Soup Sieve is the official CSS selector library in Beautiful Soup 4.7+, and with this change, Soup Sieve introduces a number of changes that break some of the expected behaviors that existed in versions prior to 4.7.
In short, Soup Sieve follows the CSS specifications fairly close, and this broke a number of non-standard behaviors. These non-standard behaviors were not allowed according to the CSS specifications. Soup Sieve has no intentions of bringing back these behaviors.
For more details on specific changes, and the reasoning why a specific change is considered a good change, or simply a feature that Soup Sieve cannot/will not support, see Beautiful Soup Differences.
In web browsers, CSS selectors do not usually select content inside an
iframe element if the selector is called on an element outside of the
iframe. Each HTML document is usually encapsulated and CSS selector leakage across this
iframe boundary is usually prevented.
In it's current iteration, Soup Sieve is not aware of the origin of the documents in the
iframe, and Soup Sieve will not prevent selectors from crossing these boundaries. Soup Sieve is not used to style documents, but to scrape documents. For this reason, it seems to be more helpful to allow selector combinators to cross these boundaries.
Soup Sieve isn't entirely unaware of
iframe elements though. In Soup Sieve 1.9.1, it was noticed that some pseudo-classes behaved in unexpected ways without awareness to
iframes, this was fixed in 1.9.1. Pseudo-classes such as
:contains() where given awareness of
iframes to ensure they behaved properly and returned the expected elements. This doesn't mean that
select won't return elements in
iframes, but it won't allow something like
:default to select a
button in an
iframe whose parent
form is outside the
iframe. Or better put, a default
button will be evaluated in the context of the document it is in.
With all of this said, if your selectors have issues with
iframes, it is most likely because
iframes are handled differently by different parsers.
html.parser will usually parse
iframe elements as it sees them.
lxml parser will often remove
body tags of an
iframe HTML document.
lxml-xml will simply ignore the content in a XHTML document. And
html5lib will HTML escape the content of an
iframe making traversal impossible.
In short, Soup Sieve will return elements from all documents, even
iframes. But certain pseudo-classes may take into consideration the context of the document they are in. But even with all of this, a parser's handling of
iframes may make handling its content difficult if it doesn't parse it as HTML elements, or augments its structure.