Re: paisley ?
19 views
Skip to first unread message
Sean McQuillan
Mar 22, 2011, 6:00:11 AM3/22/11
to Thomas Vander Stichele, paisley-d...@googlegroups.com
[[added mailing list]]
--
A few weird couchisms I discovered writing our internal function load_from_view (that does basically the same thing as views.py except in our model layer).
1. include_docs = True is a lie, if you run queries with include_docs=True under any write load you *will* consistently get invalid results back, including deleted docs.
-> in this situation, often, here will not be a 'doc' field on the result row and no other symptom of failure.
2. sometimes, in situations I don't yet understand completely due to lack of thorough debugging, rows contain an 'error' field, when this happens they don't have a 'doc' even if the kwarg include_docs=True was passed to couch.
Cheers,
Sean
On Tue, Mar 22, 2011 at 2:33 AM, Thomas Vander Stichele <tho...@apestaart.org> wrote:
On Mon, 2011-03-21 at 15:42 -0700, Sean McQuillan wrote:Ok, that explains.
> The github branch is from 0.1 as far as I can remember. I really
> didn't intend for this repo to be anything except oss compliance when
> I made it, so I'd be unsurprised if I didn't check in paisley_example
> and paisley_bench.
>
>
> Looks
> like http://bazaar.launchpad.net/~paisley-dev/paisley/trunk/revision/4
> matches
>
>
> https://github.com/smcq/paisley/commit/bc47382c3dde55f84835d680970bb42235b42de7
I was continuing from the object-views branch:
https://code.launchpad.net/~dreid/paisley/object-views
since that was where David said he was working on.
Can you give a quick look at that branch to see if revisions 4 to the
end are ok for you guys ? I'll work them into a git branch.
Thomas
--
>
>
> Cheers,
> Sean
>
> On Mon, Mar 21, 2011 at 2:41 PM, Jason J. W. Williams
> <jasonjw...@gmail.com> wrote:
> You'd have to ask Sean. That was awhile before I got involved.
>
> -J
>
> Sent via iPhone
>
> Is your e-mail Premiere?
>
> On Mar 21, 2011, at 14:40, Thomas Vander Stichele
> <tho...@apestaart.org> wrote:
>
> > On Mon, 2011-03-21 at 12:36 -0600, Jason J. W. Williams
> wrote:
> >> I'd second the motion to continue using Github. There's a
> bigger
> >> community of folks so I think the activity will be higher
> (somewhat
> >> evidenced by the flurry of pull requests lately).
> >>
> >>
> >> Google groups works for me.
> >
>
> > At what point was paisley branched ? I don't see views.py in
> the paisley
> > module, and I don't see paisley_example.py and
> paisley_bench.py
>
> >
> > T
> >
> >
> >>
> >>
> >> -J
> >>
> >> On Mon, Mar 21, 2011 at 12:29 PM, Sean McQuillan
> >> <mcquill...@gmail.com> wrote:
> >> Hey,
> >>
> >>
> >> Great to have more people working on this! David
> had asked me
> >> to be maintainer as well, and I had moved the
> project to
> >> github as is my personal preference. Sorry about
> any
> >> confusion!
> >>
> >>
> >> If it's alright with you, I'd like to see
> development stay on
> >> github and I'll make you a committer over here.
> What's your
> >> account?
> >>
> >>
> >> Not really a lot of formal process, I've just been
> using
> >> github pull requests, bug tracker, and inline code
> reviews to
> >> communicate. A mailing list sounds like a good idea
> to
> >> announce releases / developer communication. I'd
> prefer to
> >> avoid launchpad, is there any other options that are
> decent
> >> (google groups?).
> >>
> >>
> >> Cheers,
> >> Sean
> >>
> >>
> >> On Mon, Mar 21, 2011 at 10:39 AM, Jason J. W.
> Williams
> >> <jasonjw...@gmail.com> wrote:
> >> Hi Thomas,
> >>
> >>
> >> Actually, I'm not the maintainer just a
> contributor.
> >> My understanding is Sean McQuillan is the
> maintainer.
> >> I wanted to add authentication to paisley
> back in
> >> November and when I went to the Paisley LP
> page it
> >> redirected to the Github repo, hence where I
> >> contributed. Sean made me a committer over
> there after
> >> I fixed a number of issues in Paisley.
> >>
> >>
> >> Right now to my understanding the
> >> github.com/smcq/paisley repo is the
> authoritative one.
> >> We've been using the GH Issue page
> associated with the
> >> repo for tracking bugs. We've had a number
> of pull
> >> requests off the GH repo since we announced
> the
> >> release. I know Sean is working through one
> of them
> >> right now and I've integrated another.
> >>
> >>
> >> I got ahold of Thomas Herve and got him to
> add us to
> >> the pypi paisley account so we can finally
> push new
> >> releases (hence 0.3 being on pypi).
> >>
> >>
> >> We'd love to have you be a part of the GH
> repo. If you
> >> want to fork the project and open up an
> issue with
> >> your changes, I don't think we'll have any
> issues
> >> putting your code in (providing the usual
> caveats
> >> etc. :) ). I'll talk to Sean about making
> you a
> >> committer.
> >>
> >>
> >> Would that work for you?
