struct message_search_context *
message_search_init(const char *normalized_key_utf8,
normalizer_func_t *normalizer,
enum message_search_flags flags)
{
struct message_search_context *ctx;
i_assert(*normalized_key_utf8 != '\0');
ctx = i_new(struct message_search_context, 1);
ctx->flags = flags;
ctx->decoder = message_decoder_init(normalizer, 0);
ctx->str_find_ctx = str_find_init(default_pool, normalized_key_utf8);
return ctx;
}
static bool test_str_find_substring(const char *key, int expected_pos)
{
const unsigned char *text = (const unsigned char *)str_find_text;
const unsigned int text_len = strlen(str_find_text);
struct str_find_context *ctx;
unsigned int i, j, pos, max, offset;
bool ret;
ctx = str_find_init(pool_datastack_create(), key);
/* divide text into every possible block combination and test that
it matches */
i_assert(text_len > 0);
max = 1U << (text_len-1);
for (i = 0; i < max; i++) {
str_find_reset(ctx);
pos = 0; offset = 0; ret = FALSE;
for (j = 0; j < text_len; j++) {
if ((i & (1 << j)) != 0) {
if (str_find_more(ctx, text+pos, j-pos+1)) {
ret = TRUE;
break;
}
offset += j-pos + 1;
pos = j + 1;
}
}
if (pos != text_len && !ret) {
if (str_find_more(ctx, text+pos, j-pos))
ret = TRUE;
}
if (expected_pos < 0) {
if (ret)
return FALSE;
} else {
if (!ret)
return FALSE;
pos = str_find_get_match_end_pos(ctx) +
offset - strlen(key);
if ((int)pos != expected_pos)
return FALSE;
}
}
return TRUE;
}
static int
dbox_file_find_next_magic(struct dbox_file *file, uoff_t *offset_r, bool *pre_r)
{
struct istream *input = file->input;
struct str_find_context *pre_ctx, *post_ctx;
uoff_t orig_offset, pre_offset, post_offset;
const unsigned char *data;
size_t size;
int ret;
*pre_r = FALSE;
pre_ctx = str_find_init(default_pool, "\n"DBOX_MAGIC_PRE);
post_ctx = str_find_init(default_pool, DBOX_MAGIC_POST);
/* \n isn't part of the DBOX_MAGIC_PRE, but it always preceds it.
assume that at this point we've already just read the \n. when
scanning for it later we'll need to find the \n though. */
str_find_more(pre_ctx, (const unsigned char *)"\n", 1);
orig_offset = input->v_offset;
while ((ret = i_stream_read_data(input, &data, &size, 0)) > 0) {
pre_offset = (uoff_t)-1;
if (str_find_more(pre_ctx, data, size)) {
pre_offset = input->v_offset +
str_find_get_match_end_pos(pre_ctx) -
(strlen(DBOX_MAGIC_PRE) + 1);
*pre_r = TRUE;
}
if (str_find_more(post_ctx, data, size)) {
post_offset = input->v_offset +
str_find_get_match_end_pos(post_ctx) -
strlen(DBOX_MAGIC_POST);
if (pre_offset == (uoff_t)-1 ||
post_offset < pre_offset) {
pre_offset = post_offset;
*pre_r = FALSE;
}
}
if (pre_offset != (uoff_t)-1) {
if (*pre_r) {
/* LF isn't part of the magic */
pre_offset++;
}
*offset_r = pre_offset;
break;
}
i_stream_skip(input, size);
}
if (ret <= 0) {
i_assert(ret == -1);
if (input->stream_errno != 0)
dbox_file_set_syscall_error(file, "read()");
else {
ret = 0;
*offset_r = input->v_offset;
}
}
i_stream_seek(input, orig_offset);
str_find_deinit(&pre_ctx);
str_find_deinit(&post_ctx);
return ret;
}
