Add document querying #237

Kiyoshika · 2023-11-07T01:39:55Z

This implements a Query builder to pass to DocumentHandle.query() method to traverse the tree and return all node ID(s) that match the predicate. This also supports a FindFirst and FindAll search type. The searching of the tree terminates early on FindFirst.

I need to add some notes in the actual query.rs file explaining more of its usage, but given a Query instance, a node must match ALL conditions provided in the query (i.e., each condition is treated as an AND condition) in order to be returned.

I also want to add a simple extension DocumentHandle.query_multiple (or something along those lines) which would take a Vec<Query> list of predicates and check if a node matches at least one in order to be returned (to support OR-like conditionals.)

src/html5/parser/document.rs

Kiyoshika · 2023-11-09T23:25:54Z

I believe I have everything ready now for the first iteration of document querying. Squashed all commits during development into one.

emwalker · 2023-11-09T23:50:50Z

src/html5/parser/document.rs

+        let root_id = self.get().get_root().id;
+        node_stack.push(root_id);
+
+        while !node_stack.is_empty() {


while let Some(current_node_id) = node_stack.pop() { // ... }

implemented this suggestion, also moved self.get() outside the loop which probably has minimal overhead but it didn't really make sense to be inside anyways

emwalker · 2023-11-09T23:56:28Z

src/html5/parser/document.rs

+        let query = Query::new().contains_attribute("style").find_all();
+        let found_ids = doc.query(&query).unwrap();
+        assert_eq!(found_ids.len(), 3);
+        assert_eq!(found_ids, [div_id_2, div_id_3, p_id_4]);


Not for this PR: eventually we might be able to simplify these tests and omit the explicit tree construction:

let input = r#" <div> <div id="myid" style="somestyle"> <p title="hey"> <p> <div style="otherstyle" id="otherid"> <p> <p title="yo" style="cat"> "#; let doc = Document::parse_str(input).unwrap(); let query = Query::new().contains_attribute("style").find_all(); let found_ids = doc.query(&query).unwrap(); assert_eq!(found_ids, [div_id_2, div_id_3, p_id_4]); // ...

This would be nice. Manually constructing the document is pretty annoying but it’s what we have for now

src/html5/parser.rs

Kiyoshika commented Nov 7, 2023

View reviewed changes