/**

 * Relational pipes

 * Copyright © 2020 František Kučera (Frantovo.cz, GlobalCode.info)

 *

 * This program is free software: you can redistribute it and/or modify

 * it under the terms of the GNU General Public License as published by

 * the Free Software Foundation, version 3 of the License.

 *

 * This program is distributed in the hope that it will be useful,

 * but WITHOUT ANY WARRANTY; without even the implied warranty of

 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the

 * GNU General Public License for more details.

 *

 * You should have received a copy of the GNU General Public License

 * along with this program. If not, see <http://www.gnu.org/licenses/>.

 */

#include <vector>

#include <regex>

#include <sstream>

#include <stdexcept>

#include "INIReader.h"

namespace relpipe {

namespace in {

namespace ini {

namespace lib {

class INIReaderImpl : public INIReader {

private:

	std::istream& input;

	std::vector<INIContentHandler*> handlers;

	/** 

	 * By default, we ignore all leading whitespace on continuing lines.

	 * If there should be some spaces or tabs, they should be placed on the previous line before the „\“.

	 * If a line break is desired, it should be written as \n (escaped) or the value should be quoted in " or '.

	 * 

	 * TODO: several options:

	 *  - enabled, disabled

	 *  - if disabled, then: keep backslash, trim backslash, escape backslash

	 *    (keep requires support in some further unescaping phase, or it will cause an error)

	 *  - keep or trim the line end

	 *  - keep or trim the leading spaces

	 *  - allow comments interleaved with continuing lines (the freaky systemd syntax)

	 * 

	 * Related specifications:

	 *  - https://docs.oracle.com/javase/8/docs/api/index.html?java/util/Properties.html

	 *  - https://www.freedesktop.org/software/systemd/man/systemd.syntax.html

	 */

	bool trimLeadingSpacesOnContinuingLines = true;

	/**

	 * Some dialects or configuration files in general does not support sections.

	 * Then a line, that looks like an INI section, should be interpreted as a key

	 * (or error, if does not have a proper key-value separator).

	 */

	bool allowSections = true;

	/**

	 * KDE uses some weird INI dialect that allows [section][x] syntax where „x“ is kind of „tag“ that signalizes some properties of given section.

	 * Line „[section_1][$i]“ means that the „section_1“ is „locked“.

	 * We may emit this information somehow later, but for now, it is just ignored.

	 * 

	 * TODO: Is „section tag“ right name?

	 * 

	 * Related specifications:

	 *  - https://userbase.kde.org/KDE_System_Administration/Configuration_Files#Lock_Down

	 */

	bool allowSectionTags = true;

	/**

	 * If whole key is „aaa[bbb]“ then „aaa“ is considered to be the key and „bbb“ the sub-key.

	 * No \[ escaping is currently supported, so the key might not contain the bracket character.

	 * 

	 * Related specifications:

	 *  - https://userbase.kde.org/KDE_System_Administration/Configuration_Files#Shell_Expansion

	 *  - https://specifications.freedesktop.org/desktop-entry-spec/latest/ar01s05.html

	 */

	bool allowSubKeys = true;

	/**

	 * Classic INI uses „key=value“ syntax.

	 * But some other formats/dialects might use key:value.

	 * 

	 * Only single character separators are supported.

	 * If multiple separators should be recognized (e.g. both „=“ and „:“), this string will contain all of them,

	 * i.e. „:=“ does not mean that the „key:=value“ syntax, but „key=value“ or „key:value“.

	 */

	std::string keyValueSeparators = "=";

	/**

	 * Classic INI uses „; comment“ syntax.

	 * But many existing files contain „# comment“ lines.

	 * 

	 * Only single character separators are supported (works same as keyValueSeparators).

	 */

	std::string commentSeparators = ";#";

	/**

	 * INI often support both "quotes" and 'apostrophes' styles.

	 * But some dialects may support only one of them or not support quoting at all.

	 * 

	 * In such case e.g. „key="some value"“ would mean that the value is „"value"“ (including the quotes).

	 * Thus it is important to allow disabling quote recognizing (which is done by setting this parameter to empty string).

	 * 

	 * Only single character quotes are supported (works same as keyValueSeparators).

	 */

	std::string quotes = "\"'";

	int lineNumber = 1;

	int eventNumber = 0;

	/**

	 * Should be always used instead of input.peek().

	 * Skips \r.

	 */

	char peek() {

		// In 2020 there is no need to manually return the carriage. However some legacy systems still do it.

		char ch = input.peek();

		if (ch == '\r') {

			input.get();

			ch = input.peek();

		}

		return ch;

	}

	/**

	 * Should be always used instead of input.get().

	 * Counts the lines and skips \r.

