/**

 * vim: set ts=4 :

 * ===============================================================

 * SourceMod (C)2004-2007 AlliedModders LLC.  All rights reserved.

 * ===============================================================

 *

 * This file is not open source and may not be copied without explicit

 * written permission of AlliedModders LLC.  This file may not be redistributed 

 * in whole or significant part.

 * For information, see LICENSE.txt or http://www.sourcemod.net/license.php

 *

 * Version: $Id$

 */

#include <stdio.h>

#include <ctype.h>

#include <wctype.h>

#include <string.h>

#include <stdlib.h>

#include <assert.h>

#include "TextParsers.h"

#include "ShareSys.h"

TextParsers g_TextParser;

static int g_ini_chartable1[255] = {0};

static int g_ws_chartable[255] = {0};

bool IsWhitespace(const char *stream)

{

	return g_ws_chartable[(unsigned)*stream] == 1;

}

TextParsers::TextParsers()

{

	g_ini_chartable1[(unsigned)'_'] = 1;

	g_ini_chartable1[(unsigned)'-'] = 1;

	g_ini_chartable1[(unsigned)','] = 1;

	g_ini_chartable1[(unsigned)'+'] = 1;

	g_ini_chartable1[(unsigned)'.'] = 1;

	g_ini_chartable1[(unsigned)'$'] = 1;

	g_ini_chartable1[(unsigned)'?'] = 1;

	g_ini_chartable1[(unsigned)'/'] = 1;

	g_ws_chartable[(unsigned)'\n'] = 1;

	g_ws_chartable[(unsigned)'\v'] = 1;

	g_ws_chartable[(unsigned)'\r'] = 1;

	g_ws_chartable[(unsigned)'\t'] = 1;

	g_ws_chartable[(unsigned)'\f'] = 1;

	g_ws_chartable[(unsigned)' '] = 1;

}

void TextParsers::OnSourceModAllInitialized()

{

	g_ShareSys.AddInterface(NULL, this);

}

unsigned int TextParsers::GetUTF8CharBytes(const char *stream)

{

	return _GetUTF8CharBytes(stream);

}

/**

 * Character streams

 */

struct CharStream

{

	const char *curpos;

};

bool CharStreamReader(void *stream, char *buffer, size_t maxlength, unsigned int *read)

{

	CharStream *srdr = (CharStream *)stream;

	const char *ptr = srdr->curpos;

	for (size_t i=0; i<maxlength; i++)

	{

		if (*ptr == '\0')

		{

			break;

		}

		*buffer++ = *ptr++;

	}

	*read = ptr - srdr->curpos;

	srdr->curpos = ptr;

	return true;

}

SMCParseError TextParsers::ParseString_SMC(const char *stream, 

					 ITextListener_SMC *smc,

					 unsigned int *line,

					 unsigned int *col)

{

	CharStream srdr = { stream };

	return ParseStream_SMC(&srdr, CharStreamReader, smc, line, col);

}

/**

 * File streams

 */

bool FileStreamReader(void *stream, char *buffer, size_t maxlength, unsigned int *read)

{

	size_t num = fread(buffer, 1, maxlength, (FILE *)stream);

	*read = static_cast<unsigned int>(num);

	if (num == 0 && feof((FILE *)stream))

	{

		return true;

	}

	return (ferror((FILE *)stream) == 0);

}

SMCParseError TextParsers::ParseFile_SMC(const char *file, ITextListener_SMC *smc, unsigned int *line, unsigned int *col)

{

	FILE *fp = fopen(file, "rt");

	if (!fp)

	{

		return SMCParse_StreamOpen;

	}

	SMCParseError result = ParseStream_SMC(fp, FileStreamReader, smc, line, col);

	fclose(fp);

	return result;

}

/** 

 * Raw parsing of streams with helper functions

 */

struct StringInfo

{

	StringInfo() : quoted(false), ptr(NULL), end(NULL), special(false) { }

	bool quoted;

	char *ptr;

	char *end;

	bool special;

};

const char *FixupString(StringInfo &data)

{

	if (!data.ptr)

	{

		return NULL;

	}

	if (data.quoted)

	{

		data.ptr++;

	}

#if defined _DEBUG

	else {

		/* A string will never have beginning whitespace because we ignore it in the stream.

		 * Furthermore, if there is trailing whitespace, the end ptr will point to it, so it is valid

		 * to overwrite!  Lastly, the last character must be whitespace or a comment/invalid character.

