/**
 * std.regex helpers
 *
 * License:
 *   This Source Code Form is subject to the terms of
 *   the Mozilla Public License, v. 2.0. If a copy of
 *   the MPL was not distributed with this file, You
 *   can obtain one at http://mozilla.org/MPL/2.0/.
 *
 * Authors:
 *   Vladimir Panteleev <ae@cy.md>
 */

module ae.utils.regex;

import std.algorithm;
import std.conv;
import std.exception;
import std.regex;
import std.string;
import std.traits;
import std.typecons;

import ae.utils.text;

// ************************************************************************

/// Allows specifying regular expression patterns in expressions,
/// without having to compile them each time.
/// Example:
///   if (text.match(`^\d+$`)) {}    // old code - recompiles every time
///   if (text.match(re!`^\d+$`)) {} // new code - recompiles once

Regex!char re(string pattern, alias flags = [])()
{
	static Regex!char r;
	if (r.empty)
		r = regex(pattern, flags);
	return r;
}

unittest
{
	assert( "123".match(re!`^\d+$`));
	assert(!"abc".match(re!`^\d+$`));
}

private void convertCaptures(C, T...)(C captures, out T values)
{
	assert(values.length == captures.length-1, "Capture group count mismatch: %s arguments / %s capture groups".format(values.length, captures.length-1));
	foreach (n, ref value; values)
		value = to!(T[n])(captures[n+1]);
}

/// Lua-like pattern matching.
bool matchInto(S, R, Args...)(S s, R r, ref Args args)
{
	auto m = s.match(r);
	if (m)
	{
		convertCaptures(m.captures, args);
		return true;
	}
	return false;
}

///
unittest
{
	string name, fruit;
	int count;
	assert("Mary has 5 apples"
		.matchInto(`^(\w+) has (\d+) (\w+)$`, name, count, fruit));
	assert(name == "Mary" && count == 5 && fruit == "apples");
}

/// Match into a delegate.
bool matchCaptures(S, R, Ret, Args...)(S s, R r, Ret delegate(Args args) fun)
{
	auto m = s.match(r);
	if (m)
	{
		Args args;
		convertCaptures(m.captures, args);
		fun(args);
		return true;
	}
	return false;
}

///
unittest
{
	assert("Mary has 5 apples"
		.matchCaptures(`^(\w+) has (\d+) (\w+)$`,
			(string name, int count, string fruit)
			{
				assert(name == "Mary" && count == 5 && fruit == "apples");
			}
		)
	);
}

/// Call a delegate over all matches.
size_t matchAllCaptures(S, R, Ret, Args...)(S s, R r, Ret delegate(Args args) fun)
{
	size_t matches;
	foreach (m; s.matchAll(r))
	{
		Args args;
		convertCaptures(m.captures, args);
		fun(args);
		matches++;
	}
	return matches;
}

/// Returns a range which extracts a capture from text.
template extractCaptures(T...)
{
	auto extractCaptures(S, R)(S s, R r)
	{
		return s.matchAll(r).map!(
			(m)
			{
				static if (T.length == 1)
					return m.captures[1].to!T;
				else
				{
					Tuple!T r;
					foreach (n, TT; T)
						r[n] = m.captures[1+n].to!TT;
					return r;
				}
			});
	}
}

alias extractCapture = extractCaptures; /// ditto

auto extractCapture(S, R)(S s, R r)
if (isSomeString!S)
{
	alias x = .extractCaptures!S;
	return x(s, r);
} /// ditto

///
unittest
{
	auto s = "One 2 three 42";
	auto r = `(\d+)`;
	assert(s.extractCapture    (r).equal(["2", "42"]));
	assert(s.extractCapture!int(r).equal([ 2 ,  42 ]));
}

///
unittest
{
	auto s = "2.3 4.56 78.9";
	auto r = `(\d+)\.(\d+)`;
	assert(s.extractCapture!(int, int)(r).equal([tuple(2, 3), tuple(4, 56), tuple(78, 9)]));
}

// ************************************************************************

/// Take a string, and return a regular expression that matches that string
/// exactly (escape RE metacharacters).
string escapeRE(string s)
{
	// TODO: test

	string result;
	foreach (c; s)
		switch (c)
		{
		//	case '!':
		//	case '"':
		//	case '#':
			case '$':
		//	case '%':
		//	case '&':
			case '\'':
			case '(':
			case ')':
			case '*':
			case '+':
		//	case ',':
		//	case '-':
			case '.':
			case '/':
		//	case ':':
		//	case ';':
		//	case '<':
		//	case '=':
		//	case '>':
			case '?':
		//	case '@':
			case '[':
			case '\\':
			case ']':
			case '^':
		//	case '_':
		//	case '`':
			case '{':
			case '|':
			case '}':
		//	case '~':
				result ~= '\\';
				goto default;
			default:
				result ~= c;
		}
	return result;
}

