int
cha_parse_bos(cha_lat_t *lat)
{
static int path0 = -1;
lat->offset = lat->cursor = 0;
lat->anno = lat->last_anno = -1;
lat->head_path = 1;
cha_block_clear(Cha_mrph_block);
free_chars();
free_path();
Cha_path[0].start = Cha_path[0].end = 0;
Cha_path[0].path = &path0;
Cha_path[0].cost = 0;
Cha_path[0].mrph_p = 0;
Cha_path[0].state = 0;
Cha_path_num = 1;
register_bos_eos();
return 0;
}
/* Produces a string consisting of the characters available now in all decdoed
* buffers. */
static MVMString * get_all_in_buffer(MVMThreadContext *tc, MVMDecodeStream *ds) {
MVMString *result = (MVMString *)MVM_repr_alloc_init(tc, tc->instance->VMString);
result->body.storage_type = MVM_STRING_GRAPHEME_32;
/* If there's no codepoint buffer, then return the empty string. */
if (!ds->chars_head) {
result->body.storage.blob_32 = NULL;
result->body.num_graphs = 0;
}
/* If there's exactly one resulting codepoint buffer and we swallowed none
* of it, just use it. */
else if (ds->chars_head == ds->chars_tail && ds->chars_head_pos == 0) {
/* Set up result string. */
result->body.storage.blob_32 = ds->chars_head->chars;
result->body.num_graphs = ds->chars_head->length;
/* Don't free the buffer's memory itself, just the holder, as we
* stole that for the buffer into the string above. */
free_chars(tc, ds, ds->chars_head);
ds->chars_head = ds->chars_tail = NULL;
}
/* Otherwise, need to assemble all the things. */
else {
/* Calculate length. */
MVMint32 length = 0, pos = 0;
MVMDecodeStreamChars *cur_chars = ds->chars_head;
while (cur_chars) {
if (cur_chars == ds->chars_head)
length += cur_chars->length - ds->chars_head_pos;
else
length += cur_chars->length;
cur_chars = cur_chars->next;
}
/* Allocate a result buffer of the right size. */
result->body.storage.blob_32 = MVM_malloc(length * sizeof(MVMGrapheme32));
result->body.num_graphs = length;
/* Copy all the things into the target, freeing as we go. */
cur_chars = ds->chars_head;
while (cur_chars) {
MVMDecodeStreamChars *next_chars = cur_chars->next;
if (cur_chars == ds->chars_head) {
MVMint32 to_copy = ds->chars_head->length - ds->chars_head_pos;
memcpy(result->body.storage.blob_32 + pos, cur_chars->chars + ds->chars_head_pos,
to_copy * sizeof(MVMGrapheme32));
pos += to_copy;
}
else {
memcpy(result->body.storage.blob_32 + pos, cur_chars->chars,
cur_chars->length * sizeof(MVMGrapheme32));
pos += cur_chars->length;
}
MVM_free(cur_chars->chars);
free_chars(tc, ds, cur_chars);
cur_chars = next_chars;
}
ds->chars_head = ds->chars_tail = NULL;
}
return result;
}
static MVMString * take_chars(MVMThreadContext *tc, MVMDecodeStream *ds, MVMint32 chars, MVMint32 exclude) {
MVMString *result;
MVMint32 found = 0;
MVMint32 result_found = 0;
MVMint32 result_chars = chars - exclude;
if (result_chars < 0)
MVM_exception_throw_adhoc(tc, "DecodeStream take_chars: chars - exclude < 0 should never happen");
result = (MVMString *)MVM_repr_alloc_init(tc, tc->instance->VMString);
result->body.storage_type = MVM_STRING_GRAPHEME_32;
result->body.num_graphs = result_chars;
/* In the best case, the head char buffer has exactly what we need. This
* will typically happen when it a steady state of decoding lines. */
if (ds->chars_head->length == chars && ds->chars_head_pos == 0) {
MVMDecodeStreamChars *cur_chars = ds->chars_head;
result->body.storage.blob_32 = cur_chars->chars;
ds->chars_head = cur_chars->next;
if (ds->chars_head == NULL)
ds->chars_tail = NULL;
free_chars(tc, ds, cur_chars);
}
/* Otherwise, need to take and copy. */
else {
result->body.storage.blob_32 = MVM_malloc(result_chars * sizeof(MVMGrapheme32));
while (found < chars) {
MVMDecodeStreamChars *cur_chars = ds->chars_head;
MVMint32 available = cur_chars->length - ds->chars_head_pos;
if (available <= chars - found) {
/* We need all that's left in this buffer and likely
* more. */
MVMDecodeStreamChars *next_chars = cur_chars->next;
if (available <= result_chars - result_found) {
memcpy(result->body.storage.blob_32 + result_found,
cur_chars->chars + ds->chars_head_pos,
available * sizeof(MVMGrapheme32));
result_found += available;