		 */

 	}

#endif

	/* Do some extra work on strings that have special quoted characters. */

	if (data.special)

	{

		char *outptr = data.ptr;

		size_t len = data.end - data.ptr;

		if (len >= 2)

		{

			for (size_t i=0; i<len; i++)

			{

				if (data.ptr[i] == '\\' && i < len - 1)

				{

					/* Resolve the next character. */

					i++;

					if (data.ptr[i] == 'n')

					{

						data.ptr[i] = '\n';

					} else if (data.ptr[i] == 't') {

						data.ptr[i] = '\t';

					} else if (data.ptr[i] == 'r') {

						data.ptr[i] = '\r';

					} else if (data.ptr[i] != '\\'

						&& data.ptr[i] != '"') {

						/* This character is invalid, so go back one */

						i--;

					}

				}

				*outptr++ = data.ptr[i];

			}

			*outptr = '\0';

		}

	}

	*(data.end) = '\0';

	return data.ptr;

}

const char *rotate(StringInfo info[3])

{

	if (info[2].ptr != NULL)

	{

		return info[2].ptr;

	}

	if (info[0].ptr != NULL)

	{

		info[2] = info[1];

		info[1] = info[0];

		info[0] = StringInfo();

	}

	return NULL;

}

void scrap(StringInfo info[3])

{

	info[2] = StringInfo();

	info[1] = StringInfo();

	info[0] = StringInfo();

}

void reloc(StringInfo &data, unsigned int bytes)

{

	if (data.ptr)

	{

		data.ptr -= bytes;

	}

	if (data.end)

	{

		data.end -= bytes;

	}

}

char *lowstring(StringInfo info[3])

{

	for (int i=2; i>=0; i--)

	{

		if (info[i].ptr)

		{

			return info[i].ptr;

		}

	}

	return NULL;

}

SMCParseError TextParsers::ParseStream_SMC(void *stream, 

								   STREAMREADER srdr, 

								   ITextListener_SMC *smc, 

								   unsigned int *line, 

								   unsigned int *col)

{

	char in_buf[4096];

	char *reparse_point = NULL;

	char *parse_point = in_buf;

	char *line_begin = in_buf;

	unsigned int read;

	unsigned int curline = 1;

	unsigned int curtok = 0;

	unsigned int curlevel = 0;

	bool in_quote = false;

	bool ignoring = false;

	bool eol_comment = false;

	bool ml_comment = false;

	unsigned int i;

	SMCParseError err = SMCParse_Okay;

	SMCParseResult res;

	char c;

	StringInfo strings[3];

	StringInfo emptystring;

	smc->ReadSMC_ParseStart();

	while (srdr(stream, parse_point, sizeof(in_buf) - (parse_point - line_begin) - 1, &read))

	{

		if (!read)

		{

			break;

		}

		/* :TODO: do this outside of the main loop somehow

		 * This checks for BOM markings

		 */

		if (curline == 1 && 

			in_buf[0] == (char)0xEF && 

			in_buf[1] == (char)0xBB && 

			in_buf[2] == (char)0xBF)

		{

			parse_point = &in_buf[3];

		}

		if (reparse_point)

		{

			read += (parse_point - reparse_point);

			parse_point = reparse_point;

			reparse_point = NULL;

		}

		for (i=0; i<read; i++)

		{

			c = parse_point[i];

			if (c == '\n')

			{

				/* If we got a newline, there's a lot of things that could have happened in the interim.

				 * First, let's make sure the staged strings are rotated.

				 */

				if (strings[0].ptr)

				{

					strings[0].end = &parse_point[i];

					if (rotate(strings) != NULL)

					{

						err = SMCParse_InvalidTokens;

						goto failed;

					}

				}

				/* Next, let's clear some line-based values that may no longer have meaning */

				eol_comment = false;

				in_quote = false;

				if (ignoring && !ml_comment)

				{

					ignoring = false;

				}

				/* Pass the raw line onto the listener */

				if ((res=smc->ReadSMC_RawLine(line_begin, curline)) != SMCParse_Continue)

				{

					err = (res == SMCParse_HaltFail) ? SMCParse_Custom : SMCParse_Okay;

					goto failed;

				}

				/* Now we check the sanity of our staged strings! */

				if (strings[2].ptr)

				{

					if (!curlevel)

					{

						err = SMCParse_InvalidProperty1;

						goto failed;

					}

					/* Assume the next string is a property and pass the info on. */

					if ((res=smc->ReadSMC_KeyValue(

						FixupString(strings[2]),

						FixupString(strings[1]),

						strings[2].quoted,

						strings[1].quoted)) != SMCParse_Continue)

					{

						err = (res == SMCParse_HaltFail) ? SMCParse_Custom : SMCParse_Okay;

						goto failed;

					}

					scrap(strings);

				}

				/* Change the states for the next line */

				curtok = 0;

				curline++;

				line_begin = &parse_point[i+1];		//Note: safe because this gets relocated later

			} else if (ignoring) {

				if (in_quote)

				{

					/* If i was 0, this case is impossible due to reparsing */

					if ((i != 0) && c == '"' && parse_point[i-1] != '\\')

					{

						/* If we reached a quote in an ignore phase,

						 * we're staging a string and we must rotate it out.

