PHP-Parser
PHP-Parser copied to clipboard
nikic
Introducing PHPParser_Node_Stmt_Block
I'd like to get rid of the statement arrays and instead introduce a PHPParser_Node_Stmt_Block.
arrays might be a bit faster, but they make the code operating on the AST unnecessary complex by requiring constant type checking and two different APIs, e.g. traverse(Node $node) and traverseArray(array $nodes).
To reduce the BC break, we could implement ArrayAccess, and Countable on that node class.
What do you think?
The need for traversing arrays stays in any case, as arrays are used for a few more things than just statements. Simple example is the FuncCall node, which has an array of ->args. Another example is Echo which accepts several ->exprs. Etc.
But I do agree that adding a dedicated node for statements could be reasonable. It would allow to add a helper methods for handling statements. One such method for example could be ->getByType('Stmt_Method'). Do have ideas what else could be useful?
I need it for example to get the next/previous/parent relation correct, see #17.
However, I have discovered some other areas where I might want different nodes (names for example which are sometimes PHP_Node_Name, sometimes strings). So, I'm not sure how much we should change the AST in your library, or if the better approach is to just run a visitor over the AST which normalizes it into a format that I need in my code. Your current AST is better for performance at least.
Ah, okay, I have misunderstood your point at first, now I see it. I'll see whether I can add a node for statements without breaking too much stuff.
By the way, the reason that some names are Node_Name and others are just strings is that some names can be namespaced and others can't. For example when calling a function the function name can be namespaced, so it's a Node_Name. But when calling a method, the method name cannot be namespaced, so it's just a string.
By the way, the reason that some names are Node_Name and others are just strings is that some names can be namespaced and others can't. For example when calling a function the function name can be namespaced, so it's a Node_Name. But when calling a method, the method name cannot be namespaced, so it's just a string.
A string also has no offset information, which can be important in some cases. For example, when a user does go-to-definition for a method, I need to know whether he clicked on the method name or the variable the method was called on.
+1 for this, node navigation is very clumsy right now, ended up using modified relations from #17.
