange,
filterVisitsByRange,
} from './analytics/transitions.js'
const props = defineProps({
data: { type: Object, default: null },
range: { type: String, required: true },
window: { type: Object, required: true },
pageTree: { type: Array, default: null },
})
const window = computed(() => rangeWindow(props.range))
const visualScale = computed(() => {
const { t0, t1 } = window.value
const { t0, t1 } = props.window
if (t0 != null && t1 != null) return WEEK / (t1 - t0)
// 'all': scale by the actual data span.
// 'all': scale by the actual data span, but never less than the 30-day
// minimum the plot enforces, so sparse young data is not over-amplified.
const times = new Set()
for (const buckets of Object.values(props.data?.views || {})) {
for (const k of Object.keys(buckets)) times.add(Date.parse(k))
}
const arr = [...times]
if (arr.length < 2) return 1
return WEEK / (Math.max(...arr) - Math.min(...arr))
const span = Math.max(...arr) - Math.min(...arr)
return WEEK / Math.max(span, 30 * DAY)
})
const filteredData = computed(() => {
if (!props.data) return null
const { t0, t1 } = window.value
return {
transitions: filterTransitionsByRange(props.data.transitions, t0, t1),
views: filterViewsByRange(props.data.views, t0, t1),
}
})
const filteredVisits = computed(() =>
filterVisitsByRange(props.data?.visits, window.value.t0, window.value.t1),
)
const graph = computed(() =>
filteredData.value
? buildTransitionGraph(filteredData.value, props.pageTree, filteredVisits.value, visualScale.value)
props.data
? buildTransitionGraph(props.data, props.pageTree, props.data.visits || [], visualScale.value)
: null,
)
// Bead animation: every bead is simulated independently in JS. Each flow
// (one per edge direction) emits a bead every `interval` seconds; beads
// travel at BEAD_SPEED along the segment and are dropped at the end.
// cross their segment in a constant TRAVERSAL_S seconds (speed relative
// to span length) and are dropped at the end.
// There is deliberately no cap on beads in flight.
// Emitters persist across data reloads, keyed by flow.key: an unchanged
// link keeps its emission phase and in-flight beads (tracked by progress,
// not absolute time), so a count change elsewhere never reshuffles them.
const beads = shallowRef([])
let rafId = 0
const emitters = new Map() // flow.key -> { flow, interval, next, alive }
const live = [] // { e, p } — beads in flight, p = progress 0..1
let lastTick = 0
const MAX_BEAD_RATE = 120 // upper bound on total beads per second
const TRAVERSAL_S = 1.5 // seconds to cross any segment, end to end
const startBeads = (flows) => {
cancelAnimationFrame(rafId)
beads.value = []
if (!flows?.length) return
if (matchMedia('(prefers-reduced-motion: reduce)').matches) return
const syncBeads = (flows) => {
const reduced = matchMedia('(prefers-reduced-motion: reduce)').matches
if (!flows?.length || reduced) {
emitters.clear()
live.length = 0
beads.value = []
return
}
// Cap the total bead emission rate so a busy range cannot spawn enough
// beads to kill the page. Existing per-range time scaling is preserved;
// this is only a proportional emergency throttle when the limit is hit.
const totalRate = flows.reduce((s, f) => s + 1 / f.interval, 0)
const scale = totalRate > MAX_BEAD_RATE ? MAX_BEAD_RATE / totalRate : 1
const live = [] // { flow, t0 } — one entry per bead in flight
const now = performance.now()
const emitters = flows.map((flow) => {
const seen = new Set()
for (const flow of flows) {
seen.add(flow.key)
const interval = (flow.interval / scale) * 1000
// Pre-fill the traversal with evenly spaced beads (random phase), so
// the flow appears already running instead of starting empty.
const e = emitters.get(flow.key)
if (e) {
e.flow = flow // pick up new geometry/rate, keep the phase
e.interval = interval
continue
}
// New emitter: pre-fill the traversal with evenly spaced beads (random
// phase), so the flow appears already running instead of empty.
const phase = Math.random() * interval
for (let t = now - (flow.len / BEAD_SPEED) * 1000 + phase; t <= now; t += interval) {
live.push({ flow, t0: t })
const dp = interval / 1000 / TRAVERSAL_S
const ne = { flow, interval, next: now + phase, alive: true }
for (let p = 1 - phase / 1000 / TRAVERSAL_S; p > 0; p -= dp) {
live.push({ e: ne, p })
}
return { flow, interval, next: now + phase }
})
const tick = (t) => {
for (const e of emitters) {
while (e.next <= t) {
live.push({ flow: e.flow, t0: e.next })
e.next += e.interval
}
}
const out = []
for (let i = live.length - 1; i >= 0; i--) {
const b = live[i]
const p = ((t - b.t0) / 1000) * BEAD_SPEED / b.flow.len
if (p >= 1) {
live.splice(i, 1)
continue
}
out.push({
x: b.flow.x1 + (b.flow.x2 - b.flow.x1) * p,
y: b.flow.y1 + (b.flow.y2 - b.flow.y1) * p,
})
}
beads.value = out
rafId = requestAnimationFrame(tick)
emitters.set(flow.key, ne)
}
for (const [key, e] of emitters) {
if (!seen.has(key)) {
e.alive = false
emitters.delete(key)
}
}
for (let i = live.length - 1; i >= 0; i--) {
if (!live[i].e.alive) live.splice(i, 1)
}
}
const tick = (t) => {
const dt = lastTick ? (t - lastTick) / 1000 : 0
lastTick = t
for (const e of emitters.values()) {
while (e.next <= t) {
live.push({ e, p: 0 })
e.next += e.interval
}
}
const out = []
for (let i = live.length - 1; i >= 0; i--) {
const b = live[i]
b.p += dt / TRAVERSAL_S
if (b.p >= 1) {
live.splice(i, 1)
continue
}
const f = b.e.flow
out.push({ x: f.x1 + (f.x2 - f.x1) * b.p, y: f.y1 + (f.y2 - f.y1) * b.p })
}
beads.value = out
rafId = requestAnimationFrame(tick)
}
watch(() => graph.value?.flows, startBeads, { immediate: true })
watch(() => graph.value?.flows, syncBeads, { immediate: true })
onMounted(() => {
if (!matchMedia('(prefers-reduced-motion: reduce)').matches) {
rafId = requestAnimationFrame(tick)
}
})
onBeforeUnmount(() => cancelAnimationFrame(rafId))
// Text in the graph must render at a constant screen size regardless of
// how far the enlarged graph's viewBox is scaled down to fit the panel:
// measure the unit→pixel ratio and expose it as --u on the svg, which the
// font-size rules divide by. Falls back to 1 (raw units) until measured.
const svgEl = ref(null)
const pxPerUnit = ref(1)
let resizeObs = null
// The svg never renders larger than its natural size (1 viewBox unit = 1
// px, max-width below): the layout geometry is designed in pixel-like
// units, and upscaling would blow up the pills around their text. Narrow
// panels scale the graph down to fit (width: 100%), text along with it.
function updateScale() {
const el = svgEl.value
if (el && el.viewBox.baseVal.width) {
pxPerUnit.value = el.clientWidth / el.viewBox.baseVal.width
}
}
// Pill text is not truncated: text is clipped at the pill's rounded border
// (clipPath per node, inset a few units for padding). Captions center when
// they fit; overlong ones anchor left so their beginning (not their
// middle) survives the clip. Width estimate: ~0.52 em per glyph.
const fitsPill = (label, fontPx = 19) => label.length * 0.52 * fontPx <= TNODE_W - 16
onMounted(() => {
resizeObs = new ResizeObserver(updateScale)
})
watch(svgEl, (el) => {
resizeObs?.disconnect()
if (el) resizeObs?.observe(el)
})
watch(() => graph.value?.bounds, updateScale)
onBeforeUnmount(() => resizeObs?.disconnect())
// Font size (px) that fits a label inside the pill width at the current
// zoom: ~0.52 em average glyph width, 12 px padding per side, capped.
