deferred class
	EL_READABLE_ZSTRING

inherit
	READABLE_STRING_32
		rename
			area as unencoded_area,
			code as z_code,
			has_code as has_unicode,
			make_from_string_general as make_from_general,
			split as split_list,
			string_searcher as string_32_searcher,
			to_lower_area as unencoded_to_lower_area,
			to_upper_area as unencoded_to_upper_area
		export
			{NONE} unencoded_to_lower_area, unencoded_to_upper_area
			{EL_ZSTRING_IMPLEMENTATION, STRING_HANDLER} unencoded_area
		undefine
--			Initialization
			make,
--			Access
			area_lower, area_upper, fuzzy_index, index_of, last_index_of, out,
			substring_index_in_bounds, substring_index, string,
--			Status query			
			has, has_unicode, ends_with, ends_with_general, starts_with, starts_with_general, is_less,
			is_boolean, is_real, is_real_32, is_double, is_real_64,
			is_integer_8, is_integer_16, is_integer, is_integer_32, is_integer_64,
			is_natural_8, is_natural_16, is_natural, is_natural_32, is_natural_64,
			is_substring_whitespace, is_valid_as_string_8, valid_code,
--			Conversion
			as_string_8, as_string_32, split_list,
			to_boolean, to_real, to_real_32, to_double, to_real_64,
			to_integer_8, to_integer_16, to_integer, to_integer_32, to_integer_64,
			to_natural_8, to_natural_16, to_natural, to_natural_32, to_natural_64,
			to_string_8, to_string_32,
--			Comparison
			same_caseless_characters, same_characters, same_characters_general, same_caseless_characters_general,
--			Element change
			fill_character,
--			Measurement
			capacity, occurrences,
--			Implementation
			is_valid_integer_or_natural
		redefine
--			Initialization
			make_from_string,
--			Access
			hash_code, new_cursor,
--			Comparison
			is_equal,
--			Duplication
			copy
		end

	EL_CHARACTER_TESTABLE_ZSTRING

	EL_COMPARABLE_ZSTRING
		export
			{STRING_HANDLER, EL_OCCURRENCE_INTERVALS}
				compatible_string_8, empty_unencoded_buffer, item_8, order_comparison,
				set_unencoded_from_buffer
			{EL_ZSTRING_ITERATION_CURSOR, EL_STRING_8_IMPLEMENTATION}
				area_lower, area_upper, area, unencoded_area
			{EL_ZSTRING_IMPLEMENTATION}
				unencoded_i_th_substring, unencoded_first_lower, unencoded_first_upper, unencoded_interval_sequence,
				unencoded_fill_list
			{EL_ZCODEC} codec
		end

	EL_CONVERTABLE_ZSTRING

	EL_MEASUREABLE_ZSTRING

	EL_SEARCHABLE_ZSTRING
		export
			{EL_APPENDABLE_ZSTRING} internal_substring_index_list_general
		end

	EL_ZSTRING_TO_BASIC_TYPES

	READABLE_INDEXABLE [CHARACTER_32]
		rename
			upper as count
		undefine
			out, copy, is_equal
		redefine
			new_cursor
		end

feature {NONE} -- Initialization

	make (n: INTEGER)
			-- Allocate space for at least `n' characters.
		do
			internal_make (n)
			make_unencoded
		end

	make_from_cil (a_system_string: detachable SYSTEM_STRING)
		do
		end

	make_from_file (text: STRING)
		-- make from full file `text' which may have a BOM mark indicating it is encoded as
		-- either: UTF-8, little-endian UTF-16 or else Latin-1
		local
			utf: EL_UTF_CONVERTER
		do
			if utf.is_utf_8_file (text) then
				make_from_utf_8 (utf.bomless_utf_8 (text))

			elseif utf.is_utf_16_le_file (text) then
				make_from_utf_16_le (utf.bomless_utf_16_le (text))

			else
				make_filled ('%U', text.count)
				encode (text, 0)
			end
		end

	make_from_general (s: READABLE_STRING_GENERAL)
		do
			if same_type (s) and then attached {ZSTRING} s as z_str then
				make_from_other (z_str)
			else
				make_filled ('%U', s.count)
				encode (s, 0)
			end
		end

