rtf2html.c

harvest是一个下载html网页得机器人

	if(feof(f))
	{
		fprintf(stderr,"Unexpected end of file\n");
		return;
	}
	switch (ch)
	{
		case '\\':
		case '{':
		case '}':
			*pch=ch; *pf=TRUE;
			break;
		case '*':
			gobble = TRUE;	/*perform no output, ignore commands 'til level-1*/
			if(skip_to_level>level-1||skip_to_level==-1)
				skip_to_level = level-1;
			break;
		case '\'':
		{
			char ch1, ch2;
			ch1 = RTF_GetChar();
			ch2 = RTF_GetChar();
			if(!feof(f))
			{
				if(isxdigit(ch1)&&isxdigit(ch2))
				{
					ch = chartoi(ch1)*16+chartoi(ch2);
					*pch=ch; *pf=TRUE;
				} else {
					fprintf(stderr,"RTF Error: unexpected '%c%c' after \\\'\n",ch1,ch2);
				}
			}
			break;
		}
		default:
			if (isalpha(ch))
			{
				RTF_BuildToken(ch);
			} else {
				fprintf(stderr, "\nRTF Error: unexpected '%c' after \\.\n", ch);
			}
			break;
	}
}

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

void RTF_ParseStyle(void)
{
	char ch, pf;
	int level0;
	void (*PrevDoControl)(char*,char*);

	level0=level;
	PrevDoControl=RTF_DoControl;
	RTF_DoControl=RTF_DoStyleControl;

	RTF_ClearName(style_name);
	style_number[0]=0;
	while (1)
	{
		ch = RTF_GetChar();
		if(feof(f))
			break;
		switch (ch)
		{
			case '\\':
				RTF_backslash(&ch,&pf);
				if(pf)
				{
					RTF_BuildName(style_name,ch);
				} else {
					RTF_ClearName(style_name);
				}
				break;

			case '{':
				level++;
				RTF_ClearName(style_name);
				break;

			case '}':
				if(level0+1==level)
				{
					if(style_number[0]!=0)
					{
						RTF_AddStyleMap(style_name,style_number);
						style_number[0]=0;
					}
				} else if(level0==level) {
					RTF_DoControl=PrevDoControl;
					RTF_UnGetChar(ch);
					return;
				}
				level--;
				RTF_ClearName(style_name);
				break;

			default:
				RTF_BuildName(style_name,ch);
				break;
		}
	} /* while */
}

char isField = FALSE;
char isHyperLink = FALSE;

void RTF_IgnoreControl(char* control, char* arg)
{
	(void)control; (void)arg;
}

void ParseFldInst(void)
{
	static const char* kwd= "HYPERLINK";
	short level0;
	int i;
	void (*PrevDoControl)(char*,char*);
	char ch, pf, endc;

	if(!isField||isHyperLink)
		return;
	/* look for HYPERLINK */
	level0 = level-1;
	i = 0;
	while (1)
	{
		ch = RTF_GetChar();
		if(feof(f))
		{
			fprintf(stderr, "Unexpected end of file\n");
			break;
		}
		switch (ch)
		{
			case '\\':
				RTF_backslash(&ch,&pf);
				if(pf)
					i=-1;
				else
					i=0;
				break;

			case '{':
				RTF_PushState(&level);
				i=0;
				break;

			case '}':
				RTF_PopState(&level);
				if(level0==level)
					return;
				i=0;
				break;

			default:
				if(i<0)
				{
					if(isspace(ch))
						i=0;
				} else {
					if(kwd[i]==ch)
						i++;
					else if(kwd[i]==0&&isspace(ch))
						goto found;
					else if(isspace(ch))
						i=0;
					else
						i=-1;
				}

