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 else if (!strcasecmp(tag, "H1")) {
136 strcat(outbuf, "\n\n");
139 else if (!strcasecmp(tag, "H2")) {
140 strcat(outbuf, "\n\n");
143 else if (!strcasecmp(tag, "H3")) {
144 strcat(outbuf, "\n\n");
147 else if (!strcasecmp(tag, "H4")) {
148 strcat(outbuf, "\n\n");
151 else if (!strcasecmp(tag, "/H1")) {
152 strcat(outbuf, "\n");
155 else if (!strcasecmp(tag, "/H2")) {
156 strcat(outbuf, "\n");
159 else if (!strcasecmp(tag, "/H3")) {
160 strcat(outbuf, "\n");
163 else if (!strcasecmp(tag, "/H4")) {
164 strcat(outbuf, "\n");
167 else if (!strcasecmp(tag, "HR")) {
168 strcat(outbuf, "\n ");
169 for (j=0; j<screenwidth-2; ++j)
171 strcat(outbuf, "\n");
174 else if (!strcasecmp(tag, "BR")) {
175 strcat(outbuf, "\n");
178 else if (!strcasecmp(tag, "TR")) {
179 strcat(outbuf, "\n");
182 else if (!strcasecmp(tag, "/TABLE")) {
183 strcat(outbuf, "\n");
186 else if (!strcasecmp(tag, "BLOCKQUOTE")) {
187 strcat(outbuf, "\n\n <<\n");
191 else if (!strcasecmp(tag, "/BLOCKQUOTE")) {
192 strcat(outbuf, "\n >>\n\n");
198 else if ((nest > 0) && (strlen(tag)<(sizeof(tag)-1))) {
199 tag[strlen(tag)+1] = 0;
200 tag[strlen(tag)] = ch;
204 outbuf[strlen(outbuf)+1] = 0;
205 outbuf[strlen(outbuf)] = ch;
208 strcpy(inbuf, &inbuf[i]);
211 /* Convert &; tags to the forbidden characters */
212 if (strlen(outbuf)>0) for (i=0; i<strlen(outbuf); ++i) {
214 if (!strncasecmp(&outbuf[i], " ", 6)) {
216 strcpy(&outbuf[i+1], &outbuf[i+6]);
219 else if (!strncasecmp(&outbuf[i], "<", 4)) {
221 strcpy(&outbuf[i+1], &outbuf[i+4]);
224 else if (!strncasecmp(&outbuf[i], ">", 4)) {
226 strcpy(&outbuf[i+1], &outbuf[i+4]);
229 else if (!strncasecmp(&outbuf[i], "&", 5)) {
230 strcpy(&outbuf[i+1], &outbuf[i+5]);
233 else if (!strncasecmp(&outbuf[i], """, 6)) {
235 strcpy(&outbuf[i+1], &outbuf[i+6]);
238 else if (!strncasecmp(&outbuf[i], "©", 6)) {
242 strcpy(&outbuf[i+3], &outbuf[i+6]);
245 else if (!strncasecmp(&outbuf[i], "®", 5)) {
249 strcpy(&outbuf[i+3], &outbuf[i+5]);
252 /* two-digit decimal equivalents */
253 else if ((!strncmp(&outbuf[i], "&#", 2))
254 && (outbuf[i+4] == ';') ) {
256 sscanf(&outbuf[i+2], "%02d", &scanch);
258 strcpy(&outbuf[i+1], &outbuf[i+5]);
261 /* three-digit decimal equivalents */
262 else if ((!strncmp(&outbuf[i], "&#", 2))
263 && (outbuf[i+5] == ';') ) {
265 sscanf(&outbuf[i+2], "%03d", &scanch);
267 strcpy(&outbuf[i+1], &outbuf[i+6]);
272 /* Make sure the output buffer is big enough */
273 if ((output_len + strlen(outbuf) + SIZ)
274 > outptr_buffer_size) {
275 outptr_buffer_size += SIZ;
276 outptr = realloc(outptr, outptr_buffer_size);
279 /* Output any lines terminated with hard line breaks */
282 if (strlen(outbuf)>0) {
283 for (i = 0; i<strlen(outbuf); ++i) {
284 if ( (i<(screenwidth-2)) && (outbuf[i]=='\n')) {
286 strncpy(&outptr[output_len],
291 strcpy(&outptr[output_len],
296 strcpy(outbuf, &outbuf[i+1]);
304 /* Add soft line breaks */
305 if (strlen(outbuf) > (screenwidth - 2 )) {
307 for (i=0; i<(screenwidth-2); ++i) {
308 if (outbuf[i]==32) rb = i;
311 strncpy(&outptr[output_len], outbuf, rb);
313 strcpy(&outptr[output_len], "\n");
316 strcpy(&outptr[output_len], " ");
319 strcpy(outbuf, &outbuf[rb+1]);
321 strncpy(&outptr[output_len], outbuf,
323 output_len += (screenwidth-2);
324 strcpy(&outptr[output_len], "\n");
327 strcpy(&outptr[output_len], " ");
330 strcpy(outbuf, &outbuf[screenwidth-2]);
334 } while (done_reading == 0);
336 strcpy(&outptr[output_len], outbuf);
337 output_len += strlen(outbuf);
339 /* Strip leading/trailing whitespace. We can't do this with
340 * striplt() because it uses too many strlen()'s
342 while ((output_len > 0) && (isspace(outptr[0]))) {
343 strcpy(outptr, &outptr[1]);
346 while ((output_len > 0) && (isspace(outptr[output_len-1]))) {
347 outptr[output_len-1] = 0;
351 if (outptr[output_len-1] != '\n') {
352 strcat(outptr, "\n");