	make_from_latin_1_c (latin_1_ptr: POINTER)
		local
			latin: EL_STRING_8
		do
			latin := String_8.c_string (latin_1_ptr)
			if latin.is_ascii then
				make_unencoded
				set_from_ascii (latin)
			else
				make_from_general (latin)
			end
		end

	make_from_other (other: EL_READABLE_ZSTRING)
		do
			area := other.area.twin
			count := other.count
			make_unencoded_from_other (other)
		end

	make_from_string (str_32: READABLE_STRING_32)
		do
			if same_type (str_32) and then attached {ZSTRING} str_32 as z_str then
				make_from_other (z_str)
			else
				make_filled ('%U', str_32.count)
				encode (str_32, 0)
			end
		end

	make_from_string_8 (str_8: READABLE_STRING_8)
			-- initialize with string that has the same encoding as codec
		require else
			must_not_have_reserved_substitute_character: not str_8.has ('%/026/')
		do
			make_unencoded
			String_8.make_from_string (Current, str_8)
		end

	make_from_substring (general: READABLE_STRING_GENERAL; start_index, end_index: INTEGER)
		require
			valid_end_index: end_index <= general.count
		local
			l_count: INTEGER
		do
			l_count := end_index - start_index + 1
			if l_count < 1 then
				make_empty
			else
				make (l_count)
				append_substring_general (general, start_index, end_index)
			end
		end

	make_from_utf_16_le (utf_16_le_string: READABLE_STRING_8)
		local
			utf_16_le: EL_UTF_16_LE_CONVERTER; unicode_count: INTEGER
		do
			unicode_count := utf_16_le.unicode_count (utf_16_le_string)
			make (unicode_count)
			internal_append_utf (utf_16_le_string, 16, unicode_count)
		end

	make_from_utf_8 (utf_8_string: READABLE_STRING_8)
		local
			utf_8: EL_UTF_8_CONVERTER; unicode_count: INTEGER
		do
			unicode_count := utf_8.unicode_count (utf_8_string)
			make (unicode_count)
			internal_append_utf (utf_8_string, 8, unicode_count)
		end

	make_from_zcode_area (zcode_area: SPECIAL [NATURAL])
		local
			i, l_count, last_upper: INTEGER
		do
			l_count := zcode_area.count
			make (l_count)
			if attached empty_unencoded_buffer as buffer and then attached area as l_area then
				last_upper := buffer.last_upper
				from i := 0 until i = l_count loop
					if zcode_area [i] > 0xFF then
						l_area [i] := Substitute
						last_upper := buffer.extend_z_code (zcode_area [i], last_upper, i + 1)
					else
						l_area [i] := zcode_area [i].to_character_8
					end
					i := i + 1
				end
				set_count (l_count)
				buffer.set_last_upper (last_upper)
				set_unencoded_from_buffer (buffer)
			end
		end

	make_shared (other: like Current)
		do
			share (other)
		end

feature -- Access

	fuzzy_index (other: READABLE_STRING_GENERAL; start_index: INTEGER; fuzz: INTEGER): INTEGER
			-- <Precursor>
		do
			Result := string_searcher.fuzzy_index (Current, other, start_index, count, fuzz)
		end

	hash_code: INTEGER
		-- Hash code value
		local
			i, i_upper, block_index, code_i: INTEGER
			iter: EL_COMPACT_SUBSTRINGS_32_ITERATION; c_i: CHARACTER
		do
			inspect internal_hash_code
				when 0 then
					if attached unicode_table as l_unicode_table and then attached unencoded_area as area_32
						and then attached area as l_area
					then
						from i := area_lower; i_upper := area_upper until i > i_upper loop
							c_i := l_area [i]
							inspect c_i
								when Substitute then
									code_i := iter.item ($block_index, area_32, i + 1).code

								when Control_0 .. Control_25, Control_27 .. Max_ascii then
									code_i := c_i.code
							else
								code_i := l_unicode_table [c_i.code].code
							end
							Result := ((Result \\ 8388593) |<< 8) + code_i
							i := i + 1
						end
					end
					internal_hash_code := Result
			else
				Result := internal_hash_code
			end
		end

	new_cursor: EL_ZSTRING_ITERATION_CURSOR
		do
			create Result.make (Current)
			Result.start
		end

	share (other: like Current)
		do
			internal_share (other)
			unencoded_area := other.unencoded_area
		end

feature -- Status query

	encoded_with (a_codec: EL_ZCODEC): BOOLEAN
		do
			Result := a_Codec.same_type (codec)
		end

	extendible: BOOLEAN = True
			-- May new items be added? (Answer: yes.)

