1008: Added upload_progress to the connection.
[elinks.git] / src / encoding / encoding.c
blob307bafd18ea605c3ffd05a39a4438834de4aef3a
1 /* Stream reading and decoding (mostly decompression) */
3 #ifdef HAVE_CONFIG_H
4 #include "config.h"
5 #endif
7 #include <errno.h>
8 #include <stdio.h>
9 #include <string.h>
10 #include <sys/stat.h> /* OS/2 needs this after sys/types.h */
11 #include <sys/types.h>
12 #ifdef HAVE_FCNTL_H
13 #include <fcntl.h> /* OS/2 needs this after sys/types.h */
14 #endif
15 #ifdef HAVE_UNISTD_H
16 #include <unistd.h>
17 #endif
19 #include "elinks.h"
21 #include "config/options.h"
22 #include "encoding/encoding.h"
23 #include "network/state.h"
24 #include "osdep/osdep.h"
25 #include "util/memory.h"
26 #include "util/string.h"
29 /*************************************************************************
30 Dummy encoding (ENCODING_NONE)
31 *************************************************************************/
33 struct dummy_enc_data {
34 int fd;
37 static int
38 dummy_open(struct stream_encoded *stream, int fd)
40 stream->data = mem_alloc(sizeof(struct dummy_enc_data));
41 if (!stream->data) return -1;
43 ((struct dummy_enc_data *) stream->data)->fd = fd;
45 return 0;
48 static int
49 dummy_read(struct stream_encoded *stream, unsigned char *data, int len)
51 return safe_read(((struct dummy_enc_data *) stream->data)->fd, data, len);
54 static unsigned char *
55 dummy_decode_buffer(unsigned char *data, int len, int *new_len)
57 unsigned char *buffer = memacpy(data, len);
59 if (!buffer) return NULL;
61 *new_len = len;
62 return buffer;
65 static void
66 dummy_close(struct stream_encoded *stream)
68 close(((struct dummy_enc_data *) stream->data)->fd);
69 mem_free(stream->data);
72 static const unsigned char *const dummy_extensions[] = { NULL };
74 static const struct decoding_backend dummy_decoding_backend = {
75 "none",
76 dummy_extensions,
77 dummy_open,
78 dummy_read,
79 dummy_decode_buffer,
80 dummy_close,
84 /* Dynamic backend area */
86 #include "encoding/bzip2.h"
87 #include "encoding/deflate.h"
88 #include "encoding/lzma.h"
90 static const struct decoding_backend *const decoding_backends[] = {
91 &dummy_decoding_backend,
92 &gzip_decoding_backend,
93 &bzip2_decoding_backend,
94 &lzma_decoding_backend,
95 &deflate_decoding_backend,
99 /*************************************************************************
100 Public functions
101 *************************************************************************/
104 /* Associates encoded stream with a fd. */
105 struct stream_encoded *
106 open_encoded(int fd, enum stream_encoding encoding)
108 struct stream_encoded *stream;
110 stream = mem_alloc(sizeof(*stream));
111 if (!stream) return NULL;
113 stream->encoding = encoding;
114 if (decoding_backends[stream->encoding]->open(stream, fd) >= 0)
115 return stream;
117 mem_free(stream);
118 return NULL;
121 /* Read available data from stream and decode them. Note that when data change
122 * their size during decoding, 'len' indicates desired size of _returned_ data,
123 * not desired size of data read from stream. */
125 read_encoded(struct stream_encoded *stream, unsigned char *data, int len)
127 return decoding_backends[stream->encoding]->read(stream, data, len);
130 /* Decode an entire file from a buffer. This function is not suitable
131 * for parts of files. @data contains the original data, @len bytes
132 * long. The resulting decoded data chunk is *@new_len bytes long. */
133 unsigned char *
134 decode_encoded_buffer(enum stream_encoding encoding, unsigned char *data, int len,
135 int *new_len)
137 return decoding_backends[encoding]->decode_buffer(data, len, new_len);
140 /* Closes encoded stream. Note that fd associated with the stream will be
141 * closed here. */
142 void
143 close_encoded(struct stream_encoded *stream)
145 decoding_backends[stream->encoding]->close(stream);
146 mem_free(stream);
150 /* Return a list of extensions associated with that encoding. */
151 const unsigned char *const *listext_encoded(enum stream_encoding encoding)
153 return decoding_backends[encoding]->extensions;
156 enum stream_encoding
157 guess_encoding(unsigned char *filename)
159 int fname_len = strlen(filename);
160 unsigned char *fname_end = filename + fname_len;
161 int enc;
163 for (enc = 1; enc < ENCODINGS_KNOWN; enc++) {
164 const unsigned char *const *ext = decoding_backends[enc]->extensions;
166 while (ext && *ext) {
167 int len = strlen(*ext);
169 if (fname_len >= len && !strcmp(fname_end - len, *ext))
170 return enc;
172 ext++;
176 return ENCODING_NONE;
179 const unsigned char *
180 get_encoding_name(enum stream_encoding encoding)
182 return decoding_backends[encoding]->name;
186 /* File reading */
