# Per-Book Offline Bible Assets

Source: https://github.com/sajeevavahini/bibles — Zefania XML `World English Bible.xml` ~4.8MB, format `XMLBIBLE/BIBLEBOOK/CHAPTER/VERS` attrs `bnumber/bname/cnumber/vnumber`.

## Problem
- 4.8MB XML → 3.8MB blob + 304KB idx monolith = too big for RAM, packaging bloats `.app` to 4.2MB monolith load.
- Device has ~8MB PSRAM but LVGL/lwIP compete; loading whole bible at once OOM-prone.

## Solution: per-book split

Output `assets/bible/`:
- `books.json`: `[{bnumber, bname, verses, chapters}, ...]` 66 entries
- `<bnum>.bin` (01.bin..66.bin): concatenated NUL-terminated UTF-8 verse strings
- `<bnum>.idx`: binary: header 8 bytes: magic `BIBK` ascii, version u16 (1), count u32; then per entry 8 bytes: `offset u32 LE, cnum u16 LE, vnum u16 LE` sorted by chapter/verse.

Worst book: Psalms 19.bin ~220KB, still fits. Total ~4.1MB split but loaded one at a time → RAM tiny.

## Converter

`Apps/BibleVerse/tools/convert_bible.py` (also template):

Usage:
```bash
python3 tools/convert_bible.py /path/to/World\ English\ Bible.xml assets/bible/
# or fetch then convert
```

Implementation sketch (see templates/convert_bible.py):
- xml.etree.ElementTree iterative parse `BIBLEBOOK` to keep memory low.
- Accumulate bin buffer + idx entries per book.
- Write books.json last.

## Hybrid resolution on device

```c
// 1. Try embedded A: assets via tt_app_get_assets_path
// 2. Fallback scan /sdcard/bibles/WEB/ same binary format for user-provided translations
// 3. build target: load BookInfo array, then on demand load_book(bnumber) frees previous.
```

## Verify packaging

```bash
tar tvf build/BibleVerse.app | grep "\.bin" | wc -l  # expect 66
ls -lh assets/bible/*.bin | sort -k5 -hr | head
```