const fitPx = (label) =>
Math.min(15, (TNODE_W * pxPerUnit.value - 24) / (0.52 * Math.max(label.length, 1)))
const countLabel = (n) =>
n.readMin ? `${formatCount(n.views)}×${n.readMin}m` : formatCount(n.views)
</script>
<template>
<section v-if="graph">
<svg ref="svgEl" class="tmap" :style="{ '--u': pxPerUnit }" :viewBox="`${graph.bounds.x0} ${graph.bounds.y0} ${graph.bounds.x1 - graph.bounds.x0} ${graph.bounds.y1 - graph.bounds.y0}`"
<svg class="tmap" :style="{ maxWidth: `${graph.bounds.x1 - graph.bounds.x0}px` }" :viewBox="`${graph.bounds.x0} ${graph.bounds.y0} ${graph.bounds.x1 - graph.bounds.x0} ${graph.bounds.y1 - graph.bounds.y0}`"
role="img" aria-label="map of transitions between pages">
<path v-for="(a, i) in graph.arcs" :key="'a' + i"
:d="a.d" class="tarc" />
:id="`tarc${i}`" :d="a.d" :class="['tarc', a.top && 'tarc-top']" />
<template v-for="(a, i) in graph.arcs" :key="'t' + i">
<path v-if="a.ld" :id="`tarcl${i}`" :d="a.ld" fill="none" stroke="none" />
<text v-if="a.ld" class="tarclabel" :class="{ 'tarclabel-top': a.top }"><textPath :href="`#tarcl${i}`" startOffset="0">{{ a.label }}</textPath></text>
</template>
<path v-for="(e, i) in graph.edges" :key="'e' + i"
:d="e.d" :class="['tconn', e.external && 'tconn-exit']">
<title>{{ e.title }}</title>
@@ -167,35 +170,44 @@ const fitPx = (label) =>
<circle v-for="(b, i) in beads" :key="'b' + i"
:cx="b.x" :cy="b.y" :r="BEAD_R" class="tbead" />
<g v-for="(x, i) in graph.extNodes" :key="'x' + i">
<clipPath :id="`xclip${i}`">
<rect :x="x.x - TNODE_W/2 + 6" :y="x.y - TNODE_H/2" :width="TNODE_W - 12"
:height="TNODE_H" :rx="TNODE_H/2 - 4" />
</clipPath>
<a v-if="x.href" :href="x.href" target="_blank" rel="noopener">
<title>{{ x.path }}</title>
<rect :x="x.x - TNODE_W/2" :y="x.y - TNODE_H/2" :width="TNODE_W" :height="TNODE_H" :rx="TNODE_H/2"
:class="['txnode', x.kind === 'source' ? 'txnode-source' : 'txnode-exit']" />
<text :x="x.x" :y="x.y - TNODE_H*0.16" class="tnodeslug" dominant-baseline="middle"
:style="{ '--slug-px': `${fitPx(x.label)}px` }">{{ x.label }}</text>
<text :x="x.x" :y="x.y + TNODE_H*0.24" class="tnodecount" dominant-baseline="middle">{{ formatCount(x.count) }}</text>
<g :clip-path="`url(#xclip${i})`">
<text :x="fitsPill(x.label) ? x.x : x.x - TNODE_W/2 + 8" :y="x.y - TNODE_H*0.16" class="tnodeslug" dominant-baseline="middle" :style="{ textAnchor: fitsPill(x.label) ? 'middle' : 'start' }">{{ x.label }}</text>
<text :x="x.x" :y="x.y + TNODE_H*0.24" class="tnodecount" dominant-baseline="middle">{{ formatCount(x.count) }}</text>
</g>
</a>
<g v-else>
<title>{{ x.path }}</title>
<rect :x="x.x - TNODE_W/2" :y="x.y - TNODE_H/2" :width="TNODE_W" :height="TNODE_H" :rx="TNODE_H/2"
:class="['txnode', x.kind === 'source' ? 'txnode-source' : 'txnode-exit']" />
<text :x="x.x" :y="x.y - TNODE_H*0.16" class="tnodeslug" dominant-baseline="middle"
:style="{ '--slug-px': `${fitPx(x.label)}px` }">{{ x.label }}</text>
<text :x="x.x" :y="x.y + TNODE_H*0.24" class="tnodecount" dominant-baseline="middle">{{ formatCount(x.count) }}</text>
<g :clip-path="`url(#xclip${i})`">
<text :x="fitsPill(x.label) ? x.x : x.x - TNODE_W/2 + 8" :y="x.y - TNODE_H*0.16" class="tnodeslug" dominant-baseline="middle" :style="{ textAnchor: fitsPill(x.label) ? 'middle' : 'start' }">{{ x.label }}</text>
<text :x="x.x" :y="x.y + TNODE_H*0.24" class="tnodecount" dominant-baseline="middle">{{ formatCount(x.count) }}</text>
</g>
</g>
</g>
<g v-for="n in graph.nodes" :key="n.path">
<g v-for="(n, i) in graph.nodes" :key="n.path">
<clipPath :id="`nclip${i}`">
<rect :x="n.x - TNODE_W/2 + 6" :y="n.y - TNODE_H/2" :width="TNODE_W - 12"
:height="TNODE_H" :rx="TNODE_H/2 - 4" />
</clipPath>
<a :href="n.path">
<title>{{ n.title }}</title>
<rect :x="n.x - TNODE_W/2" :y="n.y - TNODE_H/2" :width="TNODE_W" :height="TNODE_H" :rx="TNODE_H/2" class="tnode" />
<text :x="n.x" :y="n.y - TNODE_H*0.16" class="tnodeslug" dominant-baseline="middle"
:style="{ '--slug-px': `${fitPx(n.label)}px` }">{{ n.label }}</text>
<text :x="n.x" :y="n.y + TNODE_H*0.24" class="tnodecount" dominant-baseline="middle">
{{ n.readMin ? `${formatCount(n.views)}×${n.readMin}m` : formatCount(n.views) }}
</text>
<g :clip-path="`url(#nclip${i})`">
<text :x="fitsPill(n.label) ? n.x : n.x - TNODE_W/2 + 8" :y="n.y - TNODE_H*0.16" class="tnodeslug" dominant-baseline="middle" :style="{ textAnchor: fitsPill(n.label) ? 'middle' : 'start' }">{{ n.label }}</text>
<text :x="n.x" :y="n.y + TNODE_H*0.24" class="tnodecount" dominant-baseline="middle">
{{ countLabel(n) }}
</text>
</g>
</a>
<text v-if="n.crumb" :x="n.x" :y="n.y - TNODE_H/2 - 10"
:class="['tnodepath', n.path === '/' && 'tnodepath-home']">{{ n.crumb }}</text>
</g>
</svg>
</section>
@@ -206,7 +218,9 @@ const fitPx = (label) =>
.tmap {
display: block;
width: 100%;
max-width: 100%;
/* max-width is set inline to the natural content width (px = viewBox
units), so wide panels never upscale the graph beyond 1:1. */
margin: 0 auto;
}
.tmap .tconn {
fill: var(--accent);
@@ -226,36 +240,46 @@ const fitPx = (label) =>
}
.tmap .txnode-source { fill: var(--text); }
.tmap .txnode-exit { fill: var(--text); }
/* Branch lanes: one wide concentric arc per path prefix, running behind
the node pills around the fan's circle center; parent levels sit one
indent (radius step) outward. Each lane's label follows a short guide
arc across the first inter-node gap (the part pills never cover). */
.tmap .tarc {
fill: none;
stroke: var(--line);
stroke-width: 1;
stroke: var(--muted);
stroke-width: 16;
opacity: 0.25;
}
.tmap .tarc-top { stroke-width: 24; }
/* Lane labels are left-aligned: each guide arc starts just past the source
pill's edge, the earliest point where the text is visible. */
.tmap .tarclabel {
fill: var(--muted);
font-size: 13px;
text-anchor: start;
}
/* The top lane is 50% thicker; its 🏠︎ label scales along. */
.tmap .tarclabel-top {
font-size: 19.5px;
}
.tmap .tnode {
fill: var(--accent);
stroke: none;
}
/* Text renders at a constant screen size: --u (set from JS) is the
viewBox-unit → pixel ratio of the rendered svg, so dividing by it makes
the sizes independent of how far the graph is scaled down. */
/* Text sizes are viewBox units: they shrink along with the graph on
narrow panels. Overlong labels are clipped at the pill border. */
.tmap .tnodeslug {
fill: var(--bg, Canvas);
font-size: calc(var(--slug-px, 15px) / var(--u, 1));
text-anchor: middle;
font-size: 19px;
text-anchor: start;
}
.tmap a { cursor: pointer; }
.tmap .tnodecount {
fill: var(--bg, Canvas);
opacity: 0.75;
font-size: calc(13px / var(--u, 1));
font-size: 15px;
text-anchor: middle;
}
.tmap .tnodepath {
fill: var(--muted);
font-size: calc(11px / var(--u, 1));
text-anchor: middle;
}
.tmap .tnodepath-home { font-size: calc(17px / var(--u, 1)); }
section { margin-top: 1.8rem; }
</style>
+93 -107
View File
@@ -4,10 +4,23 @@
*/
import { computed, onMounted, onUnmounted, ref } from 'vue'
import { makeSeries } from './analytics/time.js'
import { CHART_H, CHART_W, buildChart, fmtY } from './analytics/chart.js'
import {
CHART_H,
CHART_W,
MARGIN_B,
MARGIN_L,
VIEW_H,
VIEW_W,
buildChart,
} from './analytics/chart.js'
const DAY_REFRESH_MS = 15000
// Keep the whole svg within page bounds: full width below the natural
// size, centered with equal side margins above it (max() clamps the
// centering margin to 0 at the breakpoint, so the rule is continuous).
