Java unsigned types availability.

[scarlet forge]

@jagged vortex has a question:

bshmanair

does java have unsigned types like in c and c++

<@&987246399047479336> please have a look, thanks.

[weary adder]

not directly, no. the question would be why u think u need it

and based on ur response to that ull get proper better alternatives

[jagged vortex]

maybe there's user input and i just don't want them to insert values outside my intended range

like, for example, ID

[weary adder]

okay, u cant prevent that using unsigned types. also not in c++

[jagged vortex]

oh yea fair

[weary adder]

u never can prevent an user actually entering sth. u dont have control over their keyboard.

but u can ofc check what they entered, validate it and if wrong show an error and repeat

[jagged vortex]

so this means if / switch statements are the only solution to validation?

[weary adder]

yes, in java and also in c++ and anywhere else

[jagged vortex]

got it

[raven yoke]

Like zabu said, unsigned types don't solve that
int max value is around 2 billions, so what if you use unsigned int with a size of 4 billions so even if the user enter 3 bilion, it still work. But then what if the user enters 5 billions ? Do you use long ? what if a user enter a number bigger than any fixed size int ? Do you use BigInteger ? What if the user enters a number so big it crashes your server ? Etc
So yea, validate inputs via ifs or catches

That said, java does have methods to handle unsigned ints, it's more annoying to use than unsigned types, but it's here

[bold remnant]

when reading raw memory you usually want to have it unsigned. signed byte only is kinda rubbish

i mean sure you do not read memory in java, but maybe you want to read raw bytes from a file

[strange furnace]

If you want an unsigned type they aren't in the language, but various unsigned types exist in libraries

But they are implemented by just " boxing up" the validation logic + operations

So in a sense they aren't as special as many would want

[bright hedge]

unsigned types in java, while annoying that they're missing are actually completely not required

You just need a small mental model on what operations you can and can't do straight up, and which ones you need a specfic function for

add subtract and multiply are all fine, divide toString and parsing isn't

there's an unsigned right shift that doesn't duplicate the sign bit

And when using the value make sure to do a toUnsignedXXX conversion that widens it

Other than that, it's really not that big of a difference, just a bit more code than it needs to be

I have a few of these 😄

[verbal lake]

At times I would like their convenience. And while I don't completely agree with some of the JDK team on the 'yeah but', there is some truth to the fact that having both signed and unsigned type does add a lot of "but what happens when I" combinations.

I don't miss them enough to have that much skin in the game - at least not on projects I've been most involved in the last decade.

[bright hedge]

Yeah, I've ported one of my libraries to C# and "I'll just use unsigned" turns into "regular ints are easier because otherwise I have to cast for array accesses and a bunch of other random crap"

[verbal lake]

I don't do anything with hardware, and rarely am I transforming binary formats where it matters.

While I think they probably should have been left in... It's not at all the thing I'd put any thought into if I was the one picking what to focus on to change now...

The exceptions topic that was posted recently is interesting.

[bright hedge]

Thanks for reminding me I have to watch that video 🙂

[verbal lake]

Having encountered way too many libraries (in Java and .net) that don't document their exceptions, I really would like the ceremony and expressiveness barriers for more appropriate checked-exception use.

The compositional problem with checked exceptions is a pet topic of mine.

[bright hedge]

Can't even get my team to agree to when to use checked and unchecked exceptions