						 */

						in_quote = false;

						ignoring = false;

						/* Set our info */

						strings[0].end = &parse_point[i];

						strings[0].quoted = true;

						if (rotate(strings) != NULL)

						{

							/* If we rotated too many strings, there was too much crap on one line */

							err = SMCParse_InvalidTokens;

							goto failed;

						}

					} else if (c == '\\') {

						strings[0].special = true;

						if (i == (read - 1))

						{

							reparse_point = &parse_point[i];

							break;

						}

					}

				} else if (ml_comment) {

					if (c == '*')

					{

						/* Check if we need to get more input first */

						if (i == read - 1)

						{

							reparse_point = &parse_point[i];

							break;

						}

						if (parse_point[i+1] == '/')

						{

							ml_comment = false;

							ignoring = false;

							/* We should not be staging anything right now. */

							assert(strings[0].ptr == NULL);

							/* Advance the input stream so we don't choke on this token */

							i++;

							curtok++;

						}

					}

				}

			} else {

				/* Check if we're whitespace or not */

				if (!g_ws_chartable[(unsigned)c])

				{

					bool restage = false;

					/* Check various special tokens:

					 * ;

					 * //

					 * / *

					 * {

					 * }

					 */

					if (c == ';' || c == '/')

					{

						/* If it's a line-based comment (that is, ; or //)

						 * we will need to scrap everything until the end of the line.

						 */

						if (c == '/')

						{

							if (i == read - 1)

							{

								/* If we reached the end of the look-ahead, we need to re-check our input.

								 * Breaking out will force this to be the new reparse point!

								 */

								reparse_point = &parse_point[i];

								break;

							}

							if (parse_point[i + 1] == '/')

							{

								/* standard comment */

								ignoring = true;

								eol_comment = true;

								restage = true;

							} else if (parse_point[i+1] == '*') {

								/* inline comment - start ignoring */

								ignoring = true;

								ml_comment = true;

								/* yes, we restage, meaning that:

								 * STR/ *stuff* /ING  (space because ml comments don't nest in C++)

								 * will not generate 'STRING', but rather 'STR' and 'ING'.

								 * This should be a rare occurrence and is done here for convenience.

								 */

								restage = true;

							}

						} else {

							ignoring = true;

							eol_comment = true;

							restage = true;

						}

					} else if (c == '{') {

						/* If we are staging a string, we must rotate here */

						if (strings[0].ptr)

						{

							/* We have unacceptable tokens on this line */

							if (rotate(strings) != NULL)

							{

								err = SMCParse_InvalidSection1;

								goto failed;

							}

						}

						/* Sections must always be alone */

						if (strings[2].ptr != NULL)

						{

							err = SMCParse_InvalidSection1;

							goto failed;

						} else if (strings[1].ptr == NULL) {

							err = SMCParse_InvalidSection2;

							goto failed;

						}

						if ((res=smc->ReadSMC_NewSection(FixupString(strings[1]), strings[1].quoted))

							!= SMCParse_Continue)

						{

							err = (res == SMCParse_HaltFail) ? SMCParse_Custom : SMCParse_Okay;

							goto failed;

						}

						strings[1] = emptystring;

						curlevel++;

					} else if (c == '}') {

						/* Unlike our matching friend, this can be on the same line as something prior */

						if (rotate(strings) != NULL)

						{

							err = SMCParse_InvalidSection3;

							goto failed;

						}

						if (strings[2].ptr)

						{

							if (!curlevel)

							{

								err = SMCParse_InvalidProperty1;

								goto failed;

							}

							if ((res=smc->ReadSMC_KeyValue(

											FixupString(strings[2]),

											FixupString(strings[1]),

											strings[2].quoted,

											strings[1].quoted))

								!= SMCParse_Continue)

							{

								err = (res == SMCParse_HaltFail) ? SMCParse_Custom : SMCParse_Okay;

								goto failed;

							}

						} else if (strings[1].ptr) {

							err = SMCParse_InvalidSection3;

							goto failed;

						} else if (!curlevel) {

							err = SMCParse_InvalidSection4;

							goto failed;

						}

						/* Now it's safe to leave the section */

						scrap(strings);

						if ((res=smc->ReadSMC_LeavingSection()) != SMCParse_Continue)

						{

							err = (res == SMCParse_HaltFail) ? SMCParse_Custom : SMCParse_Okay;

							goto failed;

						}

						curlevel--;

					} else if (c == '"') {

						/* If we get a quote mark, we always restage, but we need to do it beforehand */

						if (strings[0].ptr)

						{

							strings[0].end = &parse_point[i];

							if (rotate(strings) != NULL)

							{

								err = SMCParse_InvalidTokens;

								goto failed;

							}

						}

						strings[0].ptr = &parse_point[i];

						in_quote = true;

						ignoring = true;

					} else if (!strings[0].ptr) {

						/* If we have no string, we must start one */

						strings[0].ptr = &parse_point[i];

					}

					if (restage && strings[0].ptr)

					{

						strings[0].end = &parse_point[i];

						if (rotate(strings) != NULL)

						{

							err = SMCParse_InvalidTokens;

							goto failed;

						}

					}

				} else {

					/* If we're eating a string and get whitespace, we need to restage.

					 * (Note that if we are quoted, this is being ignored)

					 */

					if (strings[0].ptr)

					{

						/*

						 * The specification says the second string in a pair does not need to be quoted.

						 * Thus, we check if there's already a string on the stack.

						 * If there's a newline, we always rotate so the newline has an empty starter.

						 */

						if (!strings[1].ptr)

						{

							/* There's no string, so we must move this one down and eat up another */