const CHART_MARGIN = `max(0px, calc(50% - ${VIEW_W / 2}px))`
const props = defineProps({
data: { type: Object, default: null },
range: { type: String, required: true },
@@ -29,6 +42,17 @@ function freqLabel(unit) {
return unit === '5min' ? '5 min' : unit === 'hour' ? 'hourly' : 'daily'
}
/** Vertical axis caption: "visits / 5 min" on the day view, else "hourly visits" style. */
function axisLabel(unit, ylabel) {
return unit === '5min' ? `${ylabel} / 5 min` : `${freqLabel(unit)} ${ylabel}`
}
/** Legend label for the overlaid past weeks: "Week M" or "Week MN". */
function pastLabel(series) {
const oldest = series.at(-1).label.slice(5) // strip "Week "
return series.length > 2 ? `Week ${oldest}${series[1].label.slice(5)}` : `Week ${oldest}`
}
const now = ref(Date.now())
let refreshInterval = null
onMounted(() => {
@@ -44,107 +68,87 @@ const viewChart = computed(() => buildChart(viewSeries.value, now.value))
<template>
<section v-for="c in [
{ ylabel: 'visits', chart: visitChart, empty: 'no visits recorded yet' },
{ ylabel: 'views', chart: viewChart, empty: 'no views recorded yet' },
{ ylabel: 'visits', chart: visitChart, legend: true, empty: 'no visits recorded yet' },
{ ylabel: 'views', chart: viewChart, legend: false, empty: 'no views recorded yet' },
]" :key="c.ylabel">
<template v-if="c.chart">
<div class="chartwrap">
<div class="plot">
<div class="plotarea">
<span class="yaxis-label">{{ freqLabel(c.chart.unit) }} {{ c.ylabel }}</span>
<svg class="chart" :viewBox="`0 0 ${CHART_W} ${CHART_H}`"
preserveAspectRatio="none" role="img" :aria-label="`${freqLabel(c.chart.unit)} ${c.ylabel}`">
<line v-for="g in c.chart.majors.slice(1)" :key="'j' + g.value"
:x1="0" :x2="CHART_W" :y1="g.y" :y2="g.y" class="major" />
<template v-for="t in c.chart.xticks" :key="'t' + t.x">
<line v-if="t.line" :x1="t.x" :x2="t.x" :y1="0" :y2="CHART_H"
class="minor vertical" />
</template>
<template v-if="c.chart.bars">
<rect v-for="(b, i) in c.chart.bars" :key="'b' + i"
:x="b.x" :y="b.y" :width="b.width" :height="b.height" class="bar" />
<path :d="c.chart.skyline" class="line" />
</template>
<template v-else>
<template v-for="(s, i) in c.chart.series" :key="i">
<path v-if="s.area" :d="s.area" class="area" />
<path :d="s.line" class="line" :style="{ opacity: s.opacity }" />
</template>
</template>
<line :x1="0" :x2="CHART_W" :y1="CHART_H - 0.5" :y2="CHART_H - 0.5"
class="axis" />
</svg>
<span v-for="g in c.chart.majors" :key="g.value" class="ylab"
:style="{ bottom: g.bottom + '%' }">{{ fmtY(g.value) }}</span>
</div>
<div class="xlabels">
<span v-for="t in c.chart.xticks" :key="t.x" class="xlab"
:style="{ left: t.left + '%' }">{{ t.label }}</span>
</div>
</div>
</div>
<div v-if="c.chart.series && c.chart.series.length > 1" class="legend">
<span v-for="(s, i) in c.chart.series" :key="i" :style="{ opacity: s.opacity }">
● {{ s.label }}
</span>
</div>
<svg class="chart" :viewBox="`${-MARGIN_L} 0 ${VIEW_W} ${VIEW_H}`"
:style="{ maxWidth: `${VIEW_W}px`, marginLeft: CHART_MARGIN }"
role="img" :aria-label="axisLabel(c.chart.unit, c.ylabel)">
<line v-for="g in c.chart.majors.slice(1)" :key="'j' + g.value"
:x1="0" :x2="CHART_W" :y1="g.y" :y2="g.y" class="major" />
<template v-for="t in c.chart.xticks" :key="'t' + t.x">
<line v-if="t.line" :x1="t.x" :x2="t.x" :y1="0" :y2="CHART_H"
class="minor vertical" />
</template>
<template v-if="c.chart.bars">
<rect v-for="(b, i) in c.chart.bars" :key="'b' + i"
:x="b.x" :y="b.y" :width="b.width" :height="b.height" class="bar" />
<path :d="c.chart.skyline" class="line" />
</template>
<template v-else>
<!-- Oldest overlay weeks first so the current week paints on top. -->
<template v-for="(s, i) in [...c.chart.series].reverse()" :key="i">
<path v-if="s.area" :d="s.area" class="area" />
<path :d="s.line" class="line" :class="{ past: s.past }"
:style="{ opacity: s.opacity }" />
</template>
</template>
<line :x1="0" :x2="CHART_W" :y1="CHART_H - 0.5" :y2="CHART_H - 0.5"
class="axis" />
<text v-for="g in c.chart.majors" :key="'y' + g.value" x="-5" :y="g.y"
text-anchor="end" dominant-baseline="middle" class="ylab">{{ g.label }}</text>
<text :x="-(MARGIN_L - 10)" :y="CHART_H / 2" text-anchor="middle"
:transform="`rotate(-90 ${-(MARGIN_L - 10)} ${CHART_H / 2})`"
class="yaxis-label">{{ axisLabel(c.chart.unit, c.ylabel) }}</text>
<text v-for="t in c.chart.xticks" :key="'x' + t.x" :x="t.x" :y="CHART_H + MARGIN_B - 8"
text-anchor="middle" class="xlab">{{ t.label }}</text>
<!-- Week overlay legend, top right inside the plot: current week in
accent, one muted specimen for the whole past range. -->
<g v-if="c.legend && c.chart.series.length > 1">
<line :x1="CHART_W - 98" :x2="CHART_W - 78" y1="10" y2="10" class="line" />
<text :x="CHART_W - 72" y="10" dominant-baseline="middle"
class="leglab">{{ c.chart.series[0].label }}</text>
<line :x1="CHART_W - 98" :x2="CHART_W - 78" y1="25" y2="25"
class="line past" style="opacity: 0.6" />
<text :x="CHART_W - 72" y="25" dominant-baseline="middle"
class="leglab">{{ pastLabel(c.chart.series) }}</text>
</g>
</svg>
</template>
<p v-else class="empty">{{ c.empty }}</p>
</section>
</template>
<style scoped>
/* The svg is stretched (preserveAspectRatio none), so all text lives in
HTML overlays positioned by the same fractions the geometry uses. */
.chartwrap {
padding-left: 2.8rem; /* y labels */
}
.plot {
display: flex;
flex-direction: column;
width: 100%;
}
.plotarea {
position: relative;
height: 8rem;
}
.xlabels {
position: relative;
height: 1.2rem;
}
/* Each chart is a self-contained SVG: the viewBox includes the axis label
margins, so nothing is positioned with HTML overlays. Never upscale past
the natural size (1 viewBox unit = 1 px, max-width set inline) — that
would blow up the constant-size text; smaller panels still scale the
chart down to fit. The margin-left (set inline) centers the chart above
its natural width; the svg always stays within page bounds.
overflow: visible lets wider fonts extend past the viewBox instead of
clipping. */
.chart {
display: block;
width: 100%;
height: 100%;
height: auto;
overflow: visible;
}
.ylab {
position: absolute;
left: -2.8rem;
width: 2.6rem;
text-align: right;
transform: translateY(50%);
font-size: 0.75rem;
color: var(--muted);
.chart .ylab,
.chart .xlab,
.chart .yaxis-label,
.chart .leglab {
font-family: system-ui, sans-serif; /* theme fonts can be overly styled */
font-size: 11px;
fill: var(--muted);
}
.chart .ylab {
font-variant-numeric: tabular-nums;
}
.xlab {
position: absolute;
top: 0.25rem;
transform: translateX(-50%);
font-size: 0.75rem;
color: var(--muted);
white-space: nowrap;
}
.xlabels .xlab:first-child { transform: none; }
.xlabels .xlab:last-child { transform: translateX(-100%); }
.chart .minor {
stroke: var(--line);
stroke-width: 1;
@@ -189,28 +193,10 @@ const viewChart = computed(() => buildChart(viewSeries.value, now.value))
stroke-linecap: round;
}
.legend {
display: flex;
gap: 1.2rem;
margin-top: 0.4rem;
font-size: 0.75rem;
color: var(--muted);
/* Past overlay weeks contrast with the current week's accent color. */
.chart .line.past {
stroke: var(--muted);
}
.legend span { color: var(--accent); }
.yaxis-label {
position: absolute;
top: 50%;
left: -2.8rem;
font-size: 0.75rem;
color: var(--muted);
writing-mode: vertical-rl;
white-space: nowrap;
transform: translateY(-50%) rotate(180deg);
}
section { margin-top: 1.8rem; }
.empty { color: var(--muted); }
</style>
+63 -110
View File
@@ -1,16 +1,21 @@
/**
* Chart geometry, smoothing, and SVG path generation for analytics charts.
*
* Fixed 720x180 viewBox, stretched to the panel width; values are per-unit
* rates (hour on the week view, day on month+).
* Fixed 720x180 plot area inside a larger viewBox that also holds the axis
* labels, so each chart SVG is self-contained; values are per-unit rates
* (hour on the week view, day on month+).
*/
import { DAY, HOUR, MIN5, WEEK, mondayUTC } from './time.js'
import { formatCount } from './format.js'
export const CHART_W = 720
export const CHART_H = 180
export const CHART_W = 1000
export const CHART_H = 150
export const PAD_TOP = 14 // room above the highest point
export const MARGIN_L = 40 // y tick labels + vertical axis label
export const MARGIN_B = 24 // x tick labels
export const VIEW_W = MARGIN_L + CHART_W + 8
export const VIEW_H = CHART_H + MARGIN_B
/**
* Y always starts at 0; the max is a multiple of a 1-2-5 major step with at
@@ -37,24 +42,19 @@ export function yScale(maxValue) {
}
/**
* Edge-aware adaptive Gaussian smoothing. A change-point detector first
* finds traffic-level shifts (two-unit totals compared on both sides of
* each bucket; strong ratio + significance marks a candidate, and each run
* of candidates keeps only its best-scoring bucket as an edge). Each
* edge-delimited segment is then smoothed independently: a broad two-unit
* pilot estimates the local traffic rate, which ramps the Gaussian sigma
* from ~0.4 units (isolated events stay narrow, peaking at ~1 event/unit)
* up to 1 unit (busy traffic gets full smoothing), and every bucket spreads
* its count with its local sigma, clipped to the segment and renormalized
* so total visitor count is preserved exactly. The unit is one hour on the
* week view and one day on the month+ views, so the smoothing time scale
* follows the range (month+ sigmas are 24x the hourly ones). The raw series
* is drawn faintly behind the curve for reference. Operates on raw counts.
* Edge-aware Gaussian smoothing with a fixed bandwidth. A change-point
* detector first finds traffic-level shifts (two-unit totals compared on
* both sides of each bucket; strong ratio + significance marks a candidate,
* and each run of candidates keeps only its best-scoring bucket as an
* edge). Each edge-delimited segment is then smoothed independently: every
* bucket spreads its count with a fixed Gaussian sigma chosen so N events
* in a single bucket peak at N events per unit, clipped to the segment and
* renormalized so total visitor count is preserved exactly. The unit is
* one hour on the week view and one day on the month+ views, so the
* smoothing time scale follows the range. The raw series is drawn faintly
* behind the curve for reference. Operates on raw counts.
*/
export function smooth(counts, binMinutes, unitMinutes, {
minSigmaMinutes = unitMinutes / Math.sqrt(2 * Math.PI),
maxSigmaMinutes = unitMinutes,
pilotSigmaMinutes = 2 * unitMinutes,
detectorWindowMinutes = 2 * unitMinutes,
// Count thresholds are defined per hour and scale with the unit, so
// "low traffic" means the same thing on hourly and daily views
@@ -62,8 +62,6 @@ export function smooth(counts, binMinutes, unitMinutes, {
highTrafficEvents = 10 * unitMinutes / 60,
minRatio = 2.5,
minSignificance = 4,
sigmaRampStart = 5 * unitMinutes / 60,
sigmaRampEnd = 20 * unitMinutes / 60,
} = {}) {
const n = counts.length
if (!n) return counts
@@ -105,68 +103,25 @@ export function smooth(counts, binMinutes, unitMinutes, {
i = j
}
const reflectIndex = (i, length) => {
while (i < 0 || i >= length) {
i = i < 0 ? -i - 1 : 2 * length - i - 1
}
return i
}
// Fixed sigma: N events in one bucket peak at N events per unit.
