4 * Functions which handle translation between HTML and plain text
5 * Copyright (c) 2000-2001 by Art Cancro and others. This program is
6 * released under the terms of the GNU General Public License.
20 #if TIME_WITH_SYS_TIME
21 # include <sys/time.h>
25 # include <sys/time.h>
37 #include "serv_extensions.h"
39 #include "sysdep_decls.h"
49 * Convert HTML to plain text.
51 * inputmsg = pointer to raw HTML message
52 * screenwidth = desired output screenwidth
53 * do_citaformat = set to 1 to indent newlines with spaces
55 char *html_to_ascii(char *inputmsg, int screenwidth, int do_citaformat) {
62 size_t outptr_buffer_size;
63 size_t output_len = 0;
64 int i, j, ch, did_out, rb, scanch;
65 int nest = 0; /* Bracket nesting level */
66 int blockquote = 0; /* BLOCKQUOTE nesting level */
72 outptr_buffer_size = strlen(inptr) + SIZ;
73 outptr = malloc(outptr_buffer_size);
74 if (outptr == NULL) return NULL;
79 /* Fill the input buffer */
80 if ( (done_reading == 0) && (strlen(inbuf) < (SIZ-128)) ) {
84 inbuf[strlen(inbuf)+1] = 0;
85 inbuf[strlen(inbuf)] = ch;
94 if (strlen(inbuf)>0) {
96 /* Fold in all the spacing */
97 for (i=0; i<strlen(inbuf); ++i) {
98 if (inbuf[i]==10) inbuf[i]=32;
99 if (inbuf[i]==13) inbuf[i]=32;
100 if (inbuf[i]==9) inbuf[i]=32;
101 if ((inbuf[i]<32) || (inbuf[i]>126)) {
105 for (i=0; i<strlen(inbuf); ++i) {
106 while ((inbuf[i]==32)&&(inbuf[i+1]==32))
107 strcpy(&inbuf[i], &inbuf[i+1]);
110 for (i=0; i<strlen(inbuf); ++i) {
119 else if (ch == '>') { /* We have a tag. */
120 if (nest > 0) --nest;
122 /* Unqualify the tag (truncate at first space) */
123 if (strchr(tag, ' ') != NULL) {
124 strcpy(strchr(tag, ' '), "");
127 if (!strcasecmp(tag, "P")) {
128 strcat(outbuf, "\n\n");
131 if (!strcasecmp(tag, "/DIV")) {
132 strcat(outbuf, "\n\n");
135 if (!strcasecmp(tag, "LI")) {
136 strcat(outbuf, "\n * ");
139 else if (!strcasecmp(tag, "/UL")) {
140 strcat(outbuf, "\n\n");
143 else if (!strcasecmp(tag, "H1")) {
144 strcat(outbuf, "\n\n");
147 else if (!strcasecmp(tag, "H2")) {
148 strcat(outbuf, "\n\n");
151 else if (!strcasecmp(tag, "H3")) {
152 strcat(outbuf, "\n\n");
155 else if (!strcasecmp(tag, "H4")) {
156 strcat(outbuf, "\n\n");
159 else if (!strcasecmp(tag, "/H1")) {
160 strcat(outbuf, "\n");
163 else if (!strcasecmp(tag, "/H2")) {
164 strcat(outbuf, "\n");
167 else if (!strcasecmp(tag, "/H3")) {
168 strcat(outbuf, "\n");
171 else if (!strcasecmp(tag, "/H4")) {
172 strcat(outbuf, "\n");
175 else if (!strcasecmp(tag, "HR")) {
176 strcat(outbuf, "\n ");
177 for (j=0; j<screenwidth-2; ++j)
179 strcat(outbuf, "\n");
182 else if (!strcasecmp(tag, "BR")) {
183 strcat(outbuf, "\n");
186 else if (!strcasecmp(tag, "TR")) {
187 strcat(outbuf, "\n");
190 else if (!strcasecmp(tag, "/TABLE")) {
191 strcat(outbuf, "\n");
194 else if (!strcasecmp(tag, "BLOCKQUOTE")) {
195 strcat(outbuf, "\n\n <<\n");
