Conversation
 |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
| const authors = (ld.author || []) | ||
| .map(a => (a.name || '').trim()) |
There was a problem hiding this comment.
If a book page supplies author as one object instead of an array, parseNovel calls .map on it and throws. The reader cannot open that book. Check the value’s shape before mapping it.
There was a problem hiding this comment.
Fixed in 855c6e0. The JSON-LD author value is now normalized before names are read: an array, a single Person object, a plain string, or a missing/null value all work. I tested each shape (plus a mixed array) with synthetic JSON-LD, and a live novel still parses its authors.
| const first = await getApi<ChapterContentData>(base); | ||
| const pages = [first]; | ||
| for (let p = 2; p <= first.pageMetas.length; p++) { | ||
| pages.push(await getApi<ChapterContentData>(base + '?weightPage=' + p)); |
There was a problem hiding this comment.
The routine plugin check reads one sample chapter but does not check a fixed two-page response. A later change to pageMetas or weightPage could drop the rest of a long chapter while that check still passes. Add a test that verifies text from both pages.
Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
There was a problem hiding this comment.
Not adding a test here: this repo has no unit-test harness for plugins, and building one is out of scope for this PR. I checked the two-page path against the live site instead: book/shadow-slave-r2k2ivbd6ez4/gpp58s44jc is split over two weight pages (50 + 53 blocks), and parseChapter returned all 103 paragraphs.
This branch has not been deployed
1 participant
Adds a standalone English plugin for ReChapters.
Closes #2485
What it supports
The site is a Next.js app with a public JSON API, so the plugin uses that API in most places instead of scraping HTML:
/api/search, 20 per page. It pages by building the site's offset cursor. "Latest" sorts by last chapter update. Filters: sort (popularity, latest update, rating, rating count, chapter count, word count) and status (ongoing / completed)./api/search?q=…sorted by relevance, with the same paging.authorfield is accepted as an array, a single object or a string); genres come from the tag list and status from the status pill. The complete chapter list is fetched from/api/book/{id}/chapters/buckets, then each bucket of about 100 chapters is fetched in ascending order, 5 buckets at a time./api/book/…/content(no login needed for these chapters). Long chapters are split into several pages, and all of them are fetched. The text blocks are plain text, and they are HTML-escaped into<p>elements, so the output can't contain markup,on*attributes orjavascript:URLs.The site has wallet and purchase endpoints. Every chapter I sampled across about 25 popular novels was free (
accessLevel: 0,isVip: false), so the plugin has no special handling for paid chapters.How it was tested
npm run check:plugin -- plugins/english/rechapters.ts: all four checks PASS (popular: 20 novels, search: 20 results, parseNovel: 1432 chapters, parseChapter: 10255 chars).[];authoris an array, a single object, a string, a mixed array, or missing/null.eslintandprettier --checkpass on the new file, andnpm run build:compilesucceeds. Repo-widenpm run lintstill reports errors, but they are in files this PR doesn't change (plugins/arabic/rewayatfans.ts,plugins/english/wetriedtls.ts).This PR was written by an AI agent (Claude Code), and no human has reviewed it yet.