	 */

	char get() {

		char ch = input.get();

		if (ch == '\n') lineNumber++;

		else if (ch == '\r') ch = get();

		return ch;

	}

	std::string readSpacesAndTabs() {

		std::stringstream result;

		for (char ch = peek(); input.good() && (ch == ' ' || ch == '\t'); ch = peek()) result.put(get());

		return result.str();

	}

	std::string readAllWhitespace() {

		std::stringstream result;

		for (char ch = peek(); input.good() && (ch == ' ' || ch == '\t' || ch == '\n' || ch == '\r'); ch = peek()) result.put(get());

		return result.str();

	}

	void processContinuingLine(std::stringstream& result) {

		if (trimLeadingSpacesOnContinuingLines) readSpacesAndTabs();

		else result.put('\n');

	}

	std::string readUntil(const char until, bool* found = nullptr) {

		return readUntil(std::string(1, until), found);

	}

	std::string readUntil(const std::string& until, bool* found = nullptr) {

		std::stringstream result;

		for (char ch = peek(); input.good() && !oneOf(ch, until); ch = peek()) {

			if (ch == '\\') {

				get();

				ch = get();

				if (oneOf(ch, until) && ch == '\n') processContinuingLine(result);

				else if (oneOf(ch, until)) result.put(ch);

				else if (ch == std::istream::traits_type::eof()) break;

				else result.put('\\').put(ch);

				// unescaping is done in two phases:

				// here we unescape just the \n (LF)

				// other escape sequences are leaved untouched and will be processed in later phases, see see UnescapingINIHandler

			} else {

				ch = get();

				result.put(ch);

			}

		}

		if (oneOf(peek(), until)) {

			get();

			if (found) *found = true;

		} else {

			if (found) *found = false;

		}

		return result.str();

	}

	std::string readToken(const char until, char* quote = nullptr, bool* found = nullptr) {

		return readToken(std::string(1, until), quote, found);

	}

	std::string readToken(const std::string& until, char* quote = nullptr, bool* found = nullptr) {

		std::string result;

		char ch = peek();

		if (isQuote(ch)) {

			if (quote) *quote = ch;

			result = readUntil(std::string(1, get()), found);

		} else {

			if (quote) *quote = 0;

			result = readUntil(until, found);

		}

		return result;

	}

	std::string readTokenAndEatTerminator(char until, char* quote = nullptr, bool* found = nullptr) {

		return readTokenAndEatTerminator(std::string(1, until), quote, found);

	}

	std::string readTokenAndEatTerminator(const std::string& until, char* quote = nullptr, bool* found = nullptr) {

		std::string result = readToken(until, quote, found);

		if (*quote) {

			readAllWhitespace();

			if (!oneOf(get(), until)) throw std::logic_error(std::string("missing „") + until + "“ after quoted section name");

		}

		return result;

	}

	bool isComment(char ch) {

		return oneOf(ch, commentSeparators);

	}

	bool isQuote(char ch) {

		return oneOf(ch, quotes);

	}

	/**

	 * @param ch character to be evaluated

	 * @param options list of options (characters)

	 * @return whether ch is one of options

	 */

	bool oneOf(char ch, const std::string& options) {

		return options.find(ch) != std::string::npos;

	}

	std::string trim(std::string s) {

		return std::regex_replace(s, std::regex("^\\s+|\\s+$"), "");

	}

	/**

	 * TODO: use a common method

	 */

	bool parseBoolean(const std::string& value) {

		if (value == "true") return true;

		else if (value == "false") return false;

		else throw std::invalid_argument(std::string("Unable to parse boolean value: ") + value + " (expecting true or false)");

	}

	void setDialect(const std::string& name) {

		if (name == "default-ini") {

			// already set

		} else if (name == "java-properties") {

			trimLeadingSpacesOnContinuingLines = true;

			allowSections = false;

			allowSectionTags = false;

			allowSubKeys = false;

			commentSeparators = "#";

			keyValueSeparators = "=:";

			quotes = "";

			// TODO: enable unicode unescaping

		} else {

			throw std::invalid_argument(std::string("Unsupported INI dialect: ") + name);

		}

	}

public:

	INIReaderImpl(std::istream& input) : input(input) {

	}

	void setOption(const std::string& uri, const std::string& value) override {

		if (uri == "trim-continuing-lines") trimLeadingSpacesOnContinuingLines = parseBoolean(value); // TODO: continuing lines modes (enum), not just boolean

		else if (uri == "allow-sections") allowSections = parseBoolean(value);

		else if (uri == "allow-section-tags") allowSectionTags = parseBoolean(value);

		else if (uri == "allow-sub-keys") allowSubKeys = parseBoolean(value);

		else if (uri == "comment-separators") commentSeparators = value;

		else if (uri == "key-value-separators") keyValueSeparators = value;

		else if (uri == "quotes") quotes = value;

		else if (uri == "dialect") setDialect(value);

		else throw std::invalid_argument(std::string("Invalid parser option: „") + uri + "“ with value: „" + value + "“");

	}

	void addHandler(INIContentHandler* handler) override {

		handlers.push_back(handler);

