惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

博客园_首页
H
Help Net Security
N
Netflix TechBlog - Medium
Apple Machine Learning Research
Apple Machine Learning Research
P
Proofpoint News Feed
A
About on SuperTechFans
V
V2EX
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
宝玉的分享
宝玉的分享
aimingoo的专栏
aimingoo的专栏
F
Fortinet All Blogs
博客园 - 【当耐特】
Microsoft Security Blog
Microsoft Security Blog
Martin Fowler
Martin Fowler
I
InfoQ
Google DeepMind News
Google DeepMind News
人人都是产品经理
人人都是产品经理
Engineering at Meta
Engineering at Meta
腾讯CDC
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
B
Blog RSS Feed
U
Unit 42
The Cloudflare Blog
Y
Y Combinator Blog

Hacker News: Show HN

PurrrrrFocus: Pomodoro Timer App - App Store Workflow Engine — Multi-Step Orchestration for Bun RapidPhoto: Pro Photo Editor App - App Store GitHub - DheerG/swarms: Achieve extraordinary results with claude code across a variety of tasks SPICE simulation → oscilloscope → verification with Claude Code — Lucas Gerads Show HN: VCoding – A 5 MB native Windows IDE with no dynamic dependencies Show HN: LLMs don't hallucinate because they're bad at math, it's the format GitHub - Agent-FM/agentfm-core: AgentFM is a peer-to-peer network that turns everyday computers into a decentralized AI supercomputer. AgentFM lets you run massive AI workloads directly across a global mesh of idle CPUs and GPUs. Show HN: Tracking Top US Science Olympiad Alumni over Last 25 Years GitHub - Potarix/agent-hub: One place to talk to all your agents Show HN: Runtime security for AI agents(injection,tool abuse, data exfiltration) GitHub - dubeyKartikay/lazyspotify: Terminal Spotify client for macOS and Linux GitHub - the-banana-tool/king-louie: Easy to use GUI Personal AI Assistant. Win/Linux/Mac. Show HN I made my vacation rental bookable by AI agents–no Airbnb, 0% commission GitHub - basteez/jsf-autoreload: maven plugin to enable hot reload on jsf projects uvm32/hosts/host-gdbstub at main · ringtailsoftware/uvm32 GitHub - labsai/EDDI: Config-driven engine that turns JSON into production-grade AI agents. Multi-agent orchestration, 12+ LLM providers, MCP/A2A protocols, RAG, persistent memory, and enterprise compliance (EU AI Act, GDPR, HIPAA). Built on Quarkus. GitHub - glitchnsec/fortyone-oss: AI Executive Assistant Platform Quickstart | Alien GitHub - muxshed/shed: One stream in, or many. Every destination, simultaneously. No cloud middleman, no per-channel fees, no limits. GitHub - ocrbase-hq/ocrbase: 📄 PDF/IMG ->.MD/JSON Document OCR API for PaddleOCR and GLMOCR. Self-hostable. GitHub - impactjo/home-memory: MCP server that lets your AI assistant remember everything about your home. GitHub - Sets88/dbcls: DbCls is a powerful terminal database client that supports various databases GitHub - neptun2000/heor-agent-mcp GitHub - SeanFDZ/macmind: Single-layer transformer in HyperTalk for the classic Macintosh RollQuation: Math Puzzles - Apps on Google Play GitHub - dropbox/witchcraft Show HN: Agent-cache – Multi-tier LLM/tool/session caching for Valkey and Redis GitHub - opentalon/opentalon: OpenTalon is an open-source platform built from the ground up in Go as a robust alternative to OpenClaw LinkedIn™ 职位抓取工具 - Chrome 应用商店
gp-treemap gallery
thad_imbue · 2026-04-30 · via Hacker News: Show HN

Introduction

Think of a treemap as a pie chart, but with hierarchy — every slice has its own pie-chart breakdown inside it, and so on, all the way down. Nested rectangles stand in for slices, each sized in proportion to a numeric value, so the whole distribution — and the relative weight of every part at every level — is visible at a glance. A hat tip to GrandPerspective, the classic macOS disk-usage viewer whose dense, label-free mosaic inspired gp-treemap's visual style.

Below: interactive sample visualizations built with gp-treemap. Each entry is a self-contained HTML file — bundle and dataset are inlined, so they render from any URL (or file://) with no server.

Features you'll find in every view below:

  • Breadcrumb bar along the top, showing the ancestor chain from the root down to the currently clicked cell:
    • The home icon on the far left is the tree root — click it to focus the root, double-click to zoom back out all the way.
    • Each segment after the home is a clickable ancestor; single-click focuses it (highlights its subtree), double-click zooms in so that node becomes the new root.
    • A magnifying glass marks the current zoom anchor — everything to its right lives inside the zoomed-in subtree.
  • Mouse-wheel focusing: with a cell clicked, scroll the wheel to walk focus up and down its ancestor chain — scroll down toward the root (shallower), scroll up toward the clicked leaf (deeper). It's the fastest way to see where a cell sits in the hierarchy.
  • Depth truncation via the −/+ depth stepper in the toolbar. Capping the depth prunes levels below the budget and rolls their contents up into the parent, which makes those parent cells big enough to actually carry a readable label. Lift the cap back to to see every leaf. The budget re-applies from the current zoom root, so you can drill in and still get a labelled, uncluttered view at each level.
  • Plus the usual: hover for exact values, click to select, double-click to zoom, and Size / Color / Path controls in the header of the tabular views.

