Keep track of canonical URLs from page markup in `source_metadata`

[Mr0grog]

Some pages have a <link rel="canonical" href="{url}"> element in their markup, indicating a correct, “canonical” URL for the page (some more info here: https://en.wikipedia.org/wiki/Canonical_link_element). When importing data from the Wayback Machine, it would be great to include the canonical URL in the source_metadata field.

We already parse HTML pages that we’re importing to get their titles, and we could get the canonical link (if present) at a similar point in the process. Ideally we should create a way to only parse the page once to get title, canonical link, and anything else we might want to extract from the page content in the future.

Where we already parse markup for titles:

web-monitoring-processing/web_monitoring/cli/cli.py

Lines 408 to 413 in 21512eb

	title = ''
	if media_type in HTML_MEDIA_TYPES:
	encoding = detect_encoding(memento.content, memento.headers)
	title = utils.extract_title(memento.content, encoding)
	elif media_type in PDF_MEDIA_TYPES or memento.content.startswith(b'%PDF-'):
	title = utils.extract_pdf_title(memento.content) or title

web-monitoring-processing/web_monitoring/utils.py

Lines 98 to 112 in 21512eb

	def extract_title(content_bytes, encoding='utf-8'):
	"Return content of <title> tag as string. On failure return empty string."
	content_str = content_bytes.decode(encoding=encoding, errors='ignore')
	# The parser expects a file-like, so we mock one.
	content_as_file = io.StringIO(content_str)
	try:
	title = lxml.html.parse(content_as_file).find(".//title")
	except Exception:
	return ''
	if title is None or title.text is None:
	return ''
	# In HTML, all consecutive whitespace (including line breaks) collapses
	return WHITESPACE_PATTERN.sub(' ', title.text.strip())

[Mr0grog]

It might be useful to do this for <link rel="shortlink">, too. I had totally forgotten about that one!