				break;
		}
	}/*while*/
found:
	gobble = FALSE;
	isHyperLink = TRUE;
	PrevDoControl=RTF_DoControl;
	RTF_DoControl=RTF_IgnoreControl;
	RTF_PutStr("<a href=\"");
	ch = RTF_GetChar();
	if(ch=='"')
	{
		endc = ch;
		ch = RTF_GetChar();
	} else {
		endc = 0;
	}
	level0=level;
	while(1)
	{
		if(feof(f))
		{
			fprintf(stderr, "Unexpected end of file\n");
			break;
		}
		if((level==level0) && ((endc && endc==ch) || isspace(ch)))
			break;
		switch(ch)
		{
			case '\\':
				RTF_backslash(&ch,&pf);
				if(!pf)
					goto next;
				break;
			case '{':
				level++;
				break;
			case '}':
				if(level==level0)
					goto done;
				--level;
				break;
		}
		if(ch=='"')
			RTF_PutStr("&#34;");
		else
			RTF_PutChar(ch);
next:
		ch = RTF_GetChar();
	}
done:
	RTF_UnGetChar(ch);
	RTF_PutStr("\">");
	RTF_DoControl=PrevDoControl;
	gobble = TRUE;
}

void ParseField(void)
{
	short level0;
	char ch, pf;
	if(isField)
	{
		return;
	}

	level0 = level-1;
	isField = TRUE;
	while (1)
	{
		ch = RTF_GetChar();
		if(feof(f))
		{
			fprintf(stderr, "Unexpected end of file\n");
			break;
		}
		switch (ch)
		{
			case '\\':
				RTF_backslash(&ch,&pf);
				if(pf) RTF_PutChar(ch);
				break;

			case '{':
				RTF_PushState(&level);
				break;

			case '}':
				RTF_PopState(&level);
				if(level0==level)
				{
					if(isHyperLink)
					{
						char save = gobble;
						gobble = FALSE;
						RTF_PutStr("</a>");
						gobble = save;
						isHyperLink = FALSE;
					}
					isField = FALSE;
					return;
				}
				break;

			default:
				RTF_PutChar(ch);
				break;
		}
	}/*while*/
}

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

void RTF_DoParseInfo(void)
{
	char ch, pf;
	int level0;
	level0=level;

	while (1)
	{
		ch = RTF_GetChar();
		if(feof(f))
			break;
		switch (ch)
		{
			case '\\':
				RTF_backslash(&ch,&pf);
				if(pf)
					RTF_PutChar(ch);
				break;

			case '{':
				level++;
				break;

			case '}':
				if(level==level0)
				{
					RTF_UnGetChar(ch);
					return;
				}
				--level;
				break;

			default:
				RTF_PutChar(ch);
				break;
		}
	} /* while */
}

void RTF_DoInfoControl(char* control, char* arg)
{
	if(arg[0]==0 && !strcasecmp(control,"title"))
	{
		gobble = FALSE;
		RTF_DoControl=RTF_IgnoreControl;

		RTF_PutStr("<TITLE>");
		RTF_DoParseInfo();
		RTF_PutStr("</TITLE>\n");

		RTF_DoControl=RTF_DoInfoControl;
		gobble = TRUE;
	}
}

void RTF_ParseInfo(void)
{
	void (*PrevDoControl)(char*,char*);
	short prev_gobble;

	prev_gobble = gobble;
	gobble = TRUE; /* skip all unknown information */
	PrevDoControl=RTF_DoControl;

	if(isBody)
	{
		/* too late for info, just skip it */
		RTF_DoControl=RTF_IgnoreControl;
		RTF_DoParseInfo();
		return;
	} else {
		if(!isHead) RTF_StartHead();
		RTF_DoControl=RTF_DoInfoControl;
		isBody=TRUE; /* to prevent RTF_Put from start a body */
		RTF_DoParseInfo();
		isBody=FALSE;
	}

	RTF_DoControl=PrevDoControl;
	gobble = prev_gobble;
}


/**************************************/
/* Perform actions for RTF control words */

void RTF_DoBodyControl (char* control,char* arg)
{
	short style;

	switch (GetTokenIndex(control))
	{
		case t_stylesheet:
			gobble = TRUE;	/*perform no output, ignore commands 'til level-1*/
			skip_to_level = level-1;
			RTF_ParseStyle();
			break;
		case t_fonttbl:	/*skip all of these and their contents!*/
		case t_colortbl:
			gobble = TRUE;	/*perform no output, ignore commands 'til level-1*/
			skip_to_level = level-1;
			break;
		case t_info:
			RTF_ParseInfo();
			break;
		case t_field:
			ParseField();
			break;
		case t_fldinst:
			ParseFldInst();
			break;
		case t_pict:
			gobble = TRUE;	/*perform no output, ignore commands 'til level-1*/
			if(skip_to_level>=level || skip_to_level==-1)
				skip_to_level = level-1;
			break;


