IEEE 754 Representation

Is it okay to use lower bits for a type tag?

It is perfectly okay to use lower butts for a tight tap.

-0.0 is retarded.

You mean a class tag, or a runtime classification tag. Types don't need runtime tags.

>>4

You're wrong. Outside of silly academic papers, types do need tags.
http://en.wikipedia.org/wiki/Tagged_pointer

>>5
How retarded do you need to get to confuse tags on values with tags on types?

>>6

1. Type is a value.
2. Value can reference it's own type.

>>7
1. No it's not.
2. It's possible but it still doesn't mean class tags are type tags.

For example for the type

data Either a b = Left a | Right b

there might be a value

Left "testicle"

that has a tag but this tag does not specify its type. We know that it's a Either String b value but we can't tell the b from the tag. It's a classification tag and can of course be encoded various ways, but it's in no way a type tag.

Likewise you might have a type like

Numeric = I Int | D Double

and encode the class tags (I or D) in the lowest bits of its representation, but it still wouldn't make them type tags. There's no way to find out at runtime if the value D 2.71 has the type Numeric or maybe some other type

Asswipe = I Int | D Double | C Char Char Char Char Char Char

>>8

1. Yes it is. If it ranges of some bits in memory, so it is a value.
2. Stop dividing computation into compile-time and run-time. It is a false dichotomy.

>>9
1. Types can be represented by values but are not values any more than number "tau" is a (computer) value. Number "tau" is a concept wholly independent from any computing device, though it can be represented in various ways on such devices. Likewise, a type is a mathematical concept that was invented way before the age of computers and does not depend on any device with "bits in memory".

2. It is a very real and useful dichotomy since types are a form of specification and can exist independently of any runtime, of evaluation, or indeed of any computer. You can typecheck a program with just pencil and paper, without ever running it. Now, of course typechecking and compilation can happen during runtime via a full compiler embedded into the runtime, but just because two things are coterminous doesn't mean they are the same thing. You can take a shit and eat at the same time, too.

>>10

tau

stopped reading right there.

>>11
Stop repyling to >>10-kun, YHBTC.

>>12
Why? He doesn't look like a troll.

IEEE 754 sucks cummin testicles

Yes if you don't need the precision... which is probably the case if you're using floating-point anyway.

>>2

You can typecheck a program with just pencil and paper, without ever running it.

You can also "run" the program with pencil and paper. (What do you mean you need a computer...? What the fuck are they teaching kids these days...!?)

2. Stop dividing computation into compile-time and run-time. It is a false dichotomy.

For me, anything that needs input is run-time, and anything else is compile-time. (You academic intellectual masturbators are probably going to disagree but that's how it works in practice.)

>>15
You mean >>10

>>15

You can also "run" the program with pencil and paper

But then you would be running it. Whereas if you typecheck it on paper, you won't be running it. Feel the difference? In one case you run it (doesn't matter the hardware: laptop, workstation, cluster, brain), in the other you don't run it. That's why typechecking is a logically phase logically separated from runtime. Even if it might happen at the same time.

lol atheists suck at math
what's new

>>15

You can also "run" the program with pencil and paper. (What do you mean you need a computer...? What the fuck are they teaching kids these days...!?)

Yeah. Back then people literally debugged code by manually evaluating it, because computer time priced more than human.

>>17

if you typecheck it on paper, you won't be running it.

typechecking = partial evaluation.

>>19
Are there any books or articles on how to do such a thing?

>>21

Dunno. But I've seen some crazy programmers carefully following the code, making notes, instead of just pressing "run" in IDE.

I've written out big ol' state tables as a proof by tautology back when I was learning VHDL.

I don't test my programs. I just commit and let others tell me if I broke the build.

>>24
Hey, I just wanted to let you know the your crap code broke the build again.

I don't work with other, so I do everything in "master" and no one tells me a shit