# epubgen

Python EPUB 3.3 generator for making test files. It will pull paragraphs from a text file randomly to generate the content of the book. It also has a torture mode which randomly adds unusual characters to try to break parsers.

## Usage

```sh
python epubgen.py -m normal -c 10 -p 5
```

| Option             | Default                        | Meaning                                           |
| ------------------ | ------------------------------ | ------------------------------------------------- |
| `-o, --out`        | `normal.epub` / `torture.epub` | output path (single mode) or name prefix (`both`) |
| `-m, --mode`       | `both`                         | `normal`, `torture`, or `both` (one of each)      |
| `-c, --chapters`   | `5`                            | number of chapters                                |
| `-p, --paragraphs` | `10`                           | paragraphs per chapter                            |
| `-t, --title`      | `The Testing EPUB`             | book title                                        |
| `-a, --author`     | `Test Author`                  | book author                                       |
| `--seed N`         | random                         | RNG seed for reproducibility                      |
| `--texts FILE`     | `paragraphs.txt`               | one paragraph per line                            |
| `--no-ncx`         | off                            | omit the EPUB 2 NCX fallback                      |
| `-v, --version`    |                                | print version                                     |

## What the generated EPUB contains

- `mimetype` — `application/epub+zip`, STORED, first entry, no extra fields
- `META-INF/container.xml`
- `OEBPS/content.opf` — EPUB 3.3 package: `urn:uuid` identifier, title + `title-type`
  refine, creator + `marc:relators` role refine, language, `dcterms:conformsTo`,
  publisher, date, `dcterms:modified` (required by spec), rights; manifest items with
  `properties="nav"`, `cover-image`, `svg`; NCX in the manifest for legacy readers
- `OEBPS/nav.xhtml` — `epub:type="toc"` (front matter / chapters / back matter groups)
  and `epub:type="landmarks"` guide
- `OEBPS/toc.ncx` — EPUB 2 NCX fallback so old readers work
- `OEBPS/cover.xhtml` + `OEBPS/cover.png` — SVG-wrapped procedural gradient cover
- `OEBPS/title.xhtml` — title page
- `OEBPS/chapNN.xhtml` — chapters with numbered paragraphs (anchors for link testing)
- `OEBPS/index.xhtml` — back-matter index with links into chapters
- `OEBPS/colophon.xhtml` — back matter, spine `linear="no"`
- `OEBPS/style.css` — stylesheet

## torture mode

A technically valid EPUB with syntax designed to trip up sloppy parsers:

- bidi override/embedding characters, combining marks, zero-width characters
- mixed RTL/LTR, CJK, emoji, private-use characters, exotic space variants
- numeric character references (`&#xNNNN;`) injected mid-text
- deeply nested `<span>`s, `<pre xml:space="preserve">`, tables, comments
- unusual-but-valid `xml:lang` (`zh-Hant-TW-u-ca-gregory`), `dir="rtl"` package
- chapter filenames with spaces and `&` (percent-encoded in every href)
- NCX with the DOCTYPE omitted and `playOrder` values out of document order
- CSS `@import`, media queries, vendor prefixes, `content` unicode escapes, `!important`