Added std.string.outdent (try #2)

[Abscissa]

This puts everything from pull request #156 into one commit.

[jmdavis]

This is a problem if C is mutable. It'll work on Windows but not on Linux (since string literals are actually immutable on Linux; I'm not sure why they aren't on Windows).

[andralex]

I may be missing something, but this implementation is more complicated than it should. If I understand the intent correctly, it should be able to do its job in two passes:

1. Compute the minimum amount of leading spaces, call it N
2. drop at most N characters off each line.

There should be no need for e.g. an array of indents and other artifacts. What am I missing?

[andralex]

I'd suspected the implementation is a fair amount more complicated than it should, and the way I read it it has issues with multibyte characters. So I wrote this:

String[] outdent(String)(String[] lines) if (isSomeString!String)
{
    size_t minLeadingSpaces = size_t.max;
    foreach (line; lines)
    {
        size_t leadingSpaces = 0;
        foreach (dchar c; line)
        {
            if (std.uni.isSpace(c))
            {
                ++leadingSpaces;
            }
            else
            {
                if (leadingSpaces < minLeadingSpaces)
                {
                    minLeadingSpaces = leadingSpaces;
                }
                break;
            }
        }
    }

    if (!minLeadingSpaces || minLeadingSpaces == size_t.max)
    {
        return lines;
    }

    foreach (ref line; lines)
    {
        line.popFrontN(minLeadingSpaces);
    }

    return lines;
}

String outdent(String)(String s) if (isSomeString!String)
{
    auto lines = outdent(splitLines(s));
    s = join(lines, "\n");
    if (s.endsWith('\n')) s ~= '\n';
    return s;
}

You may want to adapt it for this submission. Thanks.

[Abscissa]

I've cleaned up the (S str) version as per your suggestions. I've also eliminated the array of indents from the the (S[] lines) version. I don't think it can be significantly simplified further without introducing incorrect behavior upon inconsistent indentations and/or whitespace-only lines.

I also cleaned up the unittests.

I've looked carefully through, and I don't see anything that looks like a potential multibyte-char problem. Is there anything specific you see?

[Abscissa]

If you intend to accept #278, then that should probably be done before this is accepted so I can make use of it in outdent without needing to do it in a followup pull request.

[Abscissa]

Ok, I'm done with the changes I was planning. Any further issues/comments? If not, should I make it all a single commit again?

[jmdavis]

Looks good to me, so please do rebase so that it's one commit.

[Abscissa]

Superceded by #282