I've been looking at the mailbox listing code (starting from ic_list and
working into db.c) to improve efficiency, and the fix is pretty clear:
ic_list calls db_findmailbox_by_regex, which calls
db_list_mailboxes_by_regex, which does a big SELECT, compares the mailbox
names with the regex, keeps the matching mailboxidnr's, and throws away
the names.
ic_list then loops through the list of mailboxidnr's, and selects three
times for each mailbox: first for the mailbox name, then for is_selectable
then for no_inferiors.
The obvious fix is to keep all of that information from the
dbmail_mailboxes table that was retrieved with list_mailboxes in the first
place. So I'm building a structure to hold that information and eliminate
the loop of three more queries per mailbox.
But, in the process, I'd like to know if there's rhyme or reason for:
int db_findmailbox_by_regex(u64_t owner_idnr, const char *pattern,
u64_t ** children, unsigned *nchildren,
int only_subscribed);
static int db_list_mailboxes_by_regex(u64_t user_idnr, int
only_subscribed,
regex_t * preg,
u64_t ** mailboxes,
unsigned int *nr_mailboxes);
Yes, they really are the same functions, with different order, and just
with findmailbox_by_regex compiling the regex before calling
list_mailboxes.
I'd like to roll these into one function.
Aaron
