Add option to preserve line ending during check/formatting

[kwin]

Currently the lineEndings option does only allow to configure which line endings to expect. There is no single option which works on non-Git based source code (i.e. extracted from a ZIP source archive) accepting both line endings (Windows and Unix).

This is crucial though because by default source archives created with maven-source-plugin are not normalized with respect to line endings, i.e. the archives end up with Windows endings on a Windows machine and Unix endings on all other machines. As the spotless configuration is part of the source archive, building should not fail with spotless errors irrespective of the line ending.

[nedtwigg]

the archives end up with Windows endings on a Windows machine and Unix endings on all other machines

PLATFORM_NATIVE should work here, yeah?

[kwin]

No, you have arbitrary line endings in the source and it should work with the same config to do spotless:check in Windows and Unix machines. PLATFORM_NATIVE only works if source was created with the same platform, but not if created on a platform with another line ending.

[nedtwigg]

I'm open to a PRESERVE option here:

spotless/lib/src/main/java/com/diffplug/spotless/LineEnding.java

Lines 45 to 50 in b2a2cbb

	/** {@code \n} on unix systems, {@code \r\n} on windows systems. */
	PLATFORM_NATIVE,
	/** {@code \r\n} */
	WINDOWS,
	/** {@code \n} */
	UNIX,

spotless/lib/src/main/java/com/diffplug/spotless/LineEnding.java

Lines 134 to 136 in b2a2cbb

	public interface Policy extends Serializable, NoLambda {
	/** Returns the line ending appropriate for the given file. */
	String getEndingFor(File file);

Two ambiguous cases:

• file with zero line endings
• file with both kinds of line endings

My vote is that both these cases round to "UNIX", but print a warning message.

[kwin]

I agree with all you said except for the case

file with both kinds of line endings

In this case I would just throw an exception. This is IMHO never acceptable.

However for performance reasons I would probably just check for the first line ending in a file (and not check all) therefore the first line ending in a file determines the outcome.