feat(publisher): Introduce V1_1 for MexicoNewsDaily - #970
Conversation
MaxDall
left a comment
MaxDall
left a comment
There was a problem hiding this comment.
REQUEST_CHANGES — the new V1_1 carries V1's p[text()] predicate over, and it silently drops <em>-wrapped article prose.
mx.MexicoNewsDaily: read 10 of 100 scanned (6 flagged, 6 in draw); static read (VALID_UNTIL, version scope, image caption/author/cover pairing), Tier-1 coherence and a structural sweep of all 10 checked. 1 blocker inline.
Not inline:
- Nit — the whole 100-article draw came out of the 2018 end of
post-sitemap(all ten cached articles publish May–Jun 2018), so neither this review nor the newMexicoNewsDaily_2026_08_18fixture — itself a 2018 article — exercises a current, possibly paywalled page.Sitemap(..., reverse=True)inmx/__init__.pywould make crawls start from the newest posts and keep future reviews on the live layout.
| ) | ||
|
|
||
| class V1_1(V1): | ||
| _paragraph_selector = XPath("//div[@class='tts_content_wrapper_1']/p[text()]") |
There was a problem hiding this comment.
Blocker — p[text()] drops any <p> whose text sits inside <em>/<a>, so real prose is lost. “…UPDATE June 20, 4:16pm CDT: FIFA announced a fine of 10,000 Swiss francs (US $10,000) against the Mexican Football Federation…” never reaches the body [1]; a CORRECTION: paragraph goes the same way [2], and all 100 pooled articles show an uncaptured <p> (mostly the Source: … trailer). Fix: /p[normalize-space()], plus a bloat filter for the Source:/subscribe trailers, which would otherwise leak in.
…nclude `<strong>` elements
2 participants
No description provided.