cadrumo.adapters.outbound.aeat._html module

The one place this package names an HTML parser backend.

Every AEAT page read – sede status readers, the representation gate, the cotejo viewer probe, the Clave page flows – builds its document through parse_html(), so the backend is one decision rather than one per call site. BeautifulSoup raises FeatureNotFound rather than falling back when a named backend is absent, which makes the choice a live-failure axis the moment the dependency picture shifts: scattered across call sites, that failure arrives at whichever page a taxpayer happened to read first.

The backend is lxml, and the reasons are measured rather than preferred. pyproject.toml declares beautifulsoup4[lxml] as a direct dependency precisely because this package names the backend, so moving to the stdlib parser would silently invalidate a declaration the packaging smoke test also asserts. lxml is the more forgiving parser for the legacy markup AEAT actually serves. And parsing all 75 bundled AEAT-origin pages under both backends and comparing the structural signals these readers consume – table, row, cell, anchor, href, heading, input, form, select, option, span and div counts, plus extracted text length – diverged on none of them, so the standardisation costs nothing observable to a reader.

Callers pass markup and nothing else. Refusal stays with the caller: each read path raises its own typed error with its own operator-facing message, and a constructor that guessed at one would flatten that back into a generic parse failure.

parse_html(markup)

Parse markup into a document tree using this package’s parser backend.

Parameters:

markup (str) – Raw HTML body of an AEAT page.

Return type:

BeautifulSoup

Returns:

The parsed BeautifulSoup document.

Raises:

bs4.FeatureNotFound – When the declared backend is not installed. That is a broken environment rather than a bad page, so it is left to propagate rather than folded into a per-caller parse error.