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

WordPress大学
WordPress大学
A
About on SuperTechFans
小众软件
小众软件
Hugging Face - Blog
Hugging Face - Blog
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
博客园 - 叶小钗
博客园 - 聂微东
博客园 - Franky
Apple Machine Learning Research
Apple Machine Learning Research
罗磊的独立博客
量子位
博客园 - 三生石上(FineUI控件)
Recent Announcements
Recent Announcements
The GitHub Blog
The GitHub Blog
B
Blog RSS Feed
T
The Blog of Author Tim Ferriss
GbyAI
GbyAI
云风的 BLOG
云风的 BLOG
Last Week in AI
Last Week in AI
宝玉的分享
宝玉的分享
B
Blog
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
Stack Overflow Blog
Stack Overflow Blog
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant
I Reverse Engineered a Closed-Format App. Everything Was ...
Antonio Zhu · 2026-06-27 · via DEV Community

Antonio Zhu

I had accumulated over two thousand notes in Youdao Cloud Note over several years. When I decided to move to Obsidian, the first thing I checked was the export feature. There wasn't one. No batch export, no single-note export, nothing in the Mac client.

So I went looking for the local data.


Where The Data Was

On macOS, Youdao Cloud Note stores its data here:

~/Library/Containers/ynote-desktop/Data/Library/Application Support/ynote-desktop/

Inside that directory, organized by account email, were three SQLite databases:

<account>.db         # note metadata, folder hierarchy
<account>-content.db # note content (old editor)
<account>-search.db  # search index

And a file/ directory with 16 subdirectories arranged by the first character of each file ID, holding the new editor's local files. Everything was unencrypted.


What The Database Contained

The main database had two critical tables:

note — the note catalog. Key columns: fileId (UUID), title, parentId (folder reference), orgEditorType (0 for the new block editor, 1 for the old plain editor), entryPath (path to the local file), createTime (Unix timestamp in seconds), and deleted (NULL meant not deleted).

note_book — the folder tree. Folders were not in note at all. They lived in a separate table with their own fileId and parentId fields, forming a tree you could traverse with BFS.

The content database held a contenttable with a content field, but the field was truncated to around 150 characters — just enough for search snippets. The real content for old-editor notes lived here too, with variable lengths.

Where the content actually was depended on which editor wrote the note:

orgEditorType Editor Content source
0 New (block editor) entryPath → local file
1 Old (plain) contenttable.content

The local files came in two formats: JSON (449 of them) and plain Markdown (1,832).


The JSON Block Tree

The new editor stored notes as a block tree in JSON. Each block followed this structure, with the actual keys being cryptic integers:

"6" → block type (p=paragraph, h=heading, co=code, t=table, im=image, l=list, q=quote, hr=divider)
"4" → properties (heading level, code language, image URL, etc.)
"5" → array of child blocks
"7" → inline text segments with "9" format markers (b=bold, i=italic, li=link, il=inline code, etc.)

Walking the "5" array recursively converted the entire tree to Markdown. The core renderer was under 200 lines.


Three Things That Wasted My Time

The __compress__ flag. Ten JSON files had "__compress__": true. I spent an hour trying to uncompress them with LZString before realizing the "5" array was still just a plain array — not a compressed string. The flag had been set but never used.

Deleted is NULL, not 0. Every note with deleted IS NULL had been a living note. WHERE deleted = 0 returned nothing. This is standard SQL but easy to miss when you're scanning thousands of rows for anomalies.

The timestamp is in seconds. 1484115955 is January 11, 2017. If you treat it as milliseconds and divide by 1000, you land in 1970. I did that once.


A Dead-End That Wasn't

The contenttable.content field was consistently 150 characters. This looked like a deliberate truncation to defeat extraction. But the actual content was never supposed to be in that column — the new editor stored everything in local files referenced by entryPath. The content column was just a search index. Once I followed the entryPath trail, I had the full text of every note.


What The Script Does

The whole thing is a single Python file. It connects to the three SQLite databases, walks the folder tree, maps every note to its content source, converts JSON blocks to Markdown when needed, and writes the output organized by folder.

git clone https://github.com/chncaesar/youdao-to-obsidian.git
cd youdao-to-obsidian
pip3 install beautifulsoup4
python3 youdao_migrate.py

The script auto-detects your account and data directory. Output goes to ~/Desktop/obsidian/ by default. Each note gets a .md file with YAML frontmatter preserving the original title, creation date, and source ID.

I also bundled a Claude Code Skill in the repo — drop it into ~/.claude/skills/ and saying "export my Youdao notes" triggers the whole pipeline without remembering flags.


What Came Out

The final export: 2,285 notes across 253 folders. Mixed Chinese, English, and code content with no encoding issues. Tables, code blocks, images, lists, and blockquotes all converted correctly from the JSON block tree. Some notes were empty bodies with attachments only, handled gracefully.

Open ~/Desktop/obsidian as an Obsidian vault and everything is there.


Why This Worked

The most interesting part of this project was not the parser or the block converter.

It was the realization that a closed-format application with no export feature had been storing all my data in plain, unencrypted SQLite files, with a documented-enough block structure that could be reverse-engineered in an afternoon.

Two thousand notes. Years of writing. The application offered no way to take them out. They were never locked in. I just hadn't looked.

GitHub: github.com/chncaesar/youdao-to-obsidian