	is_canonically_spaced: BOOLEAN
		-- `True' if the longest substring of whitespace consists of one space character (ASCII 32)
		local
			uc_i: CHARACTER_32; i, i_final, space_count, block_index: INTEGER
			iter: EL_COMPACT_SUBSTRINGS_32_ITERATION; c32: EL_CHARACTER_32_ROUTINES
			is_space: BOOLEAN; c_i: CHARACTER
		do
			if attached area as l_area and then attached unencoded_area as area_32 then
				Result := True; i_final := count
				from i := 0 until not Result or else i = i_final loop
					c_i := l_area [i]
					inspect c_i
						when Substitute then
						uc_i := iter.item ($block_index, area_32, i + 1)
					-- `c32.is_space' is workaround for finalization bug
						is_space := c32.is_space (uc_i)
					else
						is_space := c_i.is_space
					end
					if is_space then
						space_count := space_count + 1
					else
						space_count := 0
					end
					inspect space_count
						when 0 then
							do_nothing
						when 1 then
							Result := c_i = ' '
					else
						Result := False
					end
					i := i + 1
				end
			end
		end

	is_compatible (str_8: READABLE_STRING_8): BOOLEAN
		-- `True' if `str_8' can be appended directly to `area' without any `Codec' encoding
		do
			Result := attached compatible_string_8 (str_8)
		end

	is_compatible_substring (str_8: READABLE_STRING_8; start_index, end_index: INTEGER): BOOLEAN
		-- `True' if `str_8' can be appended directly to `area' without any `Codec' encoding
		do
			Result := attached compatible_substring_8 (str_8, start_index, end_index)
		end

	is_left_adjustable: BOOLEAN
		-- True if `left_adjust' will change the `count'
		do
			Result := not is_empty and then is_space_item (1)
		end

	is_right_adjustable: BOOLEAN
		-- True if `right_adjust' will change the `count'
		do
			Result := not is_empty and then is_space_item (count)
		end

	prunable: BOOLEAN
			-- May items be removed? (Answer: yes.)
		do
			Result := True
		end

	valid_code (a_code: NATURAL_32): BOOLEAN
			-- Is `a_code' a valid code for a CHARACTER_32?
		do
			Result := True
		end

feature -- Conversion

	substring (start_index, end_index: INTEGER): like Current
			-- Copy of substring containing all characters at indices
			-- between `start_index' and `end_index'
		do
			if (1 <= start_index) and (start_index <= end_index) and (end_index <= count) then
				Result := new_string (end_index - start_index + 1)
				Result.area.copy_data (area, start_index - 1, 0, end_index - start_index + 1)
				Result.set_count (end_index - start_index + 1)
			else
				Result := new_string (0)
			end
			if has_unencoded_between_optimal (area, start_index, end_index)
				and then attached empty_unencoded_buffer as buffer
			then
				buffer.append_substring (Current, start_index, end_index, 0)
				Result.set_unencoded_from_buffer (buffer)
			end
		ensure then
			unencoded_valid: Result.is_valid
		end

	substring_to (uc: CHARACTER_32): like Current
		-- `substring_to_from' from start of string
		do
			Result := substring_to_from (uc, null)
		end

