lucenenet-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Itamar Syn-Hershko <ita...@code972.com>
Subject Re: Lucene 4.0
Date Thu, 06 Jun 2013 21:04:42 GMT
Thanks Prescott for bringing this up again :)

Paul, the problem is no one can really know what it would take until they
have deep dived into the work, and even then decisions could and will
change. I will strongly advise against starting from scratch, although I do
think a lot in the current structure should change, but its going to be
much easier to change it in place, refactoring where needed, than starting
all over again. Once we kicked this off I personally will be happy with you
taking the analysis part of Lucene and porting it, its pretty much self
contained.
Re 3.6.2 work - you can just do that on a fork and send us PRs, its much
more straight forward than the v4 upgrade

Marcos, porting class by class is the fastest way to do this, we can then
concentrate on .NETifying and optimizing using .NET constructs.

That said, I think the way to go is create a branch out of the current git
master HEAD and label it "3.x", and start working on master towards a 4.3
compatible version. The actual port should be using a process that ensures
all Java classes are ported with their tests, and that those tests pass -
but I'm against committing any Java code to our repositories. The process
should probably be working on classes in some order (alphabetically, or
core classes first) and getting each class to compile before moving
forward. I don't mind about the project not being compilable for a month or
two.
Once this is done a process of .NETifying and proofing the code can be
started, at which point we will already have a working v4 version so it
will be easier to keep up with the Java project.

The first step IMO is to stabilize the test suite so tests could more or
less be copied and pasted (e.g. implementing Java-like assertEquals methods
etc; I find xUnit to be much easier to work with than NUnit). I already did
some work there but there's still a lot to do.

Unfortunately I can't dedicate time myself at this point, but I should be
back in business in August, at which point I can dedicate several hours a
week. Until then I'll be happy to watch closely and even coordinate the
work to some extent.




On Thu, Jun 6, 2013 at 10:41 PM, Marcos Lima <marcoslimagon@gmail.com>wrote:

> It really sounds good to me, this is a kick start =). I haven't contributed
> anything
> yet, but I would like to help you all to get this job done.
>
> I'm completely agree with Paul and Prescott.
>
> I know that there is a high commitment for keep the retrocompatibility on
> Lucene. Does Java Lucene API gets big changes every release?
>
> Is the Lucene.Net a port from a stable version from a Lucene version,
> right? If the Lucene API gets only minor changes every stable release (or
> keep most of its source-code), we could compare the versions, check the
> differences and bring them to Lucene.Net. Again, I haven't contributed with
> yet, so I don't know how it is (just an idea).
>
> What's your approach for implement simple performance improvements (without
> adding references or changing methods signatures)? Does it could be done,
> or "follow the java version only"?
>
>
>
> 2013/6/6 Paul Irwin <pirwin@feature23.com>
>
> > This is just an "outsider" suggestion as I haven't contributed anything
> > yet, although I will definitely help with the 4.x work as I have a vested
> > interest in seeing that get completed. I think there should be one person
> > (or perhaps two) guiding what the structure and workflow will look like
> to
> > avoid chaos. If the 4.x work is going to be starting from scratch as a
> > fresh port, that person should set up the project, get everything going
> in
> > source control, create the folder structure, maybe stub out some classes,
> > etc. Then divide and conquer with anyone interested in contributing,
> > perhaps by namespace.
> >
> > I like the idea of throwing the java code in there so it's easy to
> > reference.
> >
> > Again, I can work on Lucene.Net.Documents, Lucene.Net.Analysis, or
> > Lucene.Net.Store -- or any others, those are just the ones I'm most
> > familiar with the inner workings. Tell me what to do and I'll get started
> > on my fork.
> >
> > Paul Irwin
> >
> >
> > On Thu, Jun 6, 2013 at 2:38 PM, Prescott Nasser <geobmx540@hotmail.com
> > >wrote:
> >
> > > Hey guys -
> > > I know I've been MIA a little while. We have a board report due soon -
> I
> > > think it prudent that we advise them that we seem to have stalled
> > somewhat.
> > > We've got a few low hanging items out of of jira and have been
> responsive
> > > on the mailing list to community questions, but I don't think we've
> done
> > > anything to move forward with 4.0.
> > > To that end - I'd like to *try* and start us back up moving forward.
> What
> > > is the best way to accomplish this? If we took the java lucene 4.0 code
> > and
> > > committed that java to our branch and then just let people pull that
> down
> > > and being changing / modifying is that one way to maybe make some
> forward
> > > progress?
> > > ~P
> >
>
>
>
> --
> --
> Marcos Lima
> Software Developer/Tech Lead
> marcoslimagon@gmail.com
> tel: +55 (19) 9798-9335
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message