catalan-adobe / acom-franklin-import-sections-mapping Goto Github PK

Original page: https://business.adobe.com/products/primetime/adobe-media-server.html

The original page contains invisible sections with content which ends up in the docx generated via the generic import script.

This issue lists Renovate updates and detected dependencies. Read the Dependency Dashboard docs to learn more.

Open

These updates have all been created already. Click a checkbox below to force a retry/rebase of any.

• fix(deps): update dependency @adobe/helix-importer to v3.3.5
• chore(deps): update external major (major) (@semantic-release/github, @semantic-release/npm, eslint, semantic-release)

Detected dependencies

• Check this box to trigger a request for Renovate to run again on this repository

Issue

Example: https://business.adobe.com/products/experience-manager/managed-services.html

Investigation

Current section detection algorithm exclude sections with x > 0 and width < ~.98 pageWidth.

Solution

With the NSiblingsElements approach used to find the page sections, we do not need such strict checks on x position and page width.

Notice the narrower section is now detected and extracted.

Issue

acom-sections-mapping prepare is not extracting some sections on some pages.

How to reproduce

Example page: https://business.adobe.com/products/adobe-experience-cloud-products.html

Run

echo 'https://business.adobe.com/products/adobe-experience-cloud-products.html' | node index.js prepare -i

Then
=> In sections-mapping/data, find the screenshots for that page, notice the bottom red marquee is missing

Investigation

business.adobe.com has akamai layer that checks for pages being loaded in "automation", in our case a headless browser. The function we use to launch the headless browser already implements basic mechanism to "hide" its headless state (like sanitizing the user-agent) but it's not enough.
By sniffing the network requests, we can see 403 responses on extra HTML resources fetched from the page (like plain HTML from fragments).

Solution

Until we find ways to not be detected as headless, we can report back in the output when we capture such issue + notify to re-run the command in non headless mode, using --no-headless.