	substring_to_from (uc: CHARACTER_32; start_index_int32_ptr: TYPED_POINTER [INTEGER]): like Current
		-- substring from INTEGER at memory location `start_index_int32_ptr' up to but not including index of `uc'
		-- or else `substring_end (start_index)' if `uc' not found
		-- `start_index' is 1 if `start_index_int32_ptr = Default_pointer'
		-- write new start_index back to `start_index_int32_ptr'
		-- if `uc' not found then new `start_index' is `count + 1'
		local
			start_index, index: INTEGER; pointer: EL_POINTER_ROUTINES
		do
			if start_index_int32_ptr.is_default_pointer then
				start_index := 1
			else
				start_index := pointer.read_integer_32 (start_index_int32_ptr)
			end
			index := index_of (uc, start_index)
			if index > 0 then
				Result := substring (start_index, index - 1)
				start_index := index + 1
			else
				Result := substring_end (start_index)
				start_index := count + 1
			end
			if not start_index_int32_ptr.is_default_pointer then
				pointer.put_integer_32 (start_index, start_index_int32_ptr)
			end
		end

	substring_to_reversed (uc: CHARACTER_32): like Current
		-- `substring_to_reversed_from' from end of string
		do
			Result := substring_to_reversed_from (uc, null)
		end

	substring_to_reversed_from (uc: CHARACTER_32; start_index_from_end_ptr: TYPED_POINTER [INTEGER]): like Current
		-- the same as `substring_to' except going from right to left
		-- if `uc' not found `start_index_from_end' is set to `0' and written back to `start_index_from_end_ptr'
		local
			start_index_from_end, index: INTEGER; pointer: EL_POINTER_ROUTINES
		do
			if start_index_from_end_ptr.is_default_pointer then
				start_index_from_end := count
			else
				start_index_from_end := pointer.read_integer_32 (start_index_from_end_ptr)
			end
			index := last_index_of (uc, start_index_from_end)
			if index > 0 then
				Result := substring (index + 1, start_index_from_end)
				start_index_from_end := index - 1
			else
				Result := substring (1, start_index_from_end)
				start_index_from_end := 0
			end
			if not start_index_from_end_ptr.is_default_pointer then
				pointer.put_integer_32 (start_index_from_end, start_index_from_end_ptr)
			end
		end

feature -- Comparison

	is_equal (other: like Current): BOOLEAN
		local
			l_count: INTEGER; l_hash, l_other_hash: like internal_hash_code
		do
			if other = Current then
				Result := True
			else
				l_count := count
				if l_count = other.count then
						-- Let's compare the content if and only if the hash_code are the same or not yet computed.
					l_hash := internal_hash_code
					l_other_hash := other.internal_hash_code
					if l_hash = 0 or else l_other_hash = 0 or else l_hash = l_other_hash then
						Result := same_string (other)
					end
				end
			end
		end

feature {EL_ZSTRING_IMPLEMENTATION} -- Duplication

	copy (other: like Current)
		-- Reinitialize by copying the characters of `other'.
		-- (This is also used by `twin'.)
		local
			old_area: like area; last_upper: INTEGER
		do
			if other /= Current then
				old_area := area
				standard_copy (other)
				copy_area (old_area, other)
				make_unencoded_from_other (other)
				last_upper := unencoded_last_upper
				if last_upper > count and then attached empty_unencoded_buffer as buffer then
				-- in case was `other' was shortened by use of `set_count'
					buffer.append_substring (other, 1, count, 0)
					set_unencoded_from_buffer (buffer)
				end
				internal_hash_code := 0
			end
		ensure then
			new_result_count: count = other.count
			-- same_characters: For every `i' in 1..`count', `item' (`i') = `other'.`item' (`i')
		end

feature {STRING_HANDLER} -- Access

	frozen set_count (number: INTEGER)
		-- Set `count' to `number' of characters.

		-- NOTE: this is dangerous to set if `has_unencoded' is true
		-- but can be temporarily set to smaller value to look up hash table.
		-- See `ZSTRING_TEST_SET.test_hash_code'
		do
			count := number
			reset_hash
		end

feature -- Deferred

	current_zstring: ZSTRING
		deferred
		end

feature {NONE} -- Implementation

	current_readable: EL_READABLE_ZSTRING
		do
			Result := Current
		end

	new_filled_unencoded_area (uc: CHARACTER_32; n: INTEGER): like unencoded_area
		do
			create Result.make_filled (uc, n)
		end

	reset_hash
		do
			internal_hash_code := 0
			internal_case_insensitive_hash_code := 0
		end

	set_unencoded_area (a_unencoded_area: like unencoded_area)
		do
			unencoded_area := a_unencoded_area
		end

end