On 15 Feb 2018, at 11:10pm, Jens Alfke <j...@mooseyard.com> wrote:
>> On Feb 15, 2018, at 10:09 AM, Simon Slavin <slav...@bigfraud.org> wrote:
>>
>>> 1) I can't be the only programmer who learned to make paired calls ("If you
>>> initialise something, it needs deinitializing, if you allocate something,
>>> deallocate it.").
>
> When a process exits, its resources are automatically cleaned up by the
> kernel — memory and address space freed, file handles and sockets closed,
> etc. So it’s generally unnecessary to do redundant cleanup like
> sqlite3_shutdown* on quit, and it can actually slow things down**, especially
> if freeing a zillion heap blocks that have been paged out or are no longer in
> CPU caches.
>
> —Jens
>
> * This doesn’t apply to sqlite3_close() on a writeable database, since SQLite
> may need to update data in the file before closing it.
Your footnote is an example of the reason I do paired calls. I didn't write
the SQLite API. I don't know whether sqlite3_shutdown() does anything. For
all I know, sqlite3_shutdown() is the thing that deletes dead journal and
shared memory files. Maybe it slows things down too much to do that on
sqlite3_close(). I don't know.
Given the way the documentation is arranged, and the lack of any mention to the
contrary, I will assume sqlite3_shutdown() is paired with sqlite_initialize().
One of the Developer team tells me it'd not needed except for cases where your
process isn't about to quit ? Fine. But that should really be in the
documentation so people who don't read this list know it. Because the name
sqlite3_shutdown(), and the documentation for it, implies you should call it
when sqlite3_initialize() has been called and you don't need SQLite any more.
Simon.
_______________________________________________
sqlite-users mailing list
sqlite-users@mailinglists.sqlite.org
http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
