[syslog-ng] [PATCH 0/3]: value-pairs() & related patches

[Balazs Scheidler]

On Sun, 2011-04-24 at 11:18 +0200, Gergely Nagy wrote:
> Hi!
> 
> For a multitude of reasons, ranging from the feature freeze looming over
> my head to hope that discrete patches are easier to review and merge
> than my awfully chaotic git tree, I'll be posting a couple of patches in
> a short while, as soon as I figure out what magic arguments I need to
> give to git send-email to make it send the batch the way I imagined.
> 
> value-pairs()
> =============
> 
> The first one will be the value-pairs() implementation (also available
> on my integration/value-pairs branch), which is the base of the two
> others that will follow.
> 
> While the value-pairs() implementation could still use some work,
> especially in the performance department, I believe it's good enough for
> initial use, and the benefits of having it (so that the mongodb driver
> can make use of it) far outweight the downsides.
> 
> The implementation itself is reasonably light, it hardly touches
> anything outside it's own two files: the grammar and the makefile only,
> to hook itself in. If not used by a driver, then it does not change
> functionality at all. Therefore, I believe this could go in before the
> feature freeze.

Ok, I've reviewed this patch and since it is really quite self-contained
it can go in.

I'm commenting on the patch in the separate mail.

> 
> afmongodb value-pairs() port
> ============================
> 
> The next patch will be the port of the already existing mongodb
> destination driver to use value-pairs(). Also available on my
> integration/afmongodb-vp branch.
> 
> This means that the driver will finally be able to do what it's good at:
> store arbitrary information in the database: whether that information
> comes via SDATA or patterndb, it doesn't matter, it will be stored in a
> structured manner.
> 
> In my opinion, this is the major selling point of the mongodb
> destination driver at the moment (I have some other grand plans with it
> in the future, but that's more invasive, less important, and can easily
> wait 'till 3.4 ;). Without value-pairs(), afmongodb is pretty much
> useless.

since mongodb is quite new, and probably not many people use it, this
can also go in.

> 
> tfson()
> =======
> 
> For extra fun & profit, Balint Kovacs' format_json has been ported to
> use value-pairs(), and this will be the third patch in the series, and
> is also available on my integration/tfjson branch.
> 
> This one's the least important, though it's dead simple, therefore I'd
> be very happy if it made past the feature freeze aswell.

I'd integrate this into 3.4 instead. mongodb is already in, and the
value-pairs stuff is a clear win there.

tfjson and substr (anything I've missed) should go to 3.4. Hopefully
that can be a faster release-cycle, as I don't intend to break
everything as I did with 3.3. (you know converting a program from single
to multi-threaded has a certain cost).

-- 
Bazsi