Print NSData's hash instead of byte description in error messages

[inket]

closes #259

Printing the byte description of a 1MB NSData is long enough to keep printing for a while, and to freeze Xcode because of the length of the resulting message.

Print "NSData" instead as it's a decent representation of identity/equality (although not perfect).

[jeffh]

Hey @inket,

Looks good to me. Although I would like to have a test to make sure this doesn't regress at a later period before merging this in.

[inket]

Alright 👍

[briancroom]

Do you think it would be useful to also include the data's byte count in the output?

[inket]

@briancroom

Thought about it but couldn't think of an appropriate format to represent it: NSData<hash=123456789,length=38849> or NSData(hash:123456789,length:38849) or other variations? I'm open to suggestions otherwise I will just use the first one.

It is indeed useful to show the byte count as it would make it easy to tell whether an NSData is empty or not (and other things) at a glance.

[briancroom]

I like the first one, myself!

[inket]

After implementing tests, it seems that the Linux test fails because NSData's hash() causes a segmentation fault.

Using the REPL on Mac:

Welcome to Apple Swift version 3.0-dev (LLVM b361b0fc05, Clang 11493b0f62, Swift 24a0c3de75). Type :help for assistance.
  1> import Foundation
  2> "foobar".dataUsingEncoding(NSUTF8StringEncoding)?.hash
$R0: Int? = 114710658

on Linux:

Welcome to Swift version 3.0-dev (LLVM b361b0fc05, Clang 11493b0f62, Swift 24a0c3de75). Type :help for assistance.
  1> import Foundation
  2> "foobar".dataUsingEncoding(NSUTF8StringEncoding)?.hash
Execution interrupted. Enter code to recover and continue.
Enter LLDB commands to investigate (type :help for assistance.)

I fixed that by using #if os(Linux) and falling back to NSData<length=*> for now. But if that's not ideal, we can use the NSData<length=*> for all OS ? Please advise.

[briancroom]

Oh shoot, that's unfortunate. There are a number of places where we have introduced conditionals for Linux behavior already, and it seems reasonable to me to do so here as well. Please do leave a comment indicating the reason for it so that it is easier to try again in a few weeks/months and see if the issue has been fixed.

It'd be great if you could log an issue in bugs.swift.org as well!

[inket]

Thanks! This PR is good for a last review now 😁

Will report the bug on the Swift bug tracker later.

[jeffh]

LGTM, thanks @inket!