// sigma_bins * sqrt(2*pi) = rate = unitMinutes / binMinutes.
const sigmaBins = unitMinutes / (binMinutes * Math.sqrt(2 * Math.PI))
const radius = Math.ceil(4 * sigmaBins)
const gaussianFilterReflect = (values, sigmaBins) => {
const length = values.length
const radius = Math.ceil(4 * sigmaBins)
const kernel = new Float64Array(radius * 2 + 1)
let sum = 0
for (let k = -radius; k <= radius; k++) {
const w = Math.exp(-0.5 * (k / sigmaBins) ** 2)
kernel[k + radius] = w
sum += w
}
for (let i = 0; i < kernel.length; i++) kernel[i] /= sum
const out = new Float64Array(length)
for (let i = 0; i < length; i++) {
let value = 0
for (let k = -radius; k <= radius; k++) {
value += values[reflectIndex(i + k, length)] * kernel[k + radius]
}
out[i] = value
}
return out
}
// Process each discontinuity-delimited regime independently so neither
// the pilot nor the final Gaussian can see through a detected boundary.
// Process each discontinuity-delimited regime independently so the
// Gaussian cannot see through a detected boundary. Each input bin spreads
// its count with the fixed sigma; the kernel is renormalized after
// clipping to the segment, preserving total visitor count apart from
// floating-point error.
const bounds = [0, ...edges, n]
const smoothed = new Float64Array(n)
for (let b = 0; b < bounds.length - 1; b++) {
const lo = bounds[b]
const length = bounds[b + 1] - lo
const segment = counts.slice(lo, lo + length)
// Broad pilot estimates only the generic local traffic level used for
// choosing sigma; it is not the final displayed curve.
const pilot = gaussianFilterReflect(segment, pilotSigmaMinutes / binMinutes)
// Keep isolated/sparse traffic at the minimum bandwidth through
// sigmaRampStart events, then ramp toward maxSigmaMinutes (thresholds
// are per-hour rates scaled to the unit: low traffic is low traffic
// on every range).
const sigmaMinutes = new Float64Array(length)
for (let i = 0; i < length; i++) {
const ratePerUnit = pilot[i] * unitMinutes / binMinutes
let mix = (ratePerUnit - sigmaRampStart) / (sigmaRampEnd - sigmaRampStart)
mix = Math.sqrt(Math.max(0, Math.min(1, mix)))
sigmaMinutes[i] = minSigmaMinutes + mix * (maxSigmaMinutes - minSigmaMinutes)
}
// Each input bin spreads its own count using its local sigma. The
// per-bin kernel is renormalized after clipping to the segment,
// preserving total visitor count apart from floating-point error.
for (let j = 0; j < length; j++) {
const count = segment[j]
if (!count) continue
const sigmaBins = sigmaMinutes[j] / binMinutes
const radius = Math.ceil(4 * sigmaBins)
const start = Math.max(0, j - radius)
const end = Math.min(length, j + radius + 1)
let weightSum = 0
@@ -201,7 +156,7 @@ export function spline(pts) {
const c1y = clampY(p1.y + (p2.y - p0.y) / 6)
const c2y = clampY(p2.y - (p3.y - p1.y) / 6)
d += `C${p1.x + (p2.x - p0.x) / 6},${c1y} `
+ `${p2.x - (p3.x - p1.x) / 6},${c2y} ${p2.x},${p2.y}`
+ `${p2.x - (p3.x - p1.x) / 6},${c2y} ${p2.x},${p2.y}`
}
return d
}
@@ -240,7 +195,7 @@ export function buildChart(input, now = Date.now()) {
const nMajor = Math.round(max / step)
for (let k = 0; k <= nMajor; k++) {
const v = k * step
majors.push({ value: v, y: y(v), bottom: (1 - PAD_TOP / CHART_H) * (v / max) * 100 })
majors.push({ value: v, y: y(v), label: fmtY(v) })
}
if (minor) {
for (let v = minor; v < max; v += minor) {
@@ -254,37 +209,37 @@ export function buildChart(input, now = Date.now()) {
// ranges: boundary lines at Mondays / months / years.
const isWeek = t1 - t0 === WEEK
const isMonth = !isWeek && t1 - t0 <= 31 * DAY
const xticks = isWeek
? Array.from({ length: 7 }, (_, d) => {
const t = t0 + d * DAY + 12 * HOUR
return {
x: x(t), left: ((t - t0) / (t1 - t0)) * 100,
label: new Date(t).toLocaleDateString(undefined, {
weekday: 'short', timeZone: 'UTC',
}),
line: false,
}
let xticks
if (isWeek) {
xticks = Array.from({ length: 7 }, (_, d) => {
const t = t0 + d * DAY + 12 * HOUR
return {
x: x(t),
label: new Date(t).toLocaleDateString(undefined, {
weekday: 'short', timeZone: 'UTC',
}),
line: false,
}
})
} else if (isMonth) {
// t0 is day-aligned; label every day whose noon falls inside the range.
xticks = []
for (let day = t0; day + 12 * HOUR < t1; day += DAY) {
const date = new Date(day)
const t = day + 12 * HOUR
xticks.push({
x: x(t),
label: date.getUTCDate() === 1
? date.toLocaleDateString(undefined, { month: 'short', timeZone: 'UTC' })
: String(date.getUTCDate()),
line: false,
})
: isMonth
? Array.from(
{ length: Math.floor((t1 - Math.ceil(t0 / DAY) * DAY) / DAY) },
(_, d) => {
const day = Math.ceil(t0 / DAY) * DAY + d * DAY
const date = new Date(day)
const t = day + 12 * HOUR
return {
x: x(t), left: ((t - t0) / (t1 - t0)) * 100,
label: date.getUTCDate() === 1
? date.toLocaleDateString(undefined, { month: 'short', timeZone: 'UTC' })
: String(date.getUTCDate()),
line: false,
}
},
)
: xticksFor(t0, t1).map((t) => ({
x: x(t), left: ((t - t0) / (t1 - t0)) * 100,
label: fmtTick(t, t1 - t0), line: true,
}))
}
} else {
xticks = xticksFor(t0, t1).map((t) => ({
x: x(t), label: fmtTick(t, t1 - t0), line: true,
}))
}
return { max, majors, minors, series: drawn, xticks, unit }
}
@@ -347,7 +302,7 @@ export function buildDayChart(input, now = Date.now()) {
const nMajor = Math.round(max / step)
for (let k = 0; k <= nMajor; k++) {
const v = k * step
majors.push({ value: v, y: y(v), bottom: (1 - PAD_TOP / CHART_H) * (v / max) * 100 })
majors.push({ value: v, y: y(v), label: fmtY(v) })
}
if (minor) {
for (let v = minor; v < max; v += minor) {
@@ -363,12 +318,10 @@ export function buildDayChart(input, now = Date.now()) {
const d = new Date(t)
xticks.push({
x: ((t - t0) / (t1 - t0)) * CHART_W,
left: ((t - t0) / (t1 - t0)) * 100,
label: `${String(d.getUTCHours()).padStart(2, '0')}:00`,
line: false,
})
}
return { bars, skyline: skyline.trim(), max, majors, minors, xticks, unit: '5min', series: [] }
}
@@ -386,7 +339,7 @@ export function xticksFor(t0, t1) {
if (span <= 4 * 365 * DAY) {
const d = new Date(t0)
let t = Date.UTC(d.getUTCFullYear(), d.getUTCMonth() + 1, 1)
for (; t <= t1; ) {
for (; t <= t1;) {
ticks.push(t)
const m = new Date(t)
t = Date.UTC(m.getUTCFullYear(), m.getUTCMonth() + 1, 1)
+25 -5
View File
@@ -240,6 +240,14 @@ export function formatWhenIso(ts) {
return `${new Date(ts).toISOString().split('.')[0]}Z`
}
/**
* Compact read time for tooltips: "50s" under a minute, "1m23s" otherwise.
*/
export function formatReadTime(seconds) {
if (seconds < 60) return `${seconds}s`
return `${Math.floor(seconds / 60)}m${seconds % 60}s`
}
/**
* Compact visitor counts: plain below 1k, then 1.2k / 10k / 1.2M.
* Truncated, not rounded.
@@ -359,13 +367,16 @@ export function formatCrawlerRows(crawlers, clients, pageTree, now = Date.now())
const start = new Date(c.start).getTime()
if (start > g.lastStart) g.lastStart = start
if (c.entry?.startsWith('/')) {
g.pages.set(c.entry, (g.pages.get(c.entry) || 0) + 1)
const existing = g.pages.get(c.entry) || { count: 0, status: c.status || 200 }
existing.count += 1
if (c.status != null) existing.status = c.status
g.pages.set(c.entry, existing)
}
groups.set(c.client, g)
}
const totalHits = (g) => {
let n = 0
for (const c of g.pages.values()) n += c
for (const p of g.pages.values()) n += p.count
return n
}
return [...groups.values()]
@@ -380,8 +391,8 @@ export function formatCrawlerRows(crawlers, clients, pageTree, now = Date.now())
lastSeenIso: formatWhenIso(g.lastStart),
lastSeenLocal: formatWhenLocal(g.lastStart),
pages: [...g.pages.entries()]
.sort((a, b) => b[1] - a[1])
.map(([path, count]) => ({ ...stepOf(path, titles), count })),
.sort((a, b) => b[1].count - a[1].count)
.map(([path, info]) => ({ ...stepOf(path, titles), count: info.count, status: info.status })),
ip: client.ip || '',
ipDisplay: isHost ? mainDomain(host) : hostIP(client.ip) || client.ip || '—',
isHost,
@@ -499,8 +510,17 @@ export function formatVisitRows(visits, clients, pageTree, now = Date.now()) {
const titles = buildTitleMap(pageTree)
return [...(visits || [])].reverse().slice(0, 20).map((v) => {
const client = (clients || {})[v.client] || {}
const read = v.read || {}
const statuses = v.statuses || {}
const trail = [v.entry, ...(v.trail || [])]
.map((p) => stepOf(p, titles))
.map((p) => {
const step = stepOf(p, titles)
if (step) {
if (read[p]) step.readSeconds = read[p]
if (statuses[p]) step.status = statuses[p]
}
return step
})
.filter(Boolean)
const utmKeys = ['utm_source', 'utm_medium', 'utm_campaign', 'utm_term', 'utm_content']
const utmValues = utmKeys.map((k) => (v.utm || {})[k]).filter(Boolean)
+87 -19
View File
@@ -26,6 +26,15 @@ export function mondayUTC(t) {
return (d - ((d + 3) % 7)) * DAY
}
/** ISO 8601 week number of the week containing t (via its Thursday). */
export function isoWeek(t) {
const d = new Date(t)
d.setUTCHours(0, 0, 0, 0)
d.setUTCDate(d.getUTCDate() + 4 - (d.getUTCDay() || 7))
const yearStart = Date.UTC(d.getUTCFullYear(), 0, 1)
return Math.ceil(((d - yearStart) / DAY + 1) / 7)
}
/** Parse sparse timestamp buckets into a { epochMs: count } map. */
export function rawTimes(buckets) {
const raw = {}
@@ -44,7 +53,9 @@ export function sumRange(raw, t0, t1) {
/**
* One series per overlaid week: [this week, 1 week ago, ...], at native
* 5-minute resolution, up to 8 weeks back (and only weeks that overlap the
* recorded data at all). The current week is truncated at the current bucket
* recorded data at all). Each older week's timestamps are shifted forward
* onto the current week's axis so all curves overlay inside the plot.
