Limitations

Hi, Marcus here. I’m Analog’s founder, and I work on structured extraction every day. These are the gaps I know about:

• Complex HTML tables. We’ve put a lot of effort into high-fidelity table capture, but complex tables can still extract poorly. The chest-to-belt throughput table on the Factorio Wiki is a good example.
• Long social feeds and walls. Only part of a very long feed may be captured; the result says how far it got.
• Deep discussion threads. Tables inside replies can sometimes become garbled.
• Restaurant menus. Menus that change structure from section to section can produce garbled or oddly split output.
• Multi-page sweeps (--pages). This is still experimental. Pages pool only where their sections share a template; fields only some pages carry are null elsewhere and named in the sweep note, and sections that don’t line up stay separate.

Expected boundaries

JavaScript-heavy pages are not a limitation by themselves: the built-in browser renders them by default. A site that blocks automated visitors is different from a weak extraction; Analog reports the block as the result and stops. See Fetching.

Report a result

A quality report is useful whether or not the page type appears above: the Feedback guide explains the symptom labels and submission routes.