Hi,
I believe I've addressed the corner case in this patch set by returning a
dict() that represents the current state of the database on connection
reset/initial sync, otherwise a list of changes. This should allow the user
to have a consistent view of the database. I've also changed the response
I am also a bit concerned by issues that might arise
from a user thinking that this is always accurate, rather
than hints. Aaron, I think you had said something
regarding this when we chatted off-list, but I don't
recall the details.
-Reid
On Fri, Aug 9, 2013 at 2:50 PM, Reid Price wrote:
>
This is pretty crude (nothing else in the IDL interface requires the
caller to understand the JSON-RPC protocol) and seems difficult for
the callers to use. The changes that it returns can also be
incomplete, at least in one corner case: if the database connection
drops and reconnects, then the re
On Fri, Aug 9, 2013 at 2:50 PM, Reid Price wrote:
> Or you could keep the original function behavior the same and expose this
> as a separate function
>
> def foo(...):
>
>
> def run(...):
> return self.foo(...)[0]
>
> where foo is a better function name - update? run_details?
> run
Or you could keep the original function behavior the same and expose this
as a separate function
def foo(...):
def run(...):
return self.foo(...)[0]
where foo is a better function name - update? run_details?
run_with_changes? run_diff? _run? No opinion there.
-Reid
On Fri, Au
Right, this would break things for anyone checking the return value of
idl.run(). The only alternative I see to that is if we pass an optional arg
to run() (i.e: def run(self, return_changes=False)). Would you prefer this
instead?
Thanks,
Aaron
On Fri, Aug 9, 2013 at 1:02 PM, Ben Pfaff wrote:
On Tue, Aug 06, 2013 at 02:45:35PM -0700, Aaron Rosen wrote:
> This patch changes what is being returned from Idl.run() to a tuple
> (changed, changes) so one can determine what changes have occurred to
> the database without having to read the entire table.
>
> Signed-off-by: Aaron Rosen
It see
NVPd also calls idl.run()
:: psi
On Aug 7, 2013, at 1:37 PM, Reid Price wrote:
> I like this one better. I am surprised that test-ovsdb has the only
> invocations of idl.run(), but if that is true, LGTM.
>
>
> On Tue, Aug 6, 2013 at 3:14 PM, Aaron Rosen wrote:
> This patch changes what is
I like this one better. I am surprised that test-ovsdb has the only
invocations of idl.run(), but if that is true, LGTM.
On Tue, Aug 6, 2013 at 3:14 PM, Aaron Rosen wrote:
> This patch changes what is being returned from Idl.run() to a tuple
> (changed, changes) so one can determine what chang
This patch changes what is being returned from Idl.run() to a tuple
(changed, changes) so one can determine what changes have occurred to
the database without having to read the entire table.
Signed-off-by: Aaron Rosen
---
python/ovs/db/idl.py | 16 ++--
tests/test-ovsdb.py |4
This patch changes what is being returned from Idl.run() to a tuple
(changed, changes) so one can determine what changes have occurred to
the database without having to read the entire table.
Signed-off-by: Aaron Rosen
---
python/ovs/db/idl.py | 16 ++--
tests/test-ovsdb.py |4
11 matches
Mail list logo
