redef class FlatText

	# First byte of the CString
	protected fun first_byte: Int do return 0

	# Last byte of the CString
	protected fun last_byte: Int do return first_byte + _byte_length - 1

	# Cache of the latest position (char) explored in the string
	var position: Int = 0

	# Cached position (bytes) in the CString underlying the String
	var bytepos: Int = 0

	# Index of the character `index` in `_items`
	fun char_to_byte_index(index: Int): Int do
		var dpos = index - _position
		var b = _bytepos
		var its = _items

		if dpos == 1 then
			if its[b] & 0x80 == 0x00 then
				b += 1
			else
				b += its.length_of_char_at(b)
			end
			_bytepos = b
			_position = index
			return b
		end
		if dpos == -1 then
			b = its.find_beginning_of_char_at(b - 1)
			_bytepos = b
			_position = index
			return b
		end
		if dpos == 0 then return b

		var ln = _length
		var pos = _position
		# Find best insertion point
		var delta_begin = index
		var delta_end = (ln - 1) - index
		var delta_cache = (pos - index).abs
		var min = delta_begin

		if delta_cache < min then min = delta_cache
		if delta_end < min then min = delta_end

		var ns_i: Int
		var my_i: Int

		if min == delta_cache then
			ns_i = _bytepos
			my_i = pos
		else if min == delta_begin then
			ns_i = first_byte
			my_i = 0
		else
			ns_i = its.find_beginning_of_char_at(last_byte)
			my_i = _length - 1
		end

		ns_i = its.char_to_byte_index_cached(index, my_i, ns_i)

		_position = index
		_bytepos = ns_i

		return ns_i
	end

	# By escaping `self` to HTML, how many more bytes will be needed ?
	fun chars_to_html_escape: Int do
		var its = _items
		var max = last_byte
		var pos = first_byte
		var endlen = 0
		while pos <= max do
			var c = its[pos]
			if c == u'<' then
				endlen += 3
			else if c == u'>' then
				endlen += 3
			else if c == u'&' then
				endlen += 4
			else if c == u'"' then
				endlen += 4
			else if c == u'\'' then
				endlen += 4
			else if c == 0x2F then
				endlen += 4
			end
			pos += 1
		end
		return endlen
	end

	redef fun html_escape
	do
		var extra = chars_to_html_escape
		if extra == 0 then return to_s
		var its = _items
		var max = last_byte
		var pos = first_byte
		var nlen = extra + _byte_length
		var nits = new CString(nlen)
		var outpos = 0
		while pos <= max do
			var c = its[pos]
			# Special codes:
			# Some HTML characters are used as meta-data, they need
			# to be replaced by an HTML-Escaped equivalent
			if c == u'<' then
				nits[outpos] = u'&'
				nits[outpos + 1] = u'l'
				nits[outpos + 2] = u't'
				nits[outpos + 3] = u';'
				outpos += 4
			else if c == u'>' then
				nits[outpos] = u'&'
				nits[outpos + 1] = u'g'
				nits[outpos + 2] = u't'
				nits[outpos + 3] = u';'
				outpos += 4
			else if c == u'&' then
				nits[outpos] = u'&'
				nits[outpos + 1] = u'a'
				nits[outpos + 2] = u'm'
				nits[outpos + 3] = u'p'
				nits[outpos + 4] = u';'
				outpos += 5
			else if c == u'"' then
				nits[outpos] = u'&'
				nits[outpos + 1] = u'#'
				nits[outpos + 2] = u'3'
				nits[outpos + 3] = u'4'
				nits[outpos + 4] = u';'
				outpos += 5
			else if c == u'\'' then
				nits[outpos] = u'&'
				nits[outpos + 1] = u'#'
				nits[outpos + 2] = u'3'
				nits[outpos + 3] = u'9'
				nits[outpos + 4] = u';'
				outpos += 5
			else if c == u'/' then
				nits[outpos] = u'&'
				nits[outpos + 1] = u'#'
				nits[outpos + 2] = u'4'
				nits[outpos + 3] = u'7'
				nits[outpos + 4] = u';'
				outpos += 5
			else
				nits[outpos] = c
				outpos += 1
			end
			pos += 1
		end
		var s = new FlatString.with_infos(nits, nlen, 0)
		return s
	end