Examples

gp-treemap source tree — disk usage

The gp-treemap repository looking at itself — the working tree plus .git and node_modules. Cells are sized by bytes on disk and colored by file extension; hover for exact sizes, click to select, scroll-wheel to focus an ancestor, double-click to zoom. To scan any directory on your own machine and get a self-contained HTML like this one:

npx -p @imbue-ai/gp-treemap gpdu ~/path/to/dir

Trees from Tables

The entries below use the path / tabular treemap (gp-treemap-table), which builds the hierarchy on the fly from a flat table of rows. The Path control in the header lists the columns used to group the data, from outermost to innermost:

Path control with draggable Fuel, Continent, Country chips

Drag the chips to reorder them and the nesting re-computes immediately: with Fuel first (as above) you get one cell per fuel with continents inside; drag Continent to the front instead and the same rows re-tile as one cell per continent with fuels inside. Same data, totally different story — and no re-export, just a drag. Click the × on a chip to drop that column from the hierarchy entirely, or the + to add another.

Global electricity generation — path: fuel → continent → country

Ember's Yearly Electricity Data (2023, TWh by country and fuel). Built with tools/table-treemap.js from tabular JSONL: use the Size, Color and Path controls in the header to re-group on the fly — drag the Path chips to reorder the hierarchy. Colored by fuel, Catppuccin Mocha theme.

Reproduce it from the pre-filtered JSONL in this repo:

curl -LO https://github.com/imbue-ai/gp-treemap/raw/main/samples/data/table/energy-2023.jsonl
npx -p @imbue-ai/gp-treemap gp-treemap-table \
  --size=Generation_TWh --color=Fuel \
  --path=Fuel,Continent,Country \
  --theme=catppuccin \
  energy-2023.jsonl

California city wages 2024 — where do city payrolls go?

Every California city employee with reported wages in 2024 — about 334,000 rows from the state's Government Compensation in California raw export, cell size = TotalWages. Grouped by department → county → city → position → _row, but with depth = 1 at first so only the top-level Department bucket is drawn — a cleaner first read of where the money goes. Police and Fire dominate; Parks & Rec, Public Works and Transportation come next. Click the + in the component toolbar to reveal another level, or double-click any cell to zoom in — the depth budget re-applies from the new root, so you can drill in without losing your orientation.

Trivia: DepartmentOrSubdivision is a free-text field with no normalization, so each spelling is a separate top-level cell. Police alone shows up as Police, Police Department, Police Services Agency, Police Administration, and Police Field Services. Fire is Fire, Fire Department, Fire-Rescue, Fire Operations, Fire Suppression, and Fire Suppression & Rescue. Parks & Rec is Recreation And Parks, Recreation and Park Commission, Parks and Recreation, Parks & Recreation, and Parks & Recreation Administration. And Airports and Airport Commission are two different cells too.

California city wages 2024 — one cell per employee

Same file as above, but with the depth budget lifted so every one of the 334k rows gets its own pixel-sliver cell inside its position / city / county / department grouping. Cell size is the employee's total wages, so the layout itself reveals pay distribution: zoom into Police or a big city and you can see a handful of disproportionately large cells (the top earners) alongside a vast mosaic of small, similarly-sized cells (the bulk of staff at comparable pay).

(See the intro above for the Police / Fire / Parks & Rec spelling carnival.)

To reproduce this plot from scratch, grab the raw GCC zip and pipe its CSV into gp-treemap-table:

curl -LO https://github.com/imbue-ai/gp-treemap/raw/main/samples/data/table/ca-raw/2024_City.zip
unzip 2024_City.zip
npx -p @imbue-ai/gp-treemap gp-treemap-table \
  --size=TotalWages --color=DepartmentOrSubdivision \
  --path=DepartmentOrSubdivision,EmployerCounty,EmployerName,Position,_row \
  --theme=one-dark \
  2024_City.csv

UC wages 2024 — campus → position → individual (Nord theme)

Public-record compensation for every University of California employee paid at least $1,000 in 2024 (~334k rows). Each top-level cell is one of the 11 campuses, sized by total wages and colored by campus; zoom into one to see its positions, and zoom again to see individual employees. UCLA and UCSF (both home to big medical centers) dominate; the Office of the President is the smallest.

Trivia: the state's raw export lists one of the campuses as "Univeristy (sic) of California - Santa Cruz". Hi, Santa Cruz! Also curious: ~10,000 employees, including some of UCLA's top earners at $3–4M, are filed under Blank Assistant 1/2/3. That's an actual UC clerical/administrative title series ("Class B – Clerical and Allied Services"), so the low-end Blank Assistants are clerical workers as advertised — but the million-dollar ones at UCLA (a big one is pre-focused) are clearly not clerical staff. We don't know what they really are; the public export doesn't say.

UC wages 2024 — position → campus → individual (Nord theme)

Same dataset as above, regrouped the other way: each top-level cell is now one job title, and inside that cell you see how that title's payroll divides among the 11 campuses. Where the previous view let you compare "what does a campus do?", this one lets you compare "who pays more for the same title?" — e.g. clinical nurses at UCSF vs. UCLA vs. UCSD, or the mutual-fund-like Professor series versus the Athletics coaches hiding inside Blank Assistant.

US federal outlays FY2024 — Treasury MTS Table 5

Gross fiscal-year-to-date outlays as of 2024-09-30, from the US Treasury's Monthly Treasury Statement. Path: Category → Agency → Bureau, colored categorically by top-level category with the Viridis palette. Click into a cell to zoom; change Color to YoY_Change_Pct to see which areas grew or shrank.

US federal outlays FY2024 — colored by year-over-year % change

Same outlays data, now colored quantitatively with a diverging Cool–Warm palette: blue = spent less than FY2023, red = spent more. A crisp example of how swapping Color between a categorical and a numeric column transforms the story the treemap tells.