I've created a small library that allows to create mock functions. You can record the calls to the function and check arguments with it. It happened to me multiple times that somebody has changed the function that was mocked and it didn't comply with the values returned by mock or the function was called differenty in the code.
I recently added a support for instrumentation of mock functions with Malli schemas. If you declare a mock for a function that has Malli function schema, it will instrument the mock. This allows to discover mocks returning invalid values or accepting invalid arguments or the number of arguments.
Here's an example:
(require '[malli.core :as m])
(m/=> my-inc [:=> [:cat :int] :int])
(defn my-inc [x]
(inc x))
;; You have to provide a symbol of function
;; or with or without namespace (aliases are supported).
(def my-inc-mock (mock-fn 'my-inc 1))
(my-inc-mock 0)
;=> 1
(my-inc-mock "foo")
;=> An exception ::invalid-input
(def my-inc-mock2 (mock-fn 'my-inc nil))
(my-inc-mock2 1)
;=> An exception ::invalid-output
This instrumentation works also with a macro for defining mocks.
(with-mocks [my-inc 2]
(my-inc "foo"))
;=> An exception ::invalid-input
(with-mocks [my-inc nil]
(my-inc 1))
;=> An exception ::invalid-output
I don’t know what is Altman’s experience, but almost nobody writes software character by character for many years. Context code completion is present for about 30 years already (take a look at Delphi or VS6 from 90ties). Refactoring tooling is about 20 years old too (see Eclipse or other Java IDEs). So if he says character by character, it is just silly. When you write prompt you also write is char-by-char right? :).