fix: detect JSON-LD injected by Next.js App Router __next_s SSR mechanism

[cyanistes]

Problem

Next.js App Router SSR pages inject JSON-LD via a JavaScript push call rather than a static <script type="application/ld+json"> tag:

<script>
(self.__next_s=self.__next_s||[]).push([0,
    {"type":"application/ld+json","async":true,"children":"{...recipe JSON...}"}
])
</script>

extruct (which recipe-scrapers uses for JSON-LD extraction) only finds static script tags, so it misses this markup entirely. The result is that SchemaOrg finds no recipe data and scraping fails silently on any site built with Next.js App Router that uses structured Recipe markup. meny.no is one confirmed affected site.

Fix

Add SchemaOrg._normalize_nextjs_jsonld(html) — a static pre-processing step called at the top of __init__ that detects the __next_s.push() pattern and lifts the embedded children JSON-LD string into a proper <script type="application/ld+json"> tag before the HTML reaches extruct.

Fast path

Two cheap in checks ("__next_s" not in html and "application/ld+json" not in html) return the original HTML unchanged on all non-Next.js pages, so there is zero overhead for the vast majority of requests.

No new dependencies

json and re are stdlib modules.

Before / after (meny.no example)

Before — SchemaOrg finds no Recipe data → scraping raises SchemaOrgException.

After — the injected JSON-LD is promoted to a static script tag before extruct runs → title, ingredients, and instructions are parsed correctly.

Changes

• recipe_scrapers/_schemaorg.py — add import json, import re; add _normalize_nextjs_jsonld() static method; call it as the first line of __init__
• tests/library/test_schemaorg.py — add test_nextjs_ssr_jsonld to TestSchemaOrg

References

• Next.js App Router injects script data via __next_s: https://github.com/vercel/next.js/search?q=__next_s

[hhursev]

That's a good one! Thanks