So if I'm reading the documentation correctly:

====
The wal-index is in shared memory, and so technically it does not have
to have a name in the host computer filesystem. Custom VFS
implementations are free to implement shared memory in any way they
see fit, but the default unix and windows drivers that come built-in
with SQLite implement shared memory using mmapped files named using
the suffix "-shm" and located in the same directory as the database
file.

...

Early (pre-release) implementations of WAL mode stored the wal-index
in volatile shared-memory, such as files created in /dev/shm on Linux
or /tmp on other unix systems. The problem with that approach is that
processes with a different root directory (changed via chroot) will
see different files and hence use different shared memory areas,
leading to database corruption.
====

It seems like the only thing preventing WAL from working with
read-only databases is this lack of a global namespace for shared
memory.

This exists in many Linux systems as "/dev/shm", or even "/tmp" would
work fine for a lot of users.  I totally understand that you can't
make this the default, because it could potentially lead to strange
behavior with chroot()s and the like.  But for those of us with
controlled environments who know that all applications using SQLite
share the same view of the filesystem, it would be great if we could
#define an option which turns this on.  For my application, chroot()ed
apps are a complete non-issue, whereas lack of read-only DB access is
a dealbreaker for WAL (which would really be a shame, the performance
benefit is substantial!)

I know I could always write my own VFS to do this, but that seems like
overkill.  :)  More importantly, it requires maintenance - I'd
probably create my VFS by copying os_unix.c, but then I wouldn't
automatically be getting any fixes/updates that you guys make to that
file going forward.

Best would be if SQLite had a #define like SQLITE_CUSTOM_WAL_LOCATION,
which defaults to undefined (and hence you use the same directory as
the DB file), but which users could define to "/dev/shm", "/tmp", or
some other location to place all the shm files there, globally.  The
changes in

http://www.sqlite.org/src/fdiff?v1=ae173c9f6afaa58b2833a1c95c6cd32021755c42&v2=a76d1952ac7984574701c48b665220b871c5c9a5

are pretty straightforward, so I could probably take a stab at this if you want.

What do you guys think?

-- 
Matthew L. Creech
_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users

Reply via email to