nice_uname_nerd

12 anni fadiscuss

I don't want to be too mean, but it looks like very little has changed since the six months or so when this last popped up on HN. The book still has none of the exercises that the exposition places so much weight on. Meanwhile, the introduction still mostly tells rather than shows—for all the discussion about how LISP is an elite hacker language, there are no actual examples backing up that statement (IS it in fact widely used by hackers?). Telling me how great LISP is and rambling on about Heinlenian mysticism is not going to work if theres no real substantive material.

nice_uname_nerd

12 anni fadiscuss

I don't think the problem here is functional programming, it's the author's over-eager use of category theory to describe something simple. In particular we're just talking about all possible lists you can create in LISP, so do we really need to resort to free monoids? The language of category theory is greate for describing complicated structures like tensors and sheaves and what have you, but in this example it's not really helping.

nice_uname_nerd

12 anni fadiscuss

It's more than just strings though, it's all lists on a set of atoms. (a b c) is treated differently from (a (b c)).

EDIT: Actually you're right, it's all strings on X. I thought the author wanted all lists on X.

nice_uname_nerd

12 anni fadiscuss

>units of measure are not "tools"...they are ways of communicating information.

How you communicate information is important and has real impacts. Given a graph, you could choose to encode it as an adjacency matrix or as a collection of adjacency lists. Both convey the same information, but nobody would say we should pick one and only one. I disagree with the assertion that measurement units aren't tools or that they are exempt from similar consideration.

Sure, the Mars incident is a call for consistency, but it's a call for consistency inside NASA. That shouldn't affect carpenters who need divisibility by 2, 3 and 4 more than they need 5. Or those who like to use inches because theyre significantly larger than centimeters and consequently easier to approximate by eye.

nice_uname_nerd

12 anni fadiscuss

I strongly disagree. Using the right tool for the right job is worth far more than enforcing arbitrary universal consistency. In a math course, you'd measure angles in radians so you're dealing with nice small multiples of pi. I think it'd be downright dangerous to use radians in a fighter jet, say, when the precision you need is much better provided by degrees.

Imagine if we enforced only one programming paradigm, since they're all equivalent. Or if we demanded that logicians and cs professors only prove things using turing machines, instead of picking the model of computation that best suits the problem. If there's a case where the english measurement system is more convenient, people should use it.

nice_uname_nerd

12 anni fadiscuss

If you haven't seen any category theory, you might be interested to know that the diagrams the author draws are valuable and important in modern pure math as well. Those last two diagrams totally capture the idea of what a product or variant (coproduct, in math terminology) is.

In general, there are a lot of algebraic structures whose essence is captured by some diagramatic property like these, such as tensor products or fiber products. They're called universal properties. Unfortunately the wikipedia page looks pretty poorly written, but you might find it interesting nonetheless.

http://en.wikipedia.org/wiki/Universal_property

comments