188 /* Tries to open @prefixname with each of the supported encoding extensions
189 * appended. */
190 static inline enum stream_encoding
191 try_encoding_extensions(struct string *filename, int *fd)
193 int length = filename->length;
194 int encoding;
196 /* No file of that name was found, try some others names. */
197 for (encoding = 1; encoding < ENCODINGS_KNOWN; encoding++) {
198 const unsigned char *const *ext = listext_encoded(encoding);
200 for (; ext && *ext; ext++) {
201 add_to_string(filename, *ext);
203 /* We try with some extensions. */
204 *fd = open(filename->source, O_RDONLY | O_NOCTTY);
206 if (*fd >= 0)
207 /* Ok, found one, use it. */
208 return encoding;
210 filename->source[length] = 0;
211 filename->length = length;
215 return ENCODING_NONE;
218 /* Reads the file from @stream in chunks of size @readsize. */
219 /* Returns a connection state. S_OK if all is well. */
220 enum connection_state
221 read_file(struct stream_encoded *stream, int readsize, struct string *page)
223 if (!init_string(page)) return S_OUT_OF_MEM;
225 /* We read with granularity of stt.st_size (given as @readsize) - this
226 * does best job for uncompressed files, and doesn't hurt for
227 * compressed ones anyway - very large files usually tend to inflate
228 * fast anyway. At least I hope ;). --pasky */
229 /* Also there because of bug in Linux. Read returns -EACCES when
230 * reading 0 bytes to invalid address so ensure never to try and
231 * allocate zero number of bytes. */
232 if (!readsize) readsize = 4096;
234 while (realloc_string(page, page->length + readsize)) {
235 unsigned char *string_pos = page->source + page->length;
236 int readlen = read_encoded(stream, string_pos, readsize);
238 if (readlen < 0) {
239 done_string(page);
241 /* If it is some I/O error (and errno is set) that will
242 * do. Since errno == 0 == S_WAIT and we cannot have
243 * that. */
244 if (errno)
245 return (enum connection_state) -errno;
247 /* FIXME: This is indeed an internal error. If readed from a
248 * corrupted encoded file nothing or only some of the
249 * data will be read. */
250 return S_ENCODE_ERROR;
252 } else if (readlen == 0) {
253 /* NUL-terminate just in case */
254 page->source[page->length] = '\0';
255 return S_OK;
258 page->length += readlen;
259 #if 0
260 /* This didn't work so well as it should (I had to implement
261 * end of stream handling to bzip2 anyway), so I rather
262 * disabled this. */
263 if (readlen < readsize) {
264 /* This is much safer. It should always mean that we
265 * already read everything possible, and it permits us
266 * more elegant of handling end of file with bzip2. */
267 break;
269 #endif
272 done_string(page);
273 return S_OUT_OF_MEM;
276 static inline int
277 is_stdin_pipe(struct stat *stt, struct string *filename)
279 /* On Mac OS X, /dev/stdin has type S_IFSOCK. (bug 616) */
280 return !strlcmp(filename->source, filename->length, "/dev/stdin", 10)
281 && (
282 #ifdef S_ISSOCK
283 S_ISSOCK(stt->st_mode) ||
284 #endif
285 S_ISFIFO(stt->st_mode));
288 enum connection_state
289 read_encoded_file(struct string *filename, struct string *page)
291 struct stream_encoded *stream;
292 struct stat stt;
293 enum stream_encoding encoding = ENCODING_NONE;
294 int fd = open(filename->source, O_RDONLY | O_NOCTTY);
295 enum connection_state state = -errno;
297 if (fd == -1 && get_opt_bool("protocol.file.try_encoding_extensions", NULL)) {
298 encoding = try_encoding_extensions(filename, &fd);
300 } else if (fd != -1) {
301 encoding = guess_encoding(filename->source);
304 if (fd == -1)
305 return state;
307 /* Some file was opened so let's get down to bi'ness */
308 set_bin(fd);
310 /* Do all the necessary checks before trying to read the file.
311 * @state code is used to block further progress. */
312 if (fstat(fd, &stt)) {
313 state = -errno;
315 } else if (!S_ISREG(stt.st_mode) && encoding != ENCODING_NONE) {
316 /* We only want to open regular encoded files. */
317 /* Leave @state being the saved errno */
319 } else if (!S_ISREG(stt.st_mode) && !is_stdin_pipe(&stt, filename)
320 && !get_opt_bool("protocol.file.allow_special_files", NULL)) {
321 state = S_FILE_TYPE;
323 } else if (!(stream = open_encoded(fd, encoding))) {
324 state = S_OUT_OF_MEM;
326 } else {
327 int readsize = (int) stt.st_size;
329 /* Check if st_size will cause overflow. */
330 /* FIXME: See bug 497 for info about support for big files. */
331 if (readsize != stt.st_size || readsize < 0) {
332 #ifdef EFBIG
333 state = (enum connection_state) -(EFBIG);
334 #else
335 state = S_FILE_ERROR;
336 #endif
338 } else {
339 state = read_file(stream, stt.st_size, page);
341 close_encoded(stream);
344 close(fd);
345 return state;