This looks fine, but any reason for only a single predicate over an object with separate predicates?

Given the single-predicate option, I imagine almost all passed predicates will be of the form

function(name, node) {
  switch (name) {
    case ...:
      return ...;
    case ...:
      return ...;
    // .. and so on
  }
}

So it might be simpler to pass an object whose keys are the allowed class names.

My thinking was that we might want to manipulate the class name before comparing. For instance, the default in esquery is that it converts the class name to lowercase. We might not want that for all languages, or we might want to otherwise alter how a class name is interpreted. To that end, it seemed like a better choice to just pass the raw class name and let us decide on the other end how we'd prefer to handle it.

@nzakas Sure, simple transforms on the class name seem like a good enough reason.

Thanks so much!