fix example codes (2019-03)

[maiha]

Hi all, it is time to catch up the examples code once a year.

• checked by https://github.com/maiha/crystal-examples
• target src: master branch ( a17a65617 Implement Int#{leading,trailing}_zeros_count (#7520))
• compiler: Crystal 0.28.0-dev [a17a656] (2019-03-20)

For convenience, commits are divided file by file, so please apply squash merge.

NOTE

Aside from PR, there are three points that I was worried about.

1. humanize

1073741824.humanize_bytes(format: :IEC)

ambiguous call, implicit cast of :IEC matches all of
Int::BinaryPrefixFormat

2. Hash#dig?

h = {"a" => {"b" => [10, 20, 30]}}
h.dig? "a", "b"                # => [10, 20, 30]
h.dig? "a", "b", "c", "d", "e" # => nil

Error in tmp/hash/all.cr:71: instantiating 'Hash(String, Hash(String, Array(Int32)))#dig?(String, String, String, String, String)'

h.dig? "a", "b", "c", "d", "e" # => nil
  ^~~~

in crystal/src/hash.cr:169: instantiating 'Hash(String, Array(Int32))#dig?(Tuple(String, String, String, String))'

      value.dig?(*subkeys)
            ^~~~

in crystal/src/hash.cr:169: no overload matches 'Array(Int32)#dig?' with types String, String, String
Overloads are:
 - Indexable(T)#dig?(index : Int)
 - Indexable(T)#dig?(index : Int, *subindexes)

      value.dig?(*subkeys)
            ^~~~

3. Int32::MAX + 1 doesn't raise Overflow

Although the compiled compiler was build with default setting -D preview_overflow -D compiler_rt ..., I can't find the overflow action.

$ crystal --version
Crystal 0.28.0-dev [a17a656] (2019-03-20)

LLVM: 4.0.0
Default target: x86_64-pc-linux-gnu

$ cat test.cr
pp! Int32::MAX + 1

$ crystal test.cr
Using compiled compiler at `.build/crystal'
Int32::MAX + 1 # => -2147483648

Best regards,

[jkthorne]

That overflow behavior seems worrying.

[straight-shoota]

The error with Int#humanize_bytes looks odd, too. Something wrong with autocasting symbols? Using :JEDEC (the other enum value) instead of :IEC results in the same message: ambiguous call, implicit cast of :JEDEC matches all of Int::BinaryPrefixFormat

[asterite]

There's a bug with named arguments and autocasting but I can't remember what was it or how to reproduce it.

[bcardiff]

3. Int32::MAX + 1 doesn't raise Overflow
   Although the compiled compiler was build with default setting -D preview_overflow -D compiler_rt ..., I can't find the overflow action.

You need to compiler the snippet with -D preview_overflow in https://github.com/maiha/crystal-examples/blob/bf2ac63fcbd51dce97065d21de3163e18111609d/src/job/compile.cr#L109 to let the compiled program have overflow behavior.

The flag that you mention is for the compiler to overflow on runtime, but does not mean that the generated program have that. It's an opt-in preview feature.

[bcardiff]

I extracted a bug from Hash#dig? 🕵️ 🐛

Thanks a lot @maiha this is great.

I guess that making all the examples a full crystal program (Ref: #7564) is something that benefits the tool you made, right?

[maiha]

I guess that making all the examples a full crystal program (Ref: #7564) is something that benefits the tool you made, right?

Yep, it helps to reduce the heuristics built in here. 😺
https://github.com/maiha/crystal-examples/blob/v0.5.1/bundled/heuristic.jnl#L110-L133

[straight-shoota]

I've created an issue for the autocasting bug (#7578). Hash.dig? is #7570.

@maiha Have you checked the overflow issue with correct flags? Then this should be good to go! 🚀

[maiha]

@maiha Have you checked the overflow issue with correct flags? Then this should be good to go! 🚀

Yep, it worked! 😺

[straight-shoota]

Thank you very much @maiha 👍