* The current week is truncated at the current bucket
* — no fake zeroes drawn for the future. Counts are rates per hour
* (bucket count * 12): a lone visit in a 5-minute bucket reads as "12/h".
* The coarser ranges use per-day rates instead (unitMinutes = 24*60).
@@ -67,12 +78,13 @@ export function weeklySeries(buckets) {
: start + WEEK
const points = []
for (let t = start; t < end; t += MIN5) {
points.push({ t, count: raw[t] || 0 })
points.push({ t: t + back * WEEK, count: raw[t] || 0 })
}
out.push({
points,
label: back === 0 ? 'this week' : `${back}w ago`,
label: `Week ${isoWeek(start)}`,
opacity: Math.max(0.15, 1 - back * 0.25),
past: back > 0,
area: back === 0,
})
}
@@ -92,27 +104,38 @@ export function weeklySeries(buckets) {
* to per-day rates (the unit the month+ charts are read in).
* Ranges without a fixed span use the full data reach, but never less than
* their configured minSpan so the chart keeps a readable minimum x scale.
* t0 is aligned to the UTC day so the x labels cover the whole range;
* t1 is now, so the scale never extends into the future. The bucket size
* follows the resulting window (6h up to 31 days, daily beyond), so ranges
* covering the same window — "all" at its 30-day minimum vs "month" —
* render the identical curve.
*/
export function rollingSeries(buckets, rangeKey) {
const raw = rawTimes(buckets)
const times = Object.keys(raw).map(Number)
if (!times.length) return null
const { span, bucket, minSpan = 0 } = RANGES[rangeKey]
const t1 = Math.floor(Date.now() / bucket) * bucket + bucket
const earliest = Math.floor(Math.min(...times) / bucket) * bucket
const t0 = span != null
const t1 = Date.now()
const earliest = Math.min(...times)
const t0 = Math.floor((span != null
? t1 - span
: Math.min(earliest, t1 - minSpan)
: Math.min(earliest, t1 - minSpan)) / DAY) * DAY
// The bucket follows the actual window length, not the range key: when
// "all" is capped to its 30-day minimum it covers the very window "month"
// does, and daily bins would draw a different curve over the same data
// (coarser edge detection, points a day apart plotted at bin starts, the
// last point stuck at today's midnight instead of reaching now).
const bucketMs = t1 - t0 <= 31 * DAY ? Math.min(bucket, 6 * HOUR) : bucket
const points = []
for (let t = t0; t < t1; t += bucket) {
points.push({ t, count: sumRange(raw, t, t + bucket) })
for (let t = t0; t < t1; t += bucketMs) {
points.push({ t, count: sumRange(raw, t, t + bucketMs) })
}
return {
series: [{ points, label: '', opacity: 1, area: true }],
t0,
t1,
rate: DAY / bucket,
binMinutes: bucket / 60e3,
rate: DAY / bucketMs,
binMinutes: bucketMs / 60e3,
unitMinutes: 24 * 60,
unit: 'day',
}
@@ -152,19 +175,64 @@ export function makeSeries(buckets, rangeKey) {
/**
* Absolute UTC time window for a given range key. Used to filter visits,
* transitions and views to the same period the charts are showing.
* transitions and views for the non-chart stats on the analytics page.
* Every bounded range is a rolling span ending at now; the charts instead
* align week to Monday 00:00 UTC (overlaying previous weeks) and month+
* to UTC day boundaries, so their x windows differ from the stats range
* on purpose.
* Returns { t0, t1 } where null means unbounded.
*/
export function rangeWindow(rangeKey) {
const now = Date.now()
if (rangeKey === 'week') {
const start = mondayUTC(now)
return { t0: start, t1: start + WEEK }
}
if (rangeKey === 'all') {
return { t0: null, t1: null }
}
const { span, bucket } = RANGES[rangeKey]
const t1 = Math.floor(now / bucket) * bucket + bucket
return { t0: t1 - span, t1 }
const span = rangeKey === 'week' ? WEEK : RANGES[rangeKey].span
return { t0: now - span, t1: now }
}
/**
* Sum the bucketed transition matrix (from -> to -> bucket ISO -> count)
* into a plain from -> to -> count matrix for the window [t0, t1).
*/
export function filterTransitionsByRange(transitions, t0, t1) {
const out = {}
for (const [fr, tos] of Object.entries(transitions || {})) {
for (const [to, buckets] of Object.entries(tos)) {
let n = 0
for (const [k, c] of Object.entries(buckets)) {
const t = Date.parse(k)
if ((t0 == null || t >= t0) && (t1 == null || t < t1)) n += c
}
if (n) {
out[fr] = out[fr] || {}
out[fr][to] = n
}
}
}
return out
}
/** Keep only the 5-minute view buckets that fall inside [t0, t1). */
export function filterViewsByRange(views, t0, t1) {
const filtered = {}
for (const [path, buckets] of Object.entries(views || {})) {
const out = {}
for (const [k, c] of Object.entries(buckets)) {
const t = Date.parse(k)
if ((t0 == null || t >= t0) && (t1 == null || t < t1)) out[k] = c
}
if (Object.keys(out).length) filtered[path] = out
}
return filtered
}
/** Keep only records whose start time falls inside [t0, t1). */
export function filterRecordsByRange(records, t0, t1) {
const out = []
for (const r of records || []) {
const t = Date.parse(r.start)
if ((t0 == null || t >= t0) && (t1 == null || t < t1)) out.push(r)
}
return out
}
+226 -211
View File
@@ -22,9 +22,10 @@
* referer, then other utm_* tags. Visits with a UTM tag are grouped under
* that tag's value, not under the referer domain. A UTM source node only
* becomes a clickable link when every visit carrying that UTM tag came
* from the same referer. External exits are full-size nodes just outside
* their source page, angled away from the center. Each distinct full exit
* URL is its own node. Self-loops (reload pings) are skipped.
* from the same referer. External exits are full-size nodes in a row below
* the map, mirroring the source row, so the site itself stays in the
* middle. Each distinct full exit URL is its own node. Self-loops (reload
* pings) are skipped.
*/
import { MIN_READ_SECONDS } from './format.js'
@@ -123,27 +124,29 @@ const pillTangent = (s, margin = 0) => {
// Edge width (half-width of the thin middle) grows logarithmically with
// the count. The constants are scaled down by ~10× so busy ranges (day,
// year) do not overwhelm the graph with fat connectors. A single recorded
// transition still renders as a faint ~0.4 px line. Connections carrying
// less than PRUNE_FRACTION of the total traffic are not drawn at all (this
// also keeps the graph under ~100 connections).
// year) do not overwhelm the graph with fat connectors. Connections whose
// thin middle would render below MIN_WMID are culled entirely: fainter
// strands are practically invisible and only their wide end flares would
// show. Connections carrying less than PRUNE_FRACTION of the total traffic
// are likewise not drawn (this also keeps the graph under ~100 connections).
const WMID_MIN = 0.2
const WIDTH_GROWTH = 0.15
const PRUNE_FRACTION = 0.01
// ~0.8 px full width at natural size (1 viewBox unit = 1 px).
const MIN_WMID = 0.4
// Beads: each edge direction emits beads at count * BEAD_RATE beads per
// second (linear in the count). The rate is reduced ~10× across all time
// scales to keep the animation lightweight. The component simulates every
// bead independently in JS at BEAD_SPEED along the edge, with no limit on
// beads in flight.
export const BEAD_SPEED = 180 // svg units per second
// bead independently in JS with a constant traversal time per edge (speed
// relative to span length), with no limit on beads in flight.
export const BEAD_R = 3.2
const BEAD_RATE = 0.012 // beads per second per recorded transition
const FLOW_OFFSET = 3 // lane offset to the right of the travel direction
const MAX_EXT_IN = 8 // referer nodes in the top row
const MAX_EXT_OUT = 12 // exit nodes, at most MAX_EXT_OUT_PER_PAGE per page
const MAX_EXT_OUT_PER_PAGE = 3
const MAX_EXT_OUT = 12 // exit nodes in the bottom row
const EXT_GAP = 12 // vertical margin of the source/exit rows to the map
/** Flatten the site tree into navigation order via DFS. */
function buildNavigationOrder(pageTree) {
@@ -184,14 +187,13 @@ function collectInternalTransitions(transitions) {
return internal
}
/** Domain-only label for an external origin (path and www. removed). */
/** Domain-only label for an external origin (path and www. removed).
* Pills clip the text at their border; no length cap needed. */
function extLabel(ext) {
try {
const host = new URL(ext).hostname.replace(/^www\./, '')
return host.length > 25 ? `${host.slice(0, 24)}` : host
return new URL(ext).hostname.replace(/^www\./, '')
} catch {
const s = ext.replace(/^https?:\/\//, '').replace(/^www\./, '').split('/')[0]
return s.length > 25 ? `${s.slice(0, 24)}` : s
return ext.replace(/^https?:\/\//, '').replace(/^www\./, '').split('/')[0]
}
}
@@ -286,17 +288,9 @@ function annotateNodes(nodes, viewsData, titles, readMinutes) {
for (const n of nodes) {
n.views = viewCount(n.path)
n.readMin = readMinutes[n.path] || 0
// Article title inside the pill (shortened with ellipsis as needed),
// slug as fallback for pages missing from the site tree. The short
// path (last two segments, no leading /) renders above the pill; the
// front page shows a home symbol there instead (larger).
const slug = n.path.split('/').pop()
const label = titles.get(n.path) || (n.path === '/' ? '🏠︎' : slug)
n.label = label.length > 24 ? `${label.slice(0, 23)}` : label
const segs = n.path.split('/').filter(Boolean)
n.crumb = n.path === '/'
? '🏠︎'
: segs.length > 2 ? `…/${segs.slice(-2).join('/')}` : segs.join('/')
// Article title inside the pill (clipped at the pill border on
// render), slug as fallback for pages missing from the site tree.
n.label = titles.get(n.path) || (n.path === '/' ? '🏠︎' : n.path.split('/').pop())
n.title = titles.get(n.path) || ''
// Category (non-leaf) pages with no views in this window are omitted:
// their children move up in their place (see layoutGroups).
