Enabling endless querying options

[funkyfuture]

atm delb faciliates querying trees with XPath and CSS selectors:

nodes = document.xpath("//foo")
nodes = document.css_select("foo")

my last night's sleepin idea was to change that to a common interface based on a new abstract class that represents any query expression and implements their evaluation. the previous would then read:

nodes = document.query(xpath("//foo"))
nodes = document.query(css_select("foo"))

that can easily be transitioned with deprecations.

the new interfaces would look somethng like this, existing implementations are referenced by an fqn:

class NodeBase:
    def query(self, query: Query) -> delb.QueryResults:
        return query.evaluate(self)

class Query(ABC):
    @abstractmethod
    def evaluate(self, node: delb.NodeBase) -> delb.QueryResults:
        pass

class XPath(Query):
    def __init__(self, expression: str, namespaces: delb.NamespaceDeclarations):
        self.ast = delb.xpath.parse(expression)
        self.namespaces = namespaces

    def evaluate(self, node: delb.NodeBase) -> delb.QueryResults:
        return self.ast.evaluate(node, self.namespaces)

def xpath(expression: str, namespaces: delb.NamespaceDeclarations) -> XPath:
    return Xpath(expression, namespaces)

okay, there's more typing hyping bloat, what may be beneficial? generally speaking, applications and libraries can use whatever they come up with directly on node and document instances (node.query(my_dsl(…)) instead of my_dsl(node, …)). hypothetical ideas for more concrete uses:

• efficient evaluations for simple lookup grammars (e.g. just looking for local names)
• caching results when the application is working on immutable data (though practically it makes more sense to expose that cache somewhere where you don't necessarily have parsed a whole document)
• is it still a thing to come up with JSON based DSLs?
• well, when Query objects are re-used they don't have to parse their input again (atm they could miss the cache around delb.xml.parse)
• people can make their XPath 4.0 xpath implementation if they believe in it

that isn't entirely convincing yet.

[funkyfuture]

snakesist is already keeping specific queries around as objects:

https://github.com/delb-xml/snakesist/blob/8ebe9bdf7a75ce99bc98caa858c450786b986224/snakesist/exist_client.py#L64-L69