# CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

## Project Overview

This is a single-file Python script (`imapdown.py`) that downloads all emails from an IMAP server into individual EML files, preserving the folder hierarchy. It uses only Python's standard library and has no external dependencies.

## Development Environment

- Python 3.6+ required
- Virtual environment is set up in `.venv` - activate it before running:
  ```bash
  source .venv/bin/activate
  ```

## Running the Script

Basic usage (incremental mode - only downloads new emails):
```bash
./imapdown.py --server imap.example.com --email user@example.com --user user@example.com --password "password" --ssl
```

Full download (ignores previous state, requires empty target directory):
```bash
./imapdown.py --server imap.example.com --email user@example.com --user user@example.com --password "password" --ssl --full
```

Testing/debugging with limited emails:
```bash
./imapdown.py --server imap.example.com --email user@example.com --user user@example.com --password "password" --ssl --limit 10
```

Custom storage directory:
```bash
./imapdown.py --server imap.example.com --email user@example.com --user user@example.com --password "password" --ssl --store /path/to/backup
```

## Architecture

### Single-File Design
The entire application is contained in `imapdown.py` (13KB). This is intentional - no modules or packages.

### State Tracking
- The script maintains a `.imapdown_state.json` file in each email account's download folder
- Tracks the highest UID (unique identifier) downloaded per IMAP folder
- Format: `{"INBOX": 19334, "INBOX.Archive": 1770, "Sent": 892}`
- Enables efficient incremental downloads (default mode)

### Download Flow
1. Parse arguments
2. Connect to IMAP server (SSL, STARTTLS, or plain)
3. List all folders and decode modified UTF-7 folder names
4. For each folder:
   - Load last downloaded UID from state file (if incremental mode)
   - Search for new messages (UID > last_uid)
   - Download each message as RFC822
   - Save as `.eml` file with naming: `{UID}_{date}_{subject}.eml`
   - Extract attachments into `.zip` file (same base name)
   - Update state with highest UID
5. Save state file

### Key Implementation Details

**Modified UTF-7 Decoding**: IMAP folder names use modified UTF-7 encoding (see `decode_modified_utf7()` at line 39). This is not standard base64 - it uses `,` instead of `/` and has special `&` handling.

**Filename Sanitization**: Two-stage process:
- `sanitize_filename()`: Removes invalid filesystem characters, max 50 chars for subjects
- `sanitize_folder_path()`: Converts IMAP folder separators (`.` or `/`) to OS path separators

**UID-Based Incremental Updates**: Uses IMAP UIDs (not sequence numbers) because UIDs are persistent. The search `UID {last_uid + 1}:*` fetches only new messages. Some servers return the highest UID even when searching for higher UIDs, so there's additional filtering at line 251.

**Full Mode Safety**: `--full` mode checks if the download folder already contains `.eml` files and refuses to run (line 325). This prevents accidental duplicates. Users must delete the folder first.

**Attachment Handling**:
- Walks message parts looking for `Content-Disposition: attachment` or `inline`
- Handles duplicate attachment filenames by appending `_{counter}`
- All attachments for one email go into a single `.zip` file

## Output Structure

```
{store_dir}/           # default: ./download
└── {email_address}/   # sanitized email address
    ├── .imapdown_state.json
    ├── INBOX/
    │   ├── 123_20240115_Meeting_notes.eml
    │   └── 124_20240116_Report.zip
    └── Sent/
        └── 456_20240114_RE_Question.eml
```

## Testing

No formal test suite exists. Manual testing approach:
- Use `--limit 10` to download a small batch for verification
- Test SSL vs STARTTLS connections
- Test incremental mode by running twice
- Verify `.eml` files open correctly in email clients
- Check that folders with special characters (non-ASCII) are handled correctly