@@ -310,15 +304,13 @@ function annotateNodes(nodes, viewsData, titles, readMinutes) {
* the row following a shallow circular sag (center lowest) so connections
* between neighbors do not overlap the pills in between. Each top item's
* whole subtree fans out from it in menu (DFS preorder) order along a
* parabola that leaves the parent heading straight down and gradually
* bends to the right — no horizontal space is reserved for fans, they
* large-radius circular arc that leaves the parent heading straight down
* and gradually bends to the right — no horizontal space is reserved for fans, they
* extend under the slots to their right. Hidden index pages are omitted
* from the fan; when the top item itself is hidden, the fan shifts one
* slot up, the first visible child taking the top position. The short
* path shown above each pill (last two segments) keeps the omitted menu
* level visible.
* Also returns curved spoke paths tracing each fan: top slot to first
* member, then member to member in menu order, each bowed to the right.
* slot up, the first visible child taking the top position. Branch lanes
* labeled with the branch slug (see the branch-lane pass at the end)
* keep the omitted menu levels visible.
*/
function layoutGroups(root) {
// Top slots are spaced well over one pill width apart regardless of
@@ -328,8 +320,10 @@ function layoutGroups(root) {
// First pass: visible members per group, in menu order. Hidden index
// pages are skipped, but their children still appear. The front page
// forms its own group.
// forms its own group. groupRoots keeps each group's subtree root for
// the branch-curve pass below.
const groups = []
const groupRoots = []
for (const g of [root, ...root.children]) {
const members = []
if (g === root) {
@@ -341,53 +335,139 @@ function layoutGroups(root) {
}
walk(g)
}
if (members.length) groups.push(members)
if (members.length) {
groups.push(members)
groupRoots.push(g)
}
}
// Top row on a true circular sag: center lowest, edges raised by SAG.
// Top row on a large-radius circular arc whose bottom point is the
// LAST top item: each earlier item sits a bit higher (drop = 15% of
// the row span). Flat row when there is a single group.
const half = ((groups.length - 1) * SLOT) / 2 || 1
const SAG = TNODE_H * 0.6
const Rc = (half * half + SAG * SAG) / (2 * SAG)
const topY = (x) => SAG - Rc + Math.sqrt(Rc * Rc - x * x)
const span = (groups.length - 1) * SLOT
const topD = span * 0.15
const R_T = span ? (span * span + topD * topD) / (2 * topD) : 0
const topY = span
? (x) => topD - R_T + Math.sqrt(R_T * R_T - (x - half) * (x - half))
: () => 0
// Second pass: place groups. Fan members follow a right-opening cubic
// p(t) = (gx + B t³, y0 + t): the tangent stays vertical near the
// parent (leaving almost straight down) and bends right gently,
// reaching ~50° from vertical at the last member. Member spacing along
// the curve is the pill clearance (dt integrated against curve speed).
const spokePairs = [] // [from node, to node] — paths emitted below
// Second pass: place groups. Fan members follow a circular arc of
// large radius FAN_R centered at (gx + FAN_R, y0): the trail leaves
// the top node heading straight down (vertical tangent) and bends
// right gently, member i at arc angle π i·CLEAR/FAN_R (spaced by
// arc length CLEAR). A circle — not a spline — so the branch lanes
// below can be concentric arcs: identical forms, only radii differ.
const FAN_R = 1000
groups.forEach((members, gi) => {
const gx = gi * SLOT - half
const y0 = topY(gx)
members[0].x = gx
members[0].y = y0
const m = members.length - 1
if (!m) return
const tMax = m * CLEAR * 0.9
const B = 0.4 / (tMax * tMax)
let t = 0
for (let i = 1; i <= m; i++) {
const bend = 3 * B * t * t
t += CLEAR / Math.hypot(bend, 1)
const n = members[i]
n.x = gx + B * t * t * t
n.y = y0 + t
spokePairs.push([members[i - 1], n])
for (let i = 1; i < members.length; i++) {
const th = Math.PI - (i * CLEAR) / FAN_R
members[i].x = gx + FAN_R * (1 + Math.cos(th))
members[i].y = y0 + FAN_R * Math.sin(th)
}
})
// Fan spokes bow to the right via a quadratic control point pushed
// rightward from the segment midpoint.
const spokes = spokePairs.map(([p, n]) => {
const mx = (p.x + n.x) / 2
const my = (p.y + n.y) / 2
const bow = Math.hypot(n.x - p.x, n.y - p.y) * 0.18
return {
d: `M ${p.x.toFixed(2)} ${p.y.toFixed(2)} `
+ `Q ${(mx + bow).toFixed(2)} ${my.toFixed(2)} ${n.x.toFixed(2)} ${n.y.toFixed(2)}`,
// Branch lanes: one wide arc per path prefix (slug depth ≥ 1) whose
// subtree holds at least two visible nodes (a branch's visible nodes
// form one contiguous run in the fan's DFS preorder). Every lane of a
// group is an arc around the group's fan center with a radius one
// INDENT larger per parent level — concentric circles, so all lanes
// share exactly one form. Lanes span their branch's nodes plus a
// little extra tucked under the first/last pill (so the line caps are
// never visible) and run behind the pills. A label arc carries the
// branch slug, left-aligned just past the first pill and free to run
// to the lane's end — longer text simply passes under later pills,
// which are drawn on top. Hidden (unplaced) index pages still
// define a lane: it follows their promoted children, so lanes reflect
// the path structure rather than page existence.
const INDENT = 20 // lane spacing (radius) per nesting level (> lane width)
const END_TUCK = 22 // arc units tucked under the first/last pill
const GAP_TRIM = 32 // label arc clearance from the pills
// Labels are left-aligned on their guide: the guide starts just past the
// source pill's edge, the earliest point where the text is visible.
const LABEL_PAD = 6
// The label guide rides GUIDE_OFF outward of the lane centerline: the
// text's alphabetic baseline sits on the guide, so this puts the
// glyph middle (not the baseline) on the lane center at any zoom —
// dominant-baseline tricks are em-based and break under downscale.
const GUIDE_OFF = 3.5
const branches = []
groups.forEach((members, gi) => {
const g = groupRoots[gi]
if (g === root || members.length < 2) return
const idx = new Map(members.map((n, i) => [n, i]))
const C = [gi * SLOT - half + FAN_R, topY(gi * SLOT - half)]
const walk = (n) => {
let first = Infinity
let last = -1
const span = (m) => {
const k = idx.get(m)
if (k !== undefined) {
first = Math.min(first, k)
last = Math.max(last, k)
}
m.children.forEach(span)
}
span(n)
if (n.depth >= 1 && last > first) {
branches.push({ depth: n.depth, name: n.path.split('/').pop(), C, first, last })
}
n.children.forEach(walk)
}
walk(g)
})
return { GAP: TNODE_H * 2.6, spokes }
const depthMax = branches.reduce((d, b) => Math.max(d, b.depth), 1)
let arcLeft = Infinity // leftmost lane point, for the bounding box
const arcs = branches.map(({ depth, name, C, first, last }) => {
const R = FAN_R + (depthMax - depth) * INDENT
const th = (i) => Math.PI - (i * CLEAR) / FAN_R
const pt = (a, r) => [C[0] + r * Math.cos(a), C[1] + r * Math.sin(a)]
// Arc from angle a down to angle b (a > b; visually counterclockwise
// from the west point downward, hence sweep flag 0).
const arc = (a, b, r) => {
const [x0, y0] = pt(a, r)
const [x1, y1] = pt(b, r)
return `M ${x0.toFixed(2)} ${y0.toFixed(2)} A ${r.toFixed(2)} ${r.toFixed(2)} 0 0 0 ${x1.toFixed(2)} ${y1.toFixed(2)}`
}
const d = arc(th(first) + END_TUCK / R, th(last) - END_TUCK / R, R)
// Start just past the first pill: lanes leave the source node nearly
// vertically, so the pill's extent along the arc is its half height.
// The guide runs to the lane's end so long slugs are never cut off.
const ld = arc(th(first) - (TNODE_H / 2 + LABEL_PAD) / R,
th(last) - END_TUCK / R, R + GUIDE_OFF)
arcLeft = Math.min(arcLeft, pt(th(first) + END_TUCK / R, R)[0])
return { d, ld, label: name }
})
// Top lane: an arc along the top row's own circle, connecting
// the top nodes of all groups and tucked under the first and last of
// them (the arc bottoms at the last item, so it continues rightward
// under its pill). Drawn 50% thicker than branch lanes. A 🏠︎ label
// marks the lane right after the home pill, on a guide arc like
// the branch labels but with the offset and clearance scaled up by the
// same 50% to keep the glyph centered on the wider lane.
if (span) {
const d = `M ${(-half - END_TUCK).toFixed(2)} ${topY(-half - END_TUCK).toFixed(2)} `
+ `A ${R_T.toFixed(2)} ${R_T.toFixed(2)} 0 0 0 ${(half + END_TUCK).toFixed(2)} ${topY(half + END_TUCK).toFixed(2)}`
const rG = R_T + GUIDE_OFF * 1.5
const ptG = (x) => [x, topD - R_T + Math.sqrt(rG * rG - (x - half) ** 2)]
// Left-aligned like the branch labels: the guide starts just past the
// home pill's edge (scaled with the lane thickness).
const g0 = -half + TNODE_W / 2 + LABEL_PAD * 1.5
const g1 = SLOT - half - TNODE_W / 2 - GAP_TRIM * 1.5
const [gx0, gy0] = ptG(g0)
const [gx1, gy1] = ptG(g1)
arcs.unshift({
d,
ld: `M ${gx0.toFixed(2)} ${gy0.toFixed(2)} A ${rG.toFixed(2)} ${rG.toFixed(2)} 0 0 0 ${gx1.toFixed(2)} ${gy1.toFixed(2)}`,
label: '🏠︎',
top: true,
})
}
return { arcs, arcLeft }
}
/** Collapse opposite transition directions into one unordered pair per page pair. */
@@ -573,16 +653,18 @@ function buildFlows(a, b, ab, ba, visualScale = 1) {
}
}
const flows = []
if (ab) flows.push(flow(ab, t0, t1))
if (ba) flows.push(flow(ba, t1, t0))
// Stable key per edge direction so the component's bead simulation can
// match flows across data reloads and keep bead phases/positions.
if (ab) flows.push({ ...flow(ab, t0, t1), key: `${a.path} ${b.path}` })
if (ba) flows.push({ ...flow(ba, t1, t0), key: `${b.path} ${a.path}` })
return flows
}
/**
* Half-width for a connection middle: logarithmic in the count, anchored
* so a single count lands exactly at WMID_MIN (~1 px line), uncapped.
