Using the Parse Tree

Using the Parse Tree

Pegged parse trees are described here: Parse Trees.

As you could see, there are a deliberately simple structure. There are not multiple types to encode AST nodes and whatnots. Since D has wonderful compile-time evaluation capabilities and code can be mixed-in, Pegged has all the power it needs:

Since parse trees are structs, they are easy to use at compile-time (classes are a bit trickier for now, so I chose structs): you can iterate on them, search for specific nodes, delete nodes or simplify them, etc.

See Semantic Actions, User-Defined Parsers and Compile-Time Parsing for more on this.

An Output contains a ParseTree member called parseTree. Since Output uses alias this, the parse tree is directly accessible:

auto p = Expr.parse(input);
writeln(p.capture); // really p.parseTree.capture
foreach(ref child; p.children) // in truth, we are iterating on p.parseTree.children
    doSomething(child);

Since parse trees are run-of-the-mill N-ary trees, I didn't add tree-walking algorithms in Pegged, but I may do so to group useful tree functions with the related structure:

• searching for all nodes with a particular name
• getting the leaves (parse trees with no children)
• filtering a tree: cutting the nodes that do not obey a particular predicate.
• modifying a tree.

I'll see. You can find tree-related algorithms in another Github project of mine: https://github.com/PhilippeSigaud/dranges (see for example https://github.com/PhilippeSigaud/dranges/blob/master/treerange.d)