Monkey Patching

## What the class keyword really does In many languages a class declaration defines a type once, at compile time. In Ruby, `class Name ... end` is executable code that runs top to bottom: 1. Ruby looks up the constant `Name` in the current scope. 2. If it is **not defined**, Ruby creates a new class, assigns it to the constant, and runs the body inside it. 3. If it **is defined and is a class**, Ruby runs the body inside **that existing class**. This is called **reopening** the class; Ruby classes are therefore called **open classes**. 4. If it is defined but holds something else (an integer, a module), Ruby raises `TypeError`. So writing `class String` in your own file does not create a new `String`; it gives you the real, built-in `String` to modify. Adding or changing methods on a class you do not own this way is **monkey patching**. ## Who sees the new method A `def` inside the reopened body adds an entry to `String`'s **method table**. Because Ruby looks methods up on every call, the effect is immediate and global: - **Existing instances**: a string created before the reopening can call the method afterwards. - **Every file**: there is no import boundary; code that never required your file still sees the method once your file has run. - **Every gem**: library code inside your dependencies sees it too, and so do other patches. - **Subclasses**: any subclass of `String` inherits the method. | Situation | Result | |---|---| | `class String; def shout = upcase + "!"; end` | `"hi".shout # => "HI!"` for every string | | Same name as an existing method | replaces it for everyone (warning only under `-w`) | | `class String < Object` | allowed: Object is already String's superclass | | `class String < Array` | `TypeError`: superclass mismatch for class String | ## Why this matters for core classes The global reach is the whole point — and the whole risk: - A helper like `String#titleize` becomes available everywhere without passing anything around, which is why libraries historically added convenience methods to core classes. - Two libraries defining the same method on `String` collide, and the one loaded last wins. - Replacing an existing method such as `String#length` changes the behaviour of code you never read, including the standard library. Because of this, most teams limit core patches to a few reviewed files, often a `core_ext` directory with one file per patched class, and prefer a helper method or a small wrapper class when the convenience is only needed in their own code. ## Related forms - `module Comparable ... end` reopens an existing module the same way. - Writing the body inside `String.class_eval { ... }` also modifies the existing class, with different scoping rules; that is its own topic. - A **refinement** limits a patch to the files that activate it with `using`, the main alternative when a global change is too risky. ## Common mistakes - **Assuming a new class was created.** There is no second `String` to find; the added method sits on the real one, next to `upcase` and `length`. - **Forgetting load order.** The method exists only after the reopening file has run, so code that runs earlier, for example during boot, raises `NoMethodError`. - **Replacing instead of adding.** Reusing a name `String` already has, such as `strip` or `length`, silently changes a core method for every library, and the `method redefined` warning appears only under `-w`. ## What interviewers are checking This is a screening question about Ruby's object model: a class is an ordinary object that stays mutable at run time. A good answer says that reopening modifies the existing class, that the change is process-wide and affects existing objects, and names the risk that follows — collisions and surprising behaviour in code you did not write.