	}

	void process() override {

		for (INIContentHandler* handler : handlers) handler->startDocument();

		bool inSection = false;

		while (input.good()) { // TODO: condition

			{

				INIContentHandler::WhitespaceEvent event;

				event.lineNumber = lineNumber;

				std::string whitespace = readAllWhitespace();

				if (whitespace.size()) {

					event.eventNumber = ++eventNumber;

					event.whitespace = whitespace;

					for (INIContentHandler* handler : handlers) handler->whitespace(event);

				}

			}

			bool found;

			char quote;

			char ch = peek();

			if (ch == std::istream::traits_type::eof()) {

				break;

			} else if (ch == '[' && allowSections) {

				if (inSection) for (INIContentHandler* handler : handlers) handler->endSection();

				inSection = true;

				INIContentHandler::SectionStartEvent event;

				event.lineNumber = lineNumber;

				event.eventNumber = ++eventNumber;

				get();

				readAllWhitespace();

				event.name = readTokenAndEatTerminator(']', &quote, &found);

				if (!quote) event.name = trim(event.name);

				readSpacesAndTabs();

				if (allowSectionTags && peek() == '[') {

					get();

					event.tag = readTokenAndEatTerminator(']', &quote, &found);

				}

				readSpacesAndTabs();

				ch = peek();

				if (isComment(ch)) {

					get();

					readSpacesAndTabs();

					event.comment = readUntil('\n', &found);

				} else if (ch == '\n') {

					get();

				} else {

					throw std::logic_error(std::string("unexpected content after the section: '") + event.name + "'");

				}

				for (INIContentHandler* handler : handlers) handler->startSection(event);

			} else if (isComment(ch)) {

				INIContentHandler::CommentEvent event;

				event.lineNumber = lineNumber;

				event.eventNumber = ++eventNumber;

				get();

				readSpacesAndTabs();

				event.comment = readUntil('\n', &found);

				for (INIContentHandler* handler : handlers) handler->comment(event);

			} else {

				INIContentHandler::EntryEvent event;

				event.lineNumber = lineNumber;

				event.eventNumber = ++eventNumber;

				std::string fullKey = readToken(keyValueSeparators, &quote, &found);

				if (!found) throw std::logic_error(std::string("missing = after key: '") + fullKey + "'");

				if (!quote) fullKey = trim(fullKey);

				readSpacesAndTabs();

				if (quote) {

					ch = get();

					if (oneOf(ch, keyValueSeparators)) readSpacesAndTabs();

					else throw std::logic_error(std::string("missing = after quoted key: '") + fullKey + "'");

				}

				std::string value = readToken('\n', &quote, &found);

				if (!quote) value = trim(value);

				event.key = fullKey;

				event.fullKey = fullKey;

				event.value = value;

				if (allowSubKeys) {

					std::smatch match;

					if (std::regex_match(fullKey, match, std::regex("([^\\[]+)\\[([^\\[]+)\\]"))) {

						event.key = match[1];

						event.subKey = match[2];

						event.fullKey = fullKey;

					}

				}

				if (quote) {

					readSpacesAndTabs();

					ch = peek();

					if (isComment(ch)) {

						get();

						readSpacesAndTabs();

						event.comment = readUntil('\n', &found);

					} else if (ch == '\n') {

						get();

					} else {

						// TODO: optional support for multiple tokens in a single entry?

						// modes: array, concatenate

						// some-array-1 = "item 1" "item 2" 'item 3' item 4

						// some-array-2 = "item 1" "item 2" 'item 3' item_4 item_5

						// some-bash-style-string-value = "this "will' be' concatenated → this will be concatenated

						throw std::logic_error(std::string("unexpected content after the quoted value: key='") + fullKey + "' value='" + event.value + "'");

					}

				}

				for (INIContentHandler* handler : handlers) handler->entry(event);

			}

		}

		// TODO: error at the end, catch premature/unexpected EOF

		// TODO: unescape + trim values + ignore \r

		// TODO: count lines

		if (inSection) for (INIContentHandler* handler : handlers) handler->endSection();

		for (INIContentHandler* handler : handlers) handler->endDocument();

	}

	// General feautres:

	// TODO: warning/error handler

	// TODO: support also escaped characters

	// TODO: support also Java .properties and manifest.mf formats?

	// TODO: support also nested sections – hierarchy

	// TODO: support also nested keys e.g. key.sub.subsub.subsubsub=value – translate them to nested sections

	// TODO: support also option for alternative key-value separator (: instead of =)

	// TODO: support also other encodings (currently only UTF-8 is supported)

	// TODO: better exceptions

	// Lossless conversions:

	// TODO: emit also the quote style ('/"/)

	// TODO: emit also the comment style (;/#) ?

	// TODO: emit also the whitespace before key name, around =, after "values"/'values', around [sections] ?

	// TODO: emit also the line-end type (LF/CRLF) ?

};

INIReader* INIReader::create(std::istream& input) {

	return new INIReaderImpl(input);

}

}

}

}

}