}
else {
MVMint32 to_copy = result_chars - result_found;
memcpy(result->body.storage.blob_32 + result_found,
cur_chars->chars + ds->chars_head_pos,
to_copy * sizeof(MVMGrapheme32));
result_found += to_copy;
}
found += available;
MVM_free(cur_chars->chars);
free_chars(tc, ds, cur_chars);
ds->chars_head = next_chars;
ds->chars_head_pos = 0;
if (ds->chars_head == NULL)
ds->chars_tail = NULL;
}
else {
/* There's enough in this buffer to satisfy us, and we'll leave
* some behind. */
MVMint32 take = chars - found;
MVMint32 to_copy = result_chars - result_found;
memcpy(result->body.storage.blob_32 + result_found,
cur_chars->chars + ds->chars_head_pos,
to_copy * sizeof(MVMGrapheme32));
result_found += to_copy;
found += take;
ds->chars_head_pos += take;
}
}
}
return result;
}
int insert_stream(buffer * const b, line_desc * ld, int64_t line, int64_t pos, const char * const stream, const int64_t stream_len) {
if (!b || !ld || !stream || stream_len < 1 || pos > ld->line_len) return ERROR;
assert_line_desc(ld, b->encoding);
assert_buffer(b);
block_signals();
if (b->opt.do_undo && !(b->undoing || b->redoing)) {
const int error = add_undo_step(b, line, pos, -stream_len);
if (error) {
release_signals();
return error;
}
}
const char *s = stream;
while(s - stream < stream_len) {
int64_t const len = strnlen_ne(s, stream_len - (s - stream));
if (len) {
/* First case; there is no character allocated on this line. We
have to freshly allocate the line. */
if (!ld->line) {
if (ld->line = alloc_chars(b, len)) {
memcpy(ld->line, s, len);
ld->line_len = len;
}
else {
release_signals();
return OUT_OF_MEMORY;
}
}
/* Second case. There are not enough characters around ld->line. Note
that the value of the check_first_before parameter depends on
the position at which the insertion will be done, and it is chosen
in such a way to minimize the number of characters to move. */
else {
const int64_t result = alloc_chars_around(b, ld, len, pos < ld->line_len / 2);
if (result < 0) {
char * const p = alloc_chars(b, ld->line_len + len);
if (p) {
memcpy(p, ld->line, pos);
memcpy(&p[pos], s, len);
memcpy(&p[pos + len], ld->line + pos, ld->line_len - pos);
free_chars(b, ld->line, ld->line_len);
ld->line = p;
ld->line_len += len;
}
else {
release_signals();
return OUT_OF_MEMORY;
}
}
else { /* Third case. There are enough free characters around ld->line. */
if (len - result) memmove(ld->line - (len - result), ld->line, pos);
if (result) memmove(ld->line + pos + result, ld->line + pos, ld->line_len - pos);
memcpy(ld->line - (len - result) + pos, s, len);
ld->line -= (len - result);
ld->line_len += len;
}
}
b->is_modified = 1;
/* We just inserted len chars at (line,pos); adjust bookmarks and mark accordingly. */
if (b->marking && b->block_start_line == line && b->block_start_pos > pos) b->block_start_pos += len;
for (int i = 0, mask = b->bookmark_mask; mask; i++, mask >>= 1)
if ((mask & 1) && b->bookmark[i].line == line && b->bookmark[i].pos > pos) b->bookmark[i].pos += len;
}
/* If the string we have inserted has a NULL at the end, we create a new
line under the current one and set ld to point to it. */
if (len + (s - stream) < stream_len) {
line_desc *new_ld;
if (new_ld = alloc_line_desc(b)) {
add(&new_ld->ld_node, &ld->ld_node);
b->num_lines++;
if (pos + len < ld->line_len) {
new_ld->line_len = ld->line_len - pos - len;
new_ld->line = &ld->line[pos + len];
ld->line_len = pos + len;
if (pos + len == 0) ld->line = NULL;
}
b->is_modified = 1;
ld = new_ld;
/* We just inserted a line break at (line,pos);
adjust the buffer bookmarks and mark accordingly. */
if (b->marking) {
if (b->block_start_line == line && b->block_start_pos > pos) {
b->block_start_pos -= pos;
b->block_start_line++;
}
else if (b->block_start_line > line) b->block_start_line++;
}
for (int i = 0, mask=b->bookmark_mask; mask; i++, mask >>= 1) {
if (mask & 1) {
if (b->bookmark[i].line == line && b->bookmark[i].pos > pos) {
b->bookmark[i].pos -= pos;
b->bookmark[i].line++;
}
else if (b->bookmark[i].line > line) b->bookmark[i].line++;
}
}
pos = 0;
line++;
}
else {