* Absolute on purpose — cool routes stay visible regardless of how hot
* the hottest connection is.
* at WMID_MIN, uncapped. Absolute on purpose — cool routes stay visible
* regardless of how hot the hottest connection is. Callers cull results
* below MIN_WMID.
*/
const scaledWidth = (count) => {
if (count <= 0) return 0
@@ -608,59 +690,13 @@ function buildInternalEdges(pairs, byPath, visualScale = 1) {
const b = byPath.get(pt)
if (a.hidden || b.hidden) continue // unplaced index pages are omitted
const wMid = scaledWidth((ab + ba) * visualScale)
if (wMid <= 0) continue
if (wMid < MIN_WMID) continue
edges.push(buildRibbon(a, b, ab, ba, wMid))
flows.push(...buildFlows(a, b, ab, ba, visualScale))
}
return { edges, flows }
}
/**
* Sum the bucketed transition matrix (from -> to -> bucket ISO -> count)
* into a plain from -> to -> count matrix for the window [t0, t1).
*/
export function filterTransitionsByRange(transitions, t0, t1) {
const out = {}
for (const [fr, tos] of Object.entries(transitions || {})) {
for (const [to, buckets] of Object.entries(tos)) {
let n = 0
for (const [k, c] of Object.entries(buckets)) {
const t = Date.parse(k)
if ((t0 == null || t >= t0) && (t1 == null || t < t1)) n += c
}
if (n) {
out[fr] = out[fr] || {}
out[fr][to] = n
}
}
}
return out
}
/** Keep only the 5-minute view buckets that fall inside [t0, t1). */
export function filterViewsByRange(views, t0, t1) {
const filtered = {}
for (const [path, buckets] of Object.entries(views || {})) {
const out = {}
for (const [k, c] of Object.entries(buckets)) {
const t = Date.parse(k)
if ((t0 == null || t >= t0) && (t1 == null || t < t1)) out[k] = c
}
if (Object.keys(out).length) filtered[path] = out
}
return filtered
}
/** Keep only visits whose start time falls inside [t0, t1). */
export function filterVisitsByRange(visits, t0, t1) {
const out = []
for (const v of visits || []) {
const t = Date.parse(v.start)
if ((t0 == null || t >= t0) && (t1 == null || t < t1)) out.push(v)
}
return out
}
const UTM_PRIORITY = ['utm_campaign', 'utm_source']
const UTM_FALLBACK = ['utm_medium', 'utm_content', 'utm_term', 'utm_id']
@@ -726,11 +762,13 @@ function collectSourcePairs(visits) {
* flows.
* Sources (incoming links) are derived from visit UTM/referer data and form
* a row centered above the map, hottest first; exits come from the
* transition matrix and sit just outside their source page.
* transition matrix and form a matching row centered below the map, so
* the site itself stays in the middle. Both rows sit EXT_GAP beyond the
* map's bounds.
* Widths and pruning use the same log scale and traffic-share rule as
* internal connections.
*/
function buildExternal({ sources, exits }, byPath, gap, innerBounds, visualScale = 1) {
function buildExternal({ sources, exits }, byPath, innerBounds, visualScale = 1) {
const extNodes = []
const edges = []
const flows = []
@@ -744,17 +782,9 @@ function buildExternal({ sources, exits }, byPath, gap, innerBounds, visualScale
const width = (count) => scaledWidth(count * visualScale)
// Pill-shape overlap test (axis-aligned pills): much tighter than the
// bounding-circle test, so diagonal placements can sit close.
const overlaps = (x, y) =>
[...byPath.values(), ...extNodes].some(
(n) => !n.hidden
&& Math.abs(n.x - x) < TNODE_W + 12
&& Math.abs(n.y - y) < TNODE_H + 12,
)
// Incoming: one source node per identified source, in a row centered
// above the map, with an edge to each page that source led to.
// above the map, with an edge to each page that source led to. A source
// whose connectors are all culled (below MIN_WMID) is dropped itself.
const bySource = new Map() // source -> pairs, sorted by total incoming count
for (const p of liveSources.filter((p) => p.in >= minCount)) {
const g = bySource.get(p.source) || []
@@ -771,9 +801,11 @@ function buildExternal({ sources, exits }, byPath, gap, innerBounds, visualScale
}))
.sort((a, b) => b.total - a.total)
.slice(0, MAX_EXT_IN)
.filter(({ ps }) =>
ps.some((p) => !byPath.get(p.page).hidden && width(p.in) >= MIN_WMID))
if (origins.length) {
const cx = (innerBounds.x0 + innerBounds.x1) / 2
const y = innerBounds.y0 - TNODE_BOUND - 64
const y = innerBounds.y0 - TNODE_BOUND - EXT_GAP
const spacing = TNODE_W + 44
const x0 = cx - ((origins.length - 1) * spacing) / 2
origins.forEach(({ source, ps, total, href, isUtm }, i) => {
@@ -781,7 +813,7 @@ function buildExternal({ sources, exits }, byPath, gap, innerBounds, visualScale
const xn = {
path: source,
href,
label: label.length > 25 ? `${label.slice(0, 24)}` : label,
label, // clipped at the pill border on render
x: x0 + i * spacing,
y,
count: total,
@@ -792,79 +824,56 @@ function buildExternal({ sources, exits }, byPath, gap, innerBounds, visualScale
const page = byPath.get(p.page)
if (page.hidden) continue
const wMid = width(p.in)
if (wMid <= 0) continue
if (wMid < MIN_WMID) continue
edges.push(buildRibbon(xn, page, p.in, 0, wMid, true))
flows.push(...buildFlows(xn, page, p.in, 0, visualScale))
}
})
}
// Outgoing: group by full URL so several links to the same domain stay
// distinct; each shows the total count across all pages linking to it.
// Placement looks for empty space around the source page, always
// leftward: diagonal down-left first (often right beside the source,
// no need to drop below the fans), then left, up-left, and steeper
// fallbacks; the distance grows until a spot is free. Several exits of
// one page start at different directions.
const GAP = gap
const DIRS = [
(3 * Math.PI) / 4, // diagonal down-left
Math.PI, // left
(5 * Math.PI) / 4, // diagonal up-left
Math.PI / 2 + 0.35, // steep down-left
Math.PI - 0.35, // shallow up-left
(3 * Math.PI) / 4 + 0.5, // far down-left
]
const outgoing = liveExits.filter((p) => p.out >= minCount)
.sort((a, b) => b.out - a.out)
const perPage = new Map()
const selected = []
for (const p of outgoing) {
const used = perPage.get(p.page) || 0
if (used >= MAX_EXT_OUT_PER_PAGE) continue
perPage.set(p.page, used + 1)
selected.push(p)
if (selected.length >= MAX_EXT_OUT) break
// Outgoing: one exit node per distinct full URL (so several links to
// the same domain stay distinct), showing the total count across all
// pages linking to it, in a row centered below the map (hottest
// first), mirroring the source row above. Each (URL, page) pair
// contributes an edge from that page. An exit whose connectors are all
// culled (below MIN_WMID) is dropped itself.
const byExt = new Map() // full URL -> { ext, out, pairs }
for (const p of liveExits.filter((p) => p.out >= minCount)) {
const g = byExt.get(p.ext) || { ext: p.ext, out: 0, pairs: [] }
g.out += p.out
g.pairs.push(p)
byExt.set(p.ext, g)
}
const exitNodes = new Map() // full URL -> node
const placedPerPage = new Map() // for the placement direction offset
for (const p of selected) {
const page = byPath.get(p.page)
if (page.hidden) continue
let xn = exitNodes.get(p.ext)
if (!xn) {
const used = placedPerPage.get(p.page) || 0
placedPerPage.set(p.page, used + 1)
let x = 0
let y = 0
let found = false
for (let di = 0; di < DIRS.length && !found; di++) {
const ang = DIRS[(used + di) % DIRS.length]
for (let dist = GAP; dist <= GAP * 3.5; dist += GAP * 0.4) {
x = page.x + Math.cos(ang) * dist
y = page.y + Math.sin(ang) * dist
if (!overlaps(x, y)) { found = true; break }
}
}
if (!found) continue // no empty space near the page: leave it out
xn = {
path: p.ext,
href: p.ext,
label: extLabel(p.ext),
x,
const targets = [...byExt.values()]
.sort((a, b) => b.out - a.out)
.slice(0, MAX_EXT_OUT)
.filter(({ pairs }) =>
pairs.some((p) => !byPath.get(p.page).hidden && width(p.out) >= MIN_WMID))
if (targets.length) {
const cx = (innerBounds.x0 + innerBounds.x1) / 2
const y = innerBounds.y1 + TNODE_BOUND + EXT_GAP
const spacing = TNODE_W + 44
const x0 = cx - ((targets.length - 1) * spacing) / 2
targets.forEach(({ ext, out, pairs }, i) => {
const xn = {
path: ext,
href: ext,
label: extLabel(ext),
x: x0 + i * spacing,
y,
count: 0,
count: out,
kind: 'exit',
}
exitNodes.set(p.ext, xn)
extNodes.push(xn)
}
xn.count += p.out
const wMid = width(p.out)
if (wMid <= 0) continue
edges.push(buildRibbon(page, xn, p.out, 0, wMid, true))
flows.push(...buildFlows(page, xn, p.out, 0, visualScale))
for (const p of pairs) {
const page = byPath.get(p.page)
if (page.hidden) continue
const wMid = width(p.out)
if (wMid < MIN_WMID) continue
edges.push(buildRibbon(page, xn, p.out, 0, wMid, true))
flows.push(...buildFlows(page, xn, p.out, 0, visualScale))
}
})
}
return { extNodes, edges, flows }
@@ -873,7 +882,7 @@ function buildExternal({ sources, exits }, byPath, gap, innerBounds, visualScale
/**
* Build the transition map model.