	# By escaping `self` to C, how many more bytes will be needed ?
	#
	# This enables a double-optimization in `escape_to_c` since if this
	# method returns 0, then `self` does not need escaping and can be
	# returned as-is
	fun chars_to_escape_to_c: Int do
		var its = _items
		var max = last_byte
		var pos = first_byte
		var req_esc = 0
		while pos <= max do
			var c = its[pos]
			if c == u'\n' then
				req_esc += 1
			else if c == u'\t' then
				req_esc += 1
			else if c == u'"' then
				req_esc += 1
			else if c == u'\'' then
				req_esc += 1
			else if c == u'\\' then
				req_esc += 1
			else if c == u'?' then
				var j = pos + 1
				if j < length then
					var next = its[j]
					# We ignore `??'` because it will be escaped as `??\'`.
					if
						next == 0x21 or
						next == 0x28 or
						next == 0x29 or
						next == 0x2D or
						next == 0x2F or
						next == 0x3C or
						next == 0x3D or
						next == 0x3E
					then req_esc += 1
				end
			else if c < 32 then
				req_esc += 3
			end
			pos += 1
		end
		return req_esc
	end

	redef fun escape_to_c do
		var ln_extra = chars_to_escape_to_c
		if ln_extra == 0 then return self.to_s
		var its = _items
		var max = last_byte
		var nlen = _byte_length + ln_extra
		var nns = new CString(nlen)
		var pos = first_byte
		var opos = 0
		while pos <= max do
			var c = its[pos]
			# Special codes:
			#
			# Any byte with value < 32 is a control character
			# All their uses will be replaced by their octal
			# value in C.
			#
			# There are two exceptions however:
			#
			# * 0x09 => \t
			# * 0x0A => \n
			#
			# Aside from the code points above, the following are:
			#
			# * 0x22 => \"
			# * 0x27 => \'
			# * 0x5C => \\
			if c == u'\t' then
				nns[opos] = u'\\'
				nns[opos + 1] = u't'
				opos += 2
			else if c == u'\n' then
				nns[opos] = u'\\'
				nns[opos + 1] = u'n'
				opos += 2
			else if c == u'"' then
				nns[opos] = u'\\'
				nns[opos + 1] = u'"'
				opos += 2
			else if c == u'\'' then
				nns[opos] = u'\\'
				nns[opos + 1] = u'\''
				opos += 2
			else if c == u'\\' then
				nns[opos] = u'\\'
				nns[opos + 1] = u'\\'
				opos += 2
			else if c == u'?' then
				var j = pos + 1
				if j < length then
					var next = its[j]
					# We ignore `??'` because it will be escaped as `??\'`.
					if
						next == 0x21 or
						next == 0x28 or
						next == 0x29 or
						next == 0x2D or
						next == 0x2F or
						next == 0x3C or
						next == 0x3D or
						next == 0x3E
					then
						nns[opos] = 0x5C
						opos += 1
					end
				end
				nns[opos] = 0x3F
				opos += 1
			else if c < 32 then
				nns[opos] = u'\\'
				nns[opos + 1] = u'0'
				nns[opos + 2] = ((c & 0x38) >> 3) + u'0'
				nns[opos + 3] = (c & 0x07) + u'0'
				opos += 4
			else
				nns[opos] = c
				opos += 1
			end
			pos += 1
		end
		return nns.to_s_unsafe(nlen, copy=false, clean=false)
	end

	redef fun [](index) do
		var len = _length

		# Statistically:
		# * ~70% want the next char
		# * ~23% want the previous
		# * ~7% want the same char
		#
		# So it makes sense to shortcut early. And early is here.
		var dpos = index - _position
		var b = _bytepos
		if dpos == 1 and index < len - 1 then
			var its = _items
			var c = its[b]
			if c & 0x80 == 0x00 then
				# We want the next, and current is easy.
				# So next is easy to find!
				b += 1
				_position = index
				_bytepos = b
				# The rest will be done by `dpos==0` bellow.
				dpos = 0
			end
		else if dpos == -1 and index > 1 then
			var its = _items
			var c = its[b-1]
			if c & 0x80 == 0x00 then
				# We want the previous, and it is easy.
				b -= 1
				dpos = 0
				_position = index
				_bytepos = b
				return c.code_point
			end
		end
		if dpos == 0 then
			# We know what we want (+0 or +1) just get it now!
			var its = _items
			var c = its[b]
			if c & 0x80 == 0x00 then return c.code_point
			return items.char_at(b)
		end

		assert index >= 0 and index < len
		return fetch_char_at(index)
	end

	# Gets a `Char` at `index` in `self`
	#
	# WARNING: Use at your own risks as no bound-checking is done
	fun fetch_char_at(index: Int): Char do
		var i = char_to_byte_index(index)
		var items = _items
		var b = items[i]
		if b & 0x80 == 0x00 then return b.code_point
		return items.char_at(i)
	end

	# If `self` contains only digits and alpha <= 'f', return the corresponding integer.
	#
	#     assert "ff".to_hex == 255
	redef fun to_hex(pos, ln) do
		var res = 0
		if pos == null then pos = 0
		if ln == null then ln = length - pos
		pos = char_to_byte_index(pos)
		var its = _items
		var max = pos + ln
		for i in [pos .. max[ do
			res <<= 4
			res += its[i].code_point.from_hex
		end
		return res
	end

	redef fun copy_to_native(dst, n, src_off, dst_off) do
		_items.copy_to(dst, n, first_byte + src_off, dst_off)
	end
end