Documentation about file_obj and url_fetcher

[SebCorbin]

Seeing

WeasyPrint/weasyprint/urls.py

Lines 230 to 231 in e5f7ec9

	If a ``file_obj`` key is given, it is the caller’s responsibility
	to call ``file_obj.close()``.

and

WeasyPrint/weasyprint/urls.py

Lines 272 to 293 in e5f7ec9

	@contextlib.contextmanager
	def fetch(url_fetcher, url):
	"""Call an url_fetcher, fill in optional data, and clean up."""
	try:
	result = url_fetcher(url)
	except Exception as exc:
	raise URLFetchingError('%s: %s' % (type(exc).__name__, str(exc)))
	result.setdefault('redirected_url', url)
	result.setdefault('mime_type', None)
	if 'file_obj' in result:
	try:
	yield result
	finally:
	try:
	result['file_obj'].close()
	except Exception:
	# May already be closed or something.
	# This is just cleanup anyway: log but make it non-fatal.
	LOGGER.warning('Error when closing stream for %s:\n%s',
	url, traceback.format_exc())
	else:
	yield result

I'm guessing the documentation is either outdated or wrong, is that right?

[liZe]

is that right?

It is, thanks!

[SebCorbin]

In a way, it's not, because fetch() is actually the caller and is closing file_obj, but updating the special section about URL fetchers in the docs to explain that would be nice.

[liZe]

Yes, I'm trying to find the exact words to say this.

[liZe]

If a file_obj key is given, it is the caller’s responsibility to call file_obj.close(). The default function used internally to fetch data in WeasyPrint tries to close file_obj after retreiving; but if this URL fetcher is used elsewhere, the file object has to be closed manually.

Is that OK for you?