* Returns { nodes, edges, flows, extNodes, arcs, bounds } or null when
* there is nothing to show. `arcs` holds the family spokes; `nodes` only
* there is nothing to show. `arcs` holds the branch curves; `nodes` only
* contains placed (visible) nodes.
*/
export function buildTransitionGraph(data, pageTree, visits = [], visualScale = 1) {
@@ -889,28 +898,34 @@ export function buildTransitionGraph(data, pageTree, visits = [], visualScale =
const { nodes, byPath, root } = buildNodeTree(internal, navOrder)
sortByNav(root, navOrder)
annotateNodes(nodes, data?.views, titles, readMinutes)
const { GAP, spokes } = layoutGroups(root)
const { arcs, arcLeft } = layoutGroups(root)
const placed = nodes.filter((n) => !n.hidden)
const pairs = aggregatePairs(internal)
const { edges, flows } = buildInternalEdges(pairs, byPath, visualScale)
// Tight bounding box of the placed page nodes; external nodes extend it.
const pad = 16
// Tight bounding box of the placed page nodes, extended to cover the
// branch curves running left of the pills; external nodes extend it.
// Margins cover just the ribbon surround (pill outline + flare margin
// S=4) plus a small pad: the pill's half width/height, not its diagonal
// radius, so the map crops tight especially at top and bottom.
const pad = 8
const MX = TNODE_W / 2 + 4 + pad
const MY = TNODE_H / 2 + 4 + pad
const xs = placed.map((n) => n.x)
const ys = placed.map((n) => n.y)
const bounds = {
x0: Math.min(...xs) - TNODE_BOUND - pad,
y0: Math.min(...ys) - TNODE_BOUND - pad,
x1: Math.max(...xs) + TNODE_BOUND + pad,
y1: Math.max(...ys) + TNODE_BOUND + pad,
x0: Math.min(Math.min(...xs) - MX, arcLeft - pad),
y0: Math.min(...ys) - MY,
x1: Math.max(...xs) + MX,
y1: Math.max(...ys) + MY,
}
const ext = buildExternal({ sources, exits }, byPath, GAP, bounds, visualScale)
const ext = buildExternal({ sources, exits }, byPath, bounds, visualScale)
for (const xn of ext.extNodes) {
bounds.x0 = Math.min(bounds.x0, xn.x - TNODE_BOUND - pad)
bounds.y0 = Math.min(bounds.y0, xn.y - TNODE_BOUND - pad)
bounds.x1 = Math.max(bounds.x1, xn.x + TNODE_BOUND + pad)
bounds.y1 = Math.max(bounds.y1, xn.y + TNODE_BOUND + pad)
bounds.x0 = Math.min(bounds.x0, xn.x - MX)
bounds.y0 = Math.min(bounds.y0, xn.y - MY)
bounds.x1 = Math.max(bounds.x1, xn.x + MX)
bounds.y1 = Math.max(bounds.y1, xn.y + MY)
}
return {
@@ -918,7 +933,7 @@ export function buildTransitionGraph(data, pageTree, visits = [], visualScale =
edges: [...edges, ...ext.edges],
flows: [...flows, ...ext.flows],
extNodes: ext.extNodes,
arcs: spokes,
arcs,
bounds,
}
}
+26 -6
View File
@@ -109,6 +109,8 @@ class Visit(msgspec.Struct, omit_defaults=True):
utm: dict[str, str] = {}
#: Active reading time per path (seconds), keyed by path.
read: dict[str, int] = {}
#: HTTP status of the response when the path was first seen (200 or 404).
statuses: dict[str, int] = {}
class CrawlerHit(msgspec.Struct, omit_defaults=True):
@@ -125,6 +127,8 @@ class CrawlerHit(msgspec.Struct, omit_defaults=True):
referer: str = ""
#: Raw query string of the landing URL (UTM tags can be parsed from it).
query: str = ""
#: HTTP status of the served response (200 or 404 for content pages).
status: int = 200
class AbuseHit(msgspec.Struct, omit_defaults=True):
@@ -246,15 +250,15 @@ def _utm_tags(query: str) -> dict[str, str]:
_CRAWLER_TIMEOUT = timedelta(seconds=10)
#: UAs of JS-running crawlers, which would register as visitors on their
#: ping. Anything calling itself a "bot" matches; known crawlers without
#: that token (GoogleOther) are listed as extra alternates. No source
#: verification: a spoofed bot UA just lands in the crawler list, and
#: ping. Anything calling itself a "bot" or "spider" matches; known crawlers
#: without those tokens (GoogleOther) are listed as extra alternates. No
#: source verification: a spoofed bot UA just lands in the crawler list, and
#: scanners that probe telltale paths are caught by the abuse rules anyway.
_BOT_UA = re.compile(r"bot|googleother", re.IGNORECASE)
_BOT_UA = re.compile(r"bot|spider|googleother", re.IGNORECASE)
def _is_bot_ua(ua: str) -> bool:
"""True when the UA claims a crawler identity (Googlebot, Applebot, ...)."""
"""True when the UA claims a crawler identity (bot or spider)."""
return bool(_BOT_UA.search(ua))
#: Plain-404 count per IP that classifies it as abuse even without a
@@ -324,6 +328,9 @@ class Store:
#: Document GETs that have not yet been matched by a ping. Kept
#: in RAM only; expired entries are written to ``data.crawlers``.
self.pending_crawlers: list[CrawlerHit] = []
#: client hash -> {path: status} for recent document GETs, consumed
#: by the matching ping to record the status of each visited path.
self.pending_statuses: dict[bytes, dict[str, int]] = {}
#: ip -> number of plain (non-telltale) 404s seen, in RAM only;
#: reaching ``_ABUSE_404_THRESHOLD`` classifies the IP as abuse.
self.not_found_counts: dict[str, int] = {}
@@ -571,6 +578,7 @@ class Store:
referer: str,
client_hash: bytes,
utm: dict[str, str] | None = None,
status: int = 200,
) -> Visit:
now = datetime.now(UTC)
visit = Visit(
@@ -580,6 +588,7 @@ class Store:
client=client_hash,
utm=utm or {},
)
visit.statuses[entry] = status
self.data.visits.append(visit)
self.sessions[client_hash] = len(self.data.visits) - 1
self._count(self.data.site_visits, _bucket(now))
@@ -595,6 +604,8 @@ class Store:
ua: str,
full_path: str,
accept_language: str = "",
*,
status: int = 200,
) -> list[bytes]:
"""Stash the entry referer/UTM tags and queue a pending crawler hit.
@@ -642,8 +653,10 @@ class Store:
client=client_hash,
referer=self.pending_referers.get(ip, ""),
query=query,
status=status,
)
)
self.pending_statuses.setdefault(client_hash, {})[entry] = status
return flushed
def _add_read(self, client_hash: bytes, path: str, seconds: int) -> None:
@@ -697,8 +710,9 @@ class Store:
if hide:
# Admin ping: cancel pending crawler hits and scrub the session.
self.pending_crawlers = [
hit for hit in self.pending_crawlers if hit.client == client_hash
hit for hit in self.pending_crawlers if hit.client != client_hash
]
self.pending_statuses.pop(client_hash, None)
index = self.sessions.pop(client_hash, None)
if index is not None and index < len(self.data.visits):
self._remove_visit(index)
@@ -730,6 +744,10 @@ class Store:
return None, flushed
index = self.sessions.get(client_hash)
fr = fr_path or "(direct)"
statuses = self.pending_statuses.setdefault(client_hash, {})
target_status = statuses.pop(target, None) or 200
if not statuses:
self.pending_statuses.pop(client_hash, None)
if index is None or index >= len(self.data.visits):
# No known session: the initial ping of a fresh page load (or
# missing data after a server restart) — start a visit.
@@ -740,6 +758,7 @@ class Store:
self.pending_referers.pop(ip, ""),
client_hash,
utm=self.pending_utms.pop(ip, {}),
status=target_status,
)
else:
visit = self.data.visits[index]
@@ -750,6 +769,7 @@ class Store:
# First-seen only: repeat pages and repeated exits don't append.
if visit.entry != target and target not in visit.trail:
visit.trail.append(target)
visit.statuses[target] = target_status
self._save()
visit_index = index if index is not None and index < len(self.data.visits) else None
return visit_index, flushed
+4 -3
View File
@@ -853,7 +853,7 @@ async def analytics_ping(ping: AnalyticsPing, request: Request) -> None:
_schedule_client_enrichment(flushed_clients)
def _track_entry(path: str, request: Request) -> list[bytes]:
def _track_entry(path: str, request: Request, *, status: int = 200) -> list[bytes]:
"""Stash the referer/UTM tags and queue a pending crawler hit for the GET.
Nothing is counted on the GET itself — the client's /_a ping starts the
@@ -890,6 +890,7 @@ def _track_entry(path: str, request: Request) -> list[bytes]:
request.headers.get("user-agent", ""),
full_path,
request.headers.get("accept-language", ""),
status=status,
)
@@ -1228,7 +1229,7 @@ async def show_page(request: Request, path: str) -> Response:
# Category label without a landing page: placeholder with the pen
# to create it (404 — no page here, but the node is real).
if _is_trackable_path(path):
flushed = _track_entry(path, request)
flushed = _track_entry(path, request, status=404)
_schedule_client_enrichment(flushed)
return _html_response(
request,
@@ -1254,6 +1255,6 @@ async def show_page(request: Request, path: str) -> Response:
accept_language,
)
asyncio.create_task(_enrich_client(client_hash))
flushed = _track_entry(path, request)
flushed = _track_entry(path, request, status=404)
_schedule_client_enrichment(flushed)
return _html_response(request, "not-found", path, 404)