> >>
> >>
> >> -J
> >>
> >> On Mon, Mar 21, 2011 at 11:00 AM, Thomas
> Vander
> >> Stichele <tho...@apestaart.org> wrote:
> >> Hi guys,
> >>
> >> imagine my surprise seeing a paisley
> 0.3
> >> release announced on the
> >> twisted mailing list. Just as I was
> in the
> >> process of preparing my own
> >> 0.3 release ! :)
> >>
> >> There's some confusion here and I
> like to get
> >> it cleared up.
> >>
> >> David invoked a last touched rule
> iirc to make
> >> me maintainer of paisley
> >> a while ago. I didn't know someone
> else was
> >> either similarly touched or
> >> took initiative to be (but I'm all
> for more
> >> work to be done on it! Good
> >> job Jason!)
> >>
> >> Can we clear things up to figure out
> how to
> >> proceed ?
> >>
> >> a) there are branches on launchpad.
> I went
> >> there today and someone
> >> actually changed the lp page to
> point to
> >> Jason's github. I wish I had
> >> seen that before. What do we do
> with the
> >> branches on launchpad ?
> >>
> >> b) there's a paisley-users list on
> launchpad.
> >> It hasn't had many mails
> >> (3 ?) Is there another mailing list
> we want to
> >> use and promote instead ?
> >>
> >> c) I've been working on readying my
> release,
> >> where the main bit is the
> >> object-views mapping which reuses
> the object
> >> view mapping code from
> >> couchdb-python. This is currently
> my main
> >> area of intrest. I assume
> >> this is still of use in whatever
> 'fork' of
> >> paisley ends up being the
> >> main tree ?
> >>
> >> d) I recently got a launchpad bug
> report
> >> against paisley. What bug
> >> tracker should paisley be using ?
> >>
> >> I was planning on doing that release
> soon; I'm
> >> going to a couchdb
> >> training in London next week so this
> week was
> >> a good opportunity to
> >> finish that part of the code, and
> start
> >> working on integrating change
> >> notification in the 'orm' layer.
> >>
> >> let me know,
> >> Thomas
> >>
> >>
> >>
> >> --
> >> It's not you
> >> It's the E talking
> >> --
> >> URGent, best radio on the net -
> 24/7 !
> >> http://urgent.fm/
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >> --
> >> Sean McQuillan
> >> 415.990.0854
> >>
> >>
> >>
> >>
> >>
> >>
> >
>
> > --
> >
> > --
> > It was a one way street.
> > I was going one way.
>
> > --
> > URGent, best radio on the net - 24/7 !
> > http://urgent.fm/
> >
> >
>
>
>
>
> --
> Sean McQuillan
> 415.990.0854
>
>
>
>
We did it by the door
We did it on the floor
how about tonight, maybe
we do it some more ?
--
Flumotion - the only way to stream!
http://www.flumotion.net/
--
Sean McQuillan
Thomas Vander Stichele
Apr 14, 2011, 5:53:51 AM4/14/11
to Sean McQuillan, paisley-d...@googlegroups.com
On Tue, 2011-03-22 at 03:00 -0700, Sean McQuillan wrote:
> [[added mailing list]]
>
>
> LGTM. A few thoughts, I've been endeavoring to keep the public api
> consistent, so please try to avoid changing it in non-backwards
> compatible ways and write tests for any new api.
>
> --
>
>
> A few weird couchisms I discovered writing our internal function
> load_from_view (that does basically the same thing as views.py except
> in our model layer).
>
>
> 1. include_docs = True is a lie, if you run queries with
> include_docs=True under any write load you *will* consistently get
> invalid results back, including deleted docs.
> -> in this situation, often, here will not be a 'doc' field on the
> result row and no other symptom of failure.
> 2. sometimes, in situations I don't yet understand completely due to
> lack of thorough debugging, rows contain an 'error' field, when this
> happens they don't have a 'doc' even if the kwarg include_docs=True
> was passed to couch.
> [[added mailing list]]
>
>
> LGTM. A few thoughts, I've been endeavoring to keep the public api
> consistent, so please try to avoid changing it in non-backwards
> compatible ways and write tests for any new api.
>
> --
>
>
> A few weird couchisms I discovered writing our internal function
> load_from_view (that does basically the same thing as views.py except
> in our model layer).
>
>
> 1. include_docs = True is a lie, if you run queries with
> include_docs=True under any write load you *will* consistently get
> invalid results back, including deleted docs.
> -> in this situation, often, here will not be a 'doc' field on the
> result row and no other symptom of failure.
> 2. sometimes, in situations I don't yet understand completely due to
> lack of thorough debugging, rows contain an 'error' field, when this
> happens they don't have a 'doc' even if the kwarg include_docs=True
> was passed to couch.
I'll keep those in mind, and see what I can do about catching them when
I run into them. I'll try and add some torture tests too.
Have you checked with the couchdb list if these are actual bugs, or just
side effects that they consider should be handled by the user of
couchdb ?
Thomas
--
--
The blood before me makes me open up my heart again
And I feel it coming over like a storm again
--
MOAP - Maintaining your projects since 2006
https://apestaart.org/moap/trac/
Reply all
Reply to author
Forward
0 new messages