// We only need to make sure that there are no unescaped forward slashes
// in the regex, which would mean the end of the search pattern part of the
// regex transform. All escaped forward slashes will be unescaped during
// parsing of the regex transform (which won't affect the regex, as forward
// slashes have no special meaning, escaped or unescaped).
private string escapeUnescapedSlashes(string s)
{
	bool escaped = false;
	string result;
	foreach (c; s)
	{
		if (escaped)
			escaped = false;
		else
		if (c == '\\')
			escaped = true;
		else
		if (c == '/')
			result ~= '\\';

		result ~= c;
	}
	assert(!escaped, "Regex ends with an escape");
	return result;
}

// For the replacement part, we just need to escape all forward and backslashes.
private string escapeSlashes(string s)
{
	return s.fastReplace(`\`, `\\`).fastReplace(`/`, `\/`);
}

// Reverse of the above
private string unescapeSlashes(string s)
{
	return s.fastReplace(`\/`, `/`).fastReplace(`\\`, `\`);
}

/// Build a RE search-and-replace transform (as used by applyRE).
string buildReplaceTransformation(string search, string replacement, string flags)
{
	return "s/" ~ escapeUnescapedSlashes(search) ~ "/" ~ escapeSlashes(replacement) ~ "/" ~ flags;
}

private struct Transformation
{
	enum Type
	{
		replace,
	}
	Type type;

	struct Replace
	{
		string search, replacement, flags;
	}

	union
	{
		Replace replace;
	}
}

private Transformation[] splitRETransformation(string s)
{
	enforce(s.length >= 2, "Bad transformation");
	Transformation[] result;
	while (s.length)
	{
		Transformation t;
		switch (s[0])
		{
			case 's':
			{
				t.type = Transformation.Type.replace;
				s = s[1..$];

				auto boundary = s[0];
				s = s[1..$];

				string readString()
				{
					bool escaped = false;
					foreach (i, c; s)
						if (escaped)
							escaped = false;
						else
						if (c=='\\')
							escaped = true;
						else
						if (c == boundary)
						{
							auto result = s[0..i];
							s = s[i+1..$];
							return result;
						}
					throw new Exception("Unexpected end of regex replace transformation");
				}

				t.replace.search = readString();
				t.replace.replacement = readString();
				foreach (i, c; s)
				{
					if (c == ';')
					{
						t.replace.flags = s[0..i];
						s = s[i+1..$];
						goto endOfReplace;
					}
					else
					if (c == boundary)
						throw new Exception("Too many regex replace transformation parameters");
				}
				t.replace.flags = s;
				s = null;
			endOfReplace:
				result ~= t;
				break;
			}
			default:
				throw new Exception("Unsupported regex transformation: " ~ s[0]);
		}
	}
	return result;
}

unittest
{
	auto actual = splitRETransformation("s/from/to/");
	Transformation expected = { type : Transformation.Type.replace, replace : { search : "from", replacement : "to", flags : "" } };
	assert(actual.length == 1 && actual[0].tupleof == expected.tupleof);
}

/// Apply sed-like regex transformation (in the form of "s/FROM/TO/FLAGS") to a string.
/// Multiple commands can be separated by ';'.
string applyRE()(string str, string transformation)
{
	import std.regex;
	auto transformations = splitRETransformation(transformation);
	foreach (t; transformations)
		final switch (t.type)
		{
			case Transformation.Type.replace:
			{
				auto r = regex(t.replace.search, t.replace.flags);
				str = replace(str, r, unescapeSlashes(t.replace.replacement));
			}
		}
	return str;
}

unittest
{
	auto transformation = buildReplaceTransformation(`(?<=\d)(?=(\d\d\d)+\b)`, `,`, "g");
	assert("12000 + 42100 = 54100".applyRE(transformation) == "12,000 + 42,100 = 54,100");

	void testSlashes(string s)
	{
		assert(s.applyRE(buildReplaceTransformation(`\/`, `\`, "g")) == s.fastReplace(`/`, `\`));
		assert(s.applyRE(buildReplaceTransformation(`\\`, `/`, "g")) == s.fastReplace(`\`, `/`));
	}
	testSlashes(`a/b\c`);
	testSlashes(`a//b\\c`);
	testSlashes(`a/\b\/c`);
	testSlashes(`a/\\b\//c`);
	testSlashes(`a//\b\\/c`);

	assert("babba".applyRE(`s/a/c/g;s/b/a/g;s/c/b/g`) == "abaab");
}

// ************************************************************************