dom: implement HTMLCollection

We can't simply use the libdom dom_document_get_elements_by_tag_name
because it follows an old version of the specifications and returns a
NodeList.
Since
190700b7c1
the spec changed in favor of returning an HTMLCollection.

So I'm trying to re-implement the HTMLCollection in zig.

src/dom/document.zig

@@ -7,6 +7,7 @@ const Case = jsruntime.test_utils.Case;
 const checkCases = jsruntime.test_utils.checkCases;
 
 const Node = @import("node.zig").Node;
+const HTMLCollection = @import("html_collection.zig").HTMLCollection;
 
 const Element = @import("element.zig").Element;
 const ElementUnion = @import("element.zig").Union;
@@ -34,6 +35,23 @@ pub const Document = struct {
         const e = parser.documentCreateElement(self, tag_name);
         return Element.toInterface(e);
     }
+
+    // We can't simply use libdom dom_document_get_elements_by_tag_name here.
+    // Indeed, netsurf implemented a previous dom spec when
+    // getElementsByTagName returned a NodeList.
+    // But since
+    // https://github.com/whatwg/dom/commit/190700b7c12ecfd3b5ebdb359ab1d6ea9cbf7749
+    // the spec changed to return an HTMLCollection instead.
+    // That's why we reimplemented getElementsByTagName by using an
+    // HTMLCollection in zig here.
+    pub fn _getElementsByTagName(self: *parser.Document, tag_name: []const u8) HTMLCollection {
+        const root = parser.documentGetDocumentNode(self);
+        return HTMLCollection{
+            .root = root,
+            // TODO handle case insensitive comparison.
+            .match = tag_name,
+        };
+    }
 };
 
 // Tests
@@ -58,6 +76,19 @@ pub fn testExecFn(
     };
     try checkCases(js_env, &getElementById);
 
+    var getElementsByTagName = [_]Case{
+        .{ .src = "let getElementsByTagName = document.getElementsByTagName('P')", .ex = "undefined" },
+        .{ .src = "getElementsByTagName.length", .ex = "2" },
+        .{ .src = "getElementsByTagName.item(0).localName", .ex = "p" },
+        .{ .src = "getElementsByTagName.item(1).localName", .ex = "p" },
+        .{ .src = "let getElementsByTagNameAll = document.getElementsByTagName('*')", .ex = "undefined" },
+        .{ .src = "getElementsByTagNameAll.length", .ex = "8" },
+        .{ .src = "getElementsByTagNameAll.item(0).localName", .ex = "html" },
+        .{ .src = "getElementsByTagNameAll.item(1).localName", .ex = "head" },
+        .{ .src = "getElementsByTagNameAll.item(2).localName", .ex = "body" },
+    };
+    try checkCases(js_env, &getElementsByTagName);
+
     const tags = comptime parser.Tag.all();
     comptime var createElements: [(tags.len) * 2]Case = undefined;
     inline for (tags, 0..) |tag, i| {

src/dom/html_collection.zig

@@ -7,22 +7,122 @@ const jsruntime = @import("jsruntime");
 const Element = @import("element.zig").Element;
 
 // WEB IDL https://dom.spec.whatwg.org/#htmlcollection
+// HTMLCollection is re implemented in zig here because libdom
+// dom_html_collection expects a comparison function callback as arguement.
+// But we wanted a dynamically comparison here, according to the match tagname.
 pub const HTMLCollection = struct {
-    pub const Self = parser.HTMLCollection;
     pub const mem_guarantied = true;
 
-    // JS funcs
-    // --------
+    root: *parser.Node,
+    // match is used to select node against their name.
+    // match comparison is case sensitive.
+    match: []const u8,
 