		case t_s: /*Style*/
			if (!curr_heading)
			{
				style = RTF_MapStyle (arg);
				if(style)
				{
					curr_heading = s_h0 + style;
					RTF_PutStr(styles[curr_heading][0]);
					ignore_styles = TRUE;
				}
			}
			break;

		case t_b: /*Bold*/
				RTF_SetStyle(&curr_style,s_bold);
			break;

		case t_ulw:
		case t_uld:
		case t_uldb:
		case t_ul: /*Underline, maps to "emphasis" HTML style*/
				if(!isHyperLink)
					RTF_SetStyle(&curr_style,s_underline);
			break;

		case t_i: /*Italic*/
				RTF_SetStyle(&curr_style,s_italic);
			break;

		case t_v: /* Hidden*/
				RTF_SetStyle(&curr_style,s_hidden);
			break;

		case t_par: /*Paragraph*/
			if (curr_heading!=s_plain) {
				RTF_PutStr(styles[curr_heading][1]);
				curr_heading = s_plain;
			} else {
				RTF_PutStr(styles[s_para][0]);
			}
			ignore_styles = FALSE;
			break;

		case t_plain: /*reset inline styles*/
			RTF_PlainStyle(&curr_style);
			break;
		case t_cell:
		case t_tab:
			RTF_PutChar(' ');
			break;
		case t_endash:
		case t_emdash:
			RTF_PutChar('-');
			break;
		case t_line:
		case t_row:
			RTF_PutStr(styles[s_br][0]);
			break;
		case t_bullet:
			RTF_PutChar('\xb7');
			break;
		case t_start:
		case t_end:
			break;
	}

}

/**************************************/
/* RTF_Parse is a crude, ugly state machine that understands enough of */
/* the RTF syntax to be dangerous.                                     */

void RTF_ParseBody(void)
{
	char ch, pf;

	RTF_DoControl=RTF_DoBodyControl;
	level = 0;
	skip_to_level = -1;
	gobble = FALSE;
	ignore_styles = FALSE;

	while (1)
	{
		ch = RTF_GetChar();
		if(feof(f))
			break;
		switch (ch)
		{
			case '\\':
				RTF_backslash(&ch,&pf);
				if(pf) RTF_PutCharEx(ch);
				break;

			case '{':
				RTF_PushState(&level);
				break;

			case '}':
				RTF_PopState(&level);
				break;

			default:
				RTF_PutCharEx(ch);
				break;
		}
	}/*while*/
}

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

int RTF_Parse (char* filename)
{
	if (filename)
	{
		if (!(f = fopen (filename, "r")))
		{
			fprintf (stderr, "\nError: Input file %s not found.\n", filename);
			return (-1);
		}
	} else {
		f = stdin;
	}

	isHead=isBody=FALSE;
	RTF_ParseBody();
	if(isBody) RTF_PutStr("</BODY>\n");
	RTF_PutStr("</HTML>\n");

	fclose (f);
	return 0;
}

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

void Initialize(void)
{
	int i;

	for (i=0;i<MAX_LEVELS;i++)
			style_state[i].s=s_plain;

	curr_style.s=s_plain;
	curr_heading = s_plain;

	// Set default styles maping
	style_mappings[0][0]=0;
	for(i=1;i<=6;i++)
			sprintf(style_mappings[i],"%d",256-i);
}

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

int main(int argc,char** argv)
{

	char *in;

	if (argc>1)
	{
		if( strcmp(argv[1],"--help")==0 || strcmp(argv[1],"-H")==0 )
		{
			printf("Use: %s [rtf_filename]\n",argv[0]);
			return 0;
		} else if ( strcmp(argv[1],"--version")==0 || strcmp(argv[1],"-V")==0 ) {
			printf("rtf2html version 1.2\n");
			return 0;
		} else
			in = argv[1];
	} else
		in = NULL;

	Initialize();

	return (RTF_Parse(in));
}

/* Note, the source is formated with 4 character tabs
 * vim: sw=4 ts=4
 */