Add EOLconstant (End-Of-Line)

[postmodern]

• Added EOL to src/kernel.cr.
• Is OS aware ("\r\n" on Windows, '\n' otherwise).

Resolves #10885

[postmodern]

Some additional thoughts:

• Having EOL be a String on Windows, but a Char elsewhere might cause some unexpected type issues?
• Should we also add explicit CR, LF and CRLF constants? HTTP also uses CRLF.
• Is the top-level really the best place for such String constants?

[straight-shoota]

Having EOL be a String on Windows, but a Char elsewhere might cause some unexpected type issues?

Yes, it should be the same type on all platforms (cf #10645). String is a perfect match for this.

Should we also add explicit CR, LF and CRLF constants? HTTP also uses CRLF.

I don't think that's necessary. Libraries like HTTP can of course define such constants as they need them.

Is the top-level really the best place for such String constants?

String::EOL or System::EOL might be viable alternatives. I might fancy the latter because it emphasizes that the constant's value is system-specific. Not settled on that, though.

[asterite]

What is the use case of this constant?

[straight-shoota]

@asterite The proposal is in #10885

[Blacksmoke16]

👍 for System::EOL. I checked a few other langs and they all seem to have it in systemish namespaces:

• Java - System.lineSeparator
• Python - os.linesep
• C# - Environment.NewLine

[postmodern]

Do we have consensus on System::EOL vs String::EOL?

[j8r]

I may be alone, I still prefer NEWLINE (in whatever namespace it lands). Already quite small and more explicit than EOL.

[Sija]

@j8r If I'd like to write code that would be read by my grandma I'd agree with you for sure, but most programmers I know have a pretty firm notion of EOF, EOL, and other abbreviations from the programming world... ;)

[j8r]

Anyway, can this constant be useful in the stdlib, and even the compiler?

[HertzDevil]

crystal/spec/std/process_spec.cr

Lines 53 to 59 in 4204bce

	private def newline
	{% if flag?(:win32) %}
	"\r\n"
	{% else %}
	"\n"
	{% end %}
	end

[Blacksmoke16]

System::EOL is still my vote. Just ran into a use case where this would be helpful.

[HertzDevil]

Another one:

crystal/spec/compiler/macro/macro_methods_spec.cr

Lines 36 to 42 in 3a360b1

	private def newline
	{% if flag?(:win32) %}
	"\r\n"
	{% else %}
	"\n"
	{% end %}
	end

[straight-shoota]

LGTM.

[Sija]

Could be shortened into:

Suggested change

	EOL = {% if flag?(:windows) %}
	"\r\n"
	{% else %}
	"\n"
	{% end %}
	EOL = {{ flag?(:windows) ? "\r\n" : "\n" }}

[postmodern]

I find ternary statements overly confusing. They only help reduce line count at the cost of readability.

[straight-shoota]

I suppose you could also write it as {{ if flag?(:windows) "\r\n" else "\n" end }} if that's more approachable?
Anyway, these are all equivalent expressions of the same thing and there's no clear preference for one or the other. So let's keep it as is.