Tell Don't Ask (1997) | Hacker News Reader

Tell Don't Ask (1997)

pragprog.com

38 pointstpaschalis8 comments

The author is incorrectly using the term "declarative" when he's really describing "imperative". Stateful OOP is in no way declarative.

It's also unclear what the thesis is. He says to not ask objects about their state, then later advocates for CQRS, which does exactly that.

Of course, imperative programming is just the Church encoding of declarative programming.

Well, no. The lambda calculus isn't imperative.

In declarative programs you specify what the program is. In imperative programs you specify what the program should do. I don’t really see the overlap.

This seems to fit a couple other things. Functions to "get device capabilities" from device drivers are misguided to some extent. Also, web sites checking capabilities of your browser should be considered an anti-pattern.

what approaches do you think would be better? do you mean something along the lines of "ask for forgiveness, not permission"?

I stopped at his argument about getting a string with the contents of RED.

That's why you have a variable. Which is a specific named property of an object.

var lastName ="RED"

Umm. It's his last name?

and then you have a function like: `def printGreeting(lastName, firstName)`

and you call it like `printGreeting(firstName, lastName)`

and whoops, the compiler won't tell you you did something wrong. Maybe your editor will, if it shows handy little param name stuff next to the args in your call, but a github PR won't.

So someone ordering the arguments in a method in a non-intuitive way makes your variable name much less useful once you have to pass it around.

Sure, this is a trivial example ("why did you order the params in such a dumb order!?!?") but that's what is meant by variables not knowing much about what they are. It's a String. That doesn't tell us much, and doesn't tell the compiler or interpreter much about which functions expect it in what places.

Reply on news.ycombinator.com