-    pub fn _get_length(self: *parser.HTMLCollection) u32 {
-        return parser.HTMLCollectionLength(self);
+    /// _get_length computes the collection's length dynamically according to
+    /// the current root structure.
+    // TODO: nodes retrieved must be de-referenced.
+    pub fn get_length(self: *HTMLCollection) u32 {
+        var len: u32 = 0;
+        var node: ?*parser.Node = self.root;
+        var ntype: parser.NodeType = undefined;
+
+        var is_wildcard = std.mem.eql(u8, self.match, "*");
+
+        while (node != null) {
+            ntype = parser.nodeType(node.?);
+            if (ntype == .element) {
+                if (is_wildcard or std.mem.eql(u8, self.match, parser.nodeName(node.?))) {
+                    len += 1;
+                }
+            }
+
+            // Iterate hover the DOM tree.
+            var next = parser.nodeFirstChild(node.?);
+            if (next != null) {
+                node = next;
+                continue;
+            }
+
+            next = parser.nodeNextSibling(node.?);
+            if (next != null) {
+                node = next;
+                continue;
+            }
+
+            var parent = parser.nodeParentNode(node.?);
+            var lastchild = parser.nodeLastChild(parent.?);
+            while (node.? != self.root and node.? == lastchild) {
+                node = parent;
+                parent = parser.nodeParentNode(node.?);
+                lastchild = parser.nodeLastChild(parent.?);
+            }
+
+            if (node.? == self.root) {
+                node = null;
+                continue;
+            }
+
+            node = parser.nodeNextSibling(node.?);
+        }
+
+        return len;
     }
 
-    pub fn _item(self: *parser.HTMLCollection, index: u32) ?*parser.Element {
-        return parser.HTMLCollectionItem(self, index);
+    pub fn _item(self: *HTMLCollection, index: u32) ?*parser.Element {
+        var len: u32 = 0;
+        var node: ?*parser.Node = self.root;
+        var ntype: parser.NodeType = undefined;
+
+        var is_wildcard = std.mem.eql(u8, self.match, "*");
+
+        while (node != null) {
+            ntype = parser.nodeType(node.?);
+            if (ntype == .element) {
+                if (is_wildcard or std.mem.eql(u8, self.match, parser.nodeName(node.?))) {
+                    len += 1;
+
+                    // check if we found the searched element.
+                    if (len == index + 1) {
+                        return @as(*parser.Element, @ptrCast(node));
+                    }
+                }
+            }
+
+            // Iterate hover the DOM tree.
+            var next = parser.nodeFirstChild(node.?);
+            if (next != null) {
+                node = next;
+                continue;
+            }
+
+            next = parser.nodeNextSibling(node.?);
+            if (next != null) {
+                node = next;
+                continue;
+            }
+
+            var parent = parser.nodeParentNode(node.?);
+            var lastchild = parser.nodeLastChild(parent.?);
+            while (node.? != self.root and node.? == lastchild) {
+                node = parent;
+                parent = parser.nodeParentNode(node.?);
+                lastchild = parser.nodeLastChild(parent.?);
+            }
+
+            if (node.? == self.root) {
+                node = null;
+                continue;
+            }
+
+            node = parser.nodeNextSibling(node.?);
+        }
+
+        return null;
     }
 
-    pub fn _namedItem(self: *parser.HTMLCollection, name: []const u8) ?*parser.Element {
-        return parser.HTMLCollectionNamedItem(self, name);
+    pub fn _namedItem(self: *HTMLCollection, name: []const u8) ?*parser.Element {
+        _ = name;
+        _ = self;
+        return null;
     }
 };

src/netsurf.zig

@@ -765,6 +765,18 @@ pub inline fn documentGetElementsByTagName(doc: *Document, tagname: []const u8)
     return nlist.?;
 }
 
+// documentGetDocumentElement returns the root document element.
+pub inline fn documentGetDocumentElement(doc: *Document) *Element {
+    var elem: ?*Element = undefined;
+    _ = documentVtable(doc).dom_document_get_document_element.?(doc, &elem);
+    return elem.?;
+}
+
+pub inline fn documentGetDocumentNode(doc: *Document) *Node {
+    const res = documentGetDocumentElement(doc);
+    return @as(*Node, @ptrCast(res));
+}
+
 pub inline fn documentCreateElement(doc: *Document, tag_name: []const u8) *Element {
     var elem: ?*Element = undefined;
     _ = documentVtable(doc).dom_document_create_element.?(doc, stringFromData(tag_name), &elem);