summaryrefslogtreecommitdiff
path: root/lib/libalpm/util.c
diff options
context:
space:
mode:
authorDan McGee <dan@archlinux.org>2010-12-14 22:26:23 -0600
committerDan McGee <dan@archlinux.org>2010-12-21 14:58:17 -0600
commitf2dff0860053f45c91e4ee301fda091a6b3d7361 (patch)
treea62d43b8041f172a67b1c0e0d2ca8336ea698fee /lib/libalpm/util.c
parent126f50ab0b5ee3ed46c5a6ecae241e8af49b0fe2 (diff)
Overhaul archive fgets function
The old function was written in a time before we relied on it for nearly every operation. Since then, we have switched to the archive backend and now fast parsing is a big deal. The former function made a per-character call to the libarchive archive_read_data() function, which resulted in some 21 million calls in a typical "load all sync dbs" operation. If we instead do some buffering of our own and read the blocks directly, and then find our newlines from there, we can cut out the multiple layers of overhead and go from archive to parsed data much quicker. Both users of the former function are switched over to the new signature, made easier by the macros now in place in the sync backend parsing code. Performance: for a `pacman -Su` (no upgrades available), _alpm_archive_fgets() goes from being 29% of the total time to 12% The time spent on the libarchive function being called dropped from 24% to 6%. This pushes _alpm_pkg_find back to the title of slowest low-level function. Signed-off-by: Dan McGee <dan@archlinux.org>
Diffstat (limited to 'lib/libalpm/util.c')
-rw-r--r--lib/libalpm/util.c100
1 files changed, 78 insertions, 22 deletions
diff --git a/lib/libalpm/util.c b/lib/libalpm/util.c
index 1291ea0f..d34eab5e 100644
--- a/lib/libalpm/util.c
+++ b/lib/libalpm/util.c
@@ -771,33 +771,89 @@ int _alpm_test_md5sum(const char *filepath, const char *md5sum)
return(ret);
}
-char *_alpm_archive_fgets(char *line, size_t size, struct archive *a)
+/* Note: does NOT handle sparse files on purpose for speed. */
+int _alpm_archive_fgets(struct archive *a, struct archive_read_buffer *b)
{
- /* for now, just read one char at a time until we get to a
- * '\n' char. we can optimize this later with an internal
- * buffer. */
- /* leave room for zero terminator */
- char *last = line + size - 1;
- char *i;
-
- for(i = line; i < last; i++) {
- int ret = archive_read_data(a, i, 1);
- /* special check for first read- if null, return null,
- * this indicates EOF */
- if(i == line && (ret <= 0 || *i == '\0')) {
- return(NULL);
+ char *i = NULL;
+ int64_t offset;
+ int done = 0;
+
+ while(1) {
+ /* have we processed this entire block? */
+ if(b->block + b->block_size == b->block_offset) {
+ if(b->ret == ARCHIVE_EOF) {
+ /* reached end of archive on the last read, now we are out of data */
+ goto cleanup;
+ }
+
+ /* zero-copy - this is the entire next block of data. */
+ b->ret = archive_read_data_block(a, (void*)&b->block,
+ &b->block_size, &offset);
+ b->block_offset = b->block;
+
+ /* error or end of archive with no data read, cleanup */
+ if(b->ret < ARCHIVE_OK ||
+ (b->block_size == 0 && b->ret == ARCHIVE_EOF)) {
+ goto cleanup;
+ }
}
- /* check if read value was null or newline */
- if(ret <= 0 || *i == '\0' || *i == '\n') {
- last = i + 1;
- break;
+
+ /* loop through the block looking for EOL characters */
+ for(i = b->block_offset; i < (b->block + b->block_size); i++) {
+ /* check if read value was null or newline */
+ if(*i == '\0' || *i == '\n') {
+ done = 1;
+ break;
+ }
}
- }
- /* always null terminate the buffer */
- *last = '\0';
+ /* allocate our buffer, or ensure our existing one is big enough */
+ if(!b->line) {
+ /* set the initial buffer to the read block_size */
+ CALLOC(b->line, b->block_size + 1, sizeof(char),
+ RET_ERR(PM_ERR_MEMORY, -1));
+ b->line_size = b->block_size + 1;
+ b->line_offset = b->line;
+ } else {
+ size_t needed = (b->line_offset - b->line) + (i - b->block_offset) + 1;
+ if(needed > b->max_line_size) {
+ RET_ERR(PM_ERR_MEMORY, -1);
+ }
+ if(needed > b->line_size) {
+ /* need to realloc + copy data to fit total length */
+ char *new;
+ CALLOC(new, needed, sizeof(char), RET_ERR(PM_ERR_MEMORY, -1));
+ memcpy(new, b->line, b->line_size);
+ b->line_size = needed;
+ b->line_offset = new + (b->line_offset - b->line);
+ free(b->line);
+ b->line = new;
+ }
+ }
+
+ if(done) {
+ size_t len = i - b->block_offset;
+ memcpy(b->line_offset, b->block_offset, len);
+ b->line_offset[len] = '\0';
+ b->block_offset = ++i;
+ /* this is the main return point; from here you can read b->line */
+ return(ARCHIVE_OK);
+ } else {
+ /* we've looked through the whole block but no newline, copy it */
+ size_t len = b->block + b->block_size - b->block_offset;
+ memcpy(b->line_offset, b->block_offset, len);
+ b->line_offset += len;
+ b->block_offset = i;
+ }
+ }
- return(line);
+cleanup:
+ {
+ int ret = b->ret;
+ FREE(b->line);
+ memset(b, 0, sizeof(b));
+ return(ret);
+ }
}
int _alpm_splitname(const char *target, pmpkg_t *pkg)