MOJi辞書 Collection Exporter

Personal backup/export tools for your MOJi辞書 (mojidict.com) 收藏 / Collection items.

This project reverse-engineers the MOJi web client’s network calls and uses the same endpoints to export:

• Saved words (targetType=102)
• Saved sentence-like items (targetType=103/120)

Security note: treat your sessionToken like a password. Don’t share it or commit it.

What’s in this repo

• moji_exporter.py: main exporter (no third-party deps)
• reversed_api_ref.md: working API notes (Parse CF + REST) and observed quirks
• run_moji_from_launch.ps1: helper that loads env vars from .vscode/launch.json and runs the exporter
• saved_webpage/: offline snapshot used for reverse engineering

Requirements

• Windows PowerShell (examples below)
• Python 3 (recommended: 3.10+)

No extra pip packages are required.

Getting credentials (from the browser)

While logged in on https://www.mojidict.com:

• MOJI_SESSION_TOKEN (Parse sessionToken)
  • Local Storage key: Parse/<APP_ID>/currentUser → JSON field sessionToken
• MOJI_INSTALLATION_ID
  • Local Storage key: Parse/<APP_ID>/installationId
• MOJI_DEVICE_ID
  • Local Storage key: MOJi-PC-DeviceID
• MOJI_ROOT_FOLDER_ID (often)
  • Looks like: ROOT#com.mojitec.mojidict#zh-CN_ja

See reversed_api_ref.md for exact DevTools snippets.

Quick start (PowerShell)

Set env vars and run:

$env:MOJI_SESSION_TOKEN = '<SESSION_TOKEN>'
$env:MOJI_INSTALLATION_ID = '<INSTALLATION_ID>'
$env:MOJI_DEVICE_ID = '<DEVICE_ID>'
$env:MOJI_ROOT_FOLDER_ID = 'ROOT#com.mojitec.mojidict#zh-CN_ja'

python .\moji_exporter.py --help

Export words (text)

python .\moji_exporter.py --all --mode words --count 50 --output all_words.txt

Export sentences (text)

python .\moji_exporter.py --all --mode sentences --count 50 --output all_sentences.txt

Important server behavior (paging clamp)

MOJi’s folder-fetchContentWithRelatives appears to clamp deep paging:

• For a given (fid, sortType, targetTypes) combination, requests with pageIndex > 20 repeat page 20.
• With count=50, that’s an effective cap of ~1000 items per “view”.

Workaround: union across sort types and/or folders

The exporter supports unioning multiple partitions and de-duplicating globally:

• --all-folders: union across all folders returned by fetchMyFolders
• --sort-types: union across multiple sort orders
  • Supports ranges: 0..200 or 0-200

Example:

python .\moji_exporter.py --all --mode words --count 50 --all-folders --sort-types 0,1,2,3,4,5,6,10,100 --output all_words_union.txt

Stop early once you reached your goal

If you know roughly how many items you expect, you can stop once that many unique items have been printed:

python .\moji_exporter.py --all --mode words --count 50 --all-folders --sort-types 0..200 --expected 4200 --output all_words.txt

JSON progress output (rewrite-on-iteration)

--json makes the exporter store results in an in-memory dict and rewrite the JSON file after each page iteration. This is useful for long runs where you want to inspect partial progress.

python .\moji_exporter.py --all --mode words --count 50 --all-folders --sort-types 0..200 --expected 4200 --json --output progress_words.json

The output JSON contains:

• meta: progress info (last folder/sort/page, counts)
• itemsById: dictionary keyed by the exporter’s dedupe key (e.g. 102:<id>, 103:<id>)

Using the launcher script

If you’ve stored the env vars inside .vscode/launch.json, you can run with:

.\run_moji_from_launch.ps1 --all --mode words --count 50 --output all_words.txt

Troubleshooting

• 403 / blocked requests: the exporter already sends browser-like headers, but tokens/IDs must be correct.
• Empty results for words: ensure --mode words (or --target-types 102). The server may default to sentence-like items if targetTypes isn’t sent.
• Excel/Notepad encoding: text output uses UTF-8 with BOM when --output is used.

Disclaimer

These scripts are for personal backup/interoperability. The APIs are reverse-engineered and may change or stop working at any time.