199 else if (!strcasecmp(tag, "/BLOCKQUOTE")) {
200 strcat(outbuf, "\n >>\n\n");
206 else if ((nest > 0) && (strlen(tag)<(sizeof(tag)-1))) {
207 tag[strlen(tag)+1] = 0;
208 tag[strlen(tag)] = ch;
212 outbuf[strlen(outbuf)+1] = 0;
213 outbuf[strlen(outbuf)] = ch;
216 strcpy(inbuf, &inbuf[i]);
219 /* Convert &; tags to the forbidden characters */
220 if (strlen(outbuf)>0) for (i=0; i<strlen(outbuf); ++i) {
222 if (!strncasecmp(&outbuf[i], " ", 6)) {
224 strcpy(&outbuf[i+1], &outbuf[i+6]);
227 else if (!strncasecmp(&outbuf[i], "<", 4)) {
229 strcpy(&outbuf[i+1], &outbuf[i+4]);
232 else if (!strncasecmp(&outbuf[i], ">", 4)) {
234 strcpy(&outbuf[i+1], &outbuf[i+4]);
237 else if (!strncasecmp(&outbuf[i], "&", 5)) {
238 strcpy(&outbuf[i+1], &outbuf[i+5]);
241 else if (!strncasecmp(&outbuf[i], """, 6)) {
243 strcpy(&outbuf[i+1], &outbuf[i+6]);
246 else if (!strncasecmp(&outbuf[i], "©", 6)) {
250 strcpy(&outbuf[i+3], &outbuf[i+6]);
253 else if (!strncasecmp(&outbuf[i], "®", 5)) {
257 strcpy(&outbuf[i+3], &outbuf[i+5]);
260 /* two-digit decimal equivalents */
261 else if ((!strncmp(&outbuf[i], "&#", 2))
262 && (outbuf[i+4] == ';') ) {
264 sscanf(&outbuf[i+2], "%02d", &scanch);
266 strcpy(&outbuf[i+1], &outbuf[i+5]);
269 /* three-digit decimal equivalents */
270 else if ((!strncmp(&outbuf[i], "&#", 2))
271 && (outbuf[i+5] == ';') ) {
273 sscanf(&outbuf[i+2], "%03d", &scanch);
275 strcpy(&outbuf[i+1], &outbuf[i+6]);
280 /* Make sure the output buffer is big enough */
281 if ((output_len + strlen(outbuf) + SIZ)
282 > outptr_buffer_size) {
283 outptr_buffer_size += SIZ;
284 outptr = realloc(outptr, outptr_buffer_size);
287 /* Output any lines terminated with hard line breaks */
290 if (strlen(outbuf)>0) {
291 for (i = 0; i<strlen(outbuf); ++i) {
292 if ( (i<(screenwidth-2)) && (outbuf[i]=='\n')) {
294 strncpy(&outptr[output_len],
299 strcpy(&outptr[output_len],
304 strcpy(outbuf, &outbuf[i+1]);
312 /* Add soft line breaks */
313 if (strlen(outbuf) > (screenwidth - 2 )) {
315 for (i=0; i<(screenwidth-2); ++i) {
316 if (outbuf[i]==32) rb = i;
319 strncpy(&outptr[output_len], outbuf, rb);
321 strcpy(&outptr[output_len], "\n");
324 strcpy(&outptr[output_len], " ");
327 strcpy(outbuf, &outbuf[rb+1]);
329 strncpy(&outptr[output_len], outbuf,
331 output_len += (screenwidth-2);
332 strcpy(&outptr[output_len], "\n");
335 strcpy(&outptr[output_len], " ");
338 strcpy(outbuf, &outbuf[screenwidth-2]);
342 } while (done_reading == 0);
344 strcpy(&outptr[output_len], outbuf);
345 output_len += strlen(outbuf);
347 /* Strip leading/trailing whitespace. We can't do this with
348 * striplt() because it uses too many strlen()'s
350 while ((output_len > 0) && (isspace(outptr[0]))) {
351 strcpy(outptr, &outptr[1]);
354 while ((output_len > 0) && (isspace(outptr[output_len-1]))) {
355 outptr[output_len-1] = 0;
359 if (outptr[output_len-1] != '\n') {
360 strcat(outptr, "\n");