Thrift and Python love

[Terry Jones <te...@jon.es>]

Hi all!

There are at least half a dozen issues in Jira that relate to Python, all
with patches, but these don't seem to get to the top of the attention stack
for the Thrift team. Is that right? My apologies if not, it's just that a
couple of those patches are things I'd like to be using without needing to
keep a patched local version of Thrift.

There are people who'd be happy to help out, if help is needed/wanted.

Regards & thanks,
Terry

Re: Thrift and Python love

[Esteve Fernandez <es...@sindominio.net>]

On Wednesday 15 October 2008 22:37:58 David Reiss wrote:
> FWIW, I just received this message from someone at Facebook:
>
> - I was dragged into some other stuff, but as I have a code running under
> twisted, and want to expose data I am collecting via thrift, it is
> important for me. I will go back to this either this weekend, or early next
> week.

Great! We don't know if Twisted+Thrift is being used elsewhere than in 
Fluidinfo, so it's nice to see that there are other users interested in it.

I'll happily add more examples and/or tests if needed.

Cheers.

Re: Thrift and Python love

[David Reiss <dr...@facebook.com>]

FWIW, I just received this message from someone at Facebook:

- I was dragged into some other stuff, but as I have a code running under twisted,
and want to expose data I am collecting via thrift, it is important for me.
I will go back to this either this weekend, or early next week.

David Reiss wrote:
> I feel that this would be acceptable, and that the conflict of interest
> would be minor at most.
> 
> --David
> 
> Terry Jones wrote:
>> Hi David
>>
>>> As to the specific issue of the Twisted patch, this is a pretty significant
>>> amount of code that takes some specialized knowledge to review competently.
>>> I know someone who might be able to invest the time to do so, but I think
>>> contributors are going to have to be patient with such difficult-to-review
>>> patches.
>> I'm qualified to review the Twisted patch, and happy to do so.
>>
>> BUT, Esteve works with me, so I'm not fully independent. However, I didn't
>> have any input into that code - apart from agreeing that it was something
>> he should do. I have not looked at the code at all, though I've used it
>> with success.
>>
>> If I'm considered independent enough that my review would count (and I
>> meant count in the sense of getting the patch across the line, not just
>> count as an additional thumbs up), I'll go for it. I don't mean that in
>> negative sense. I mean that if I'm not considered independent enough, I'd
>> rather put time to better use, like reviewing some other Python patch where
>> there'd be no question of conflict of interest, etc.
>>
>> Regards,
>> Terry

Re: Thrift and Python love

[David Reiss <dr...@facebook.com>]

I feel that this would be acceptable, and that the conflict of interest
would be minor at most.

--David

Terry Jones wrote:
> Hi David
> 
>> As to the specific issue of the Twisted patch, this is a pretty significant
>> amount of code that takes some specialized knowledge to review competently.
>> I know someone who might be able to invest the time to do so, but I think
>> contributors are going to have to be patient with such difficult-to-review
>> patches.
> 
> I'm qualified to review the Twisted patch, and happy to do so.
> 
> BUT, Esteve works with me, so I'm not fully independent. However, I didn't
> have any input into that code - apart from agreeing that it was something
> he should do. I have not looked at the code at all, though I've used it
> with success.
> 
> If I'm considered independent enough that my review would count (and I
> meant count in the sense of getting the patch across the line, not just
> count as an additional thumbs up), I'll go for it. I don't mean that in
> negative sense. I mean that if I'm not considered independent enough, I'd
> rather put time to better use, like reviewing some other Python patch where
> there'd be no question of conflict of interest, etc.
> 
> Regards,
> Terry

Re: Thrift and Python love

[Terry Jones <te...@jon.es>]

Hi David

> As to the specific issue of the Twisted patch, this is a pretty significant
> amount of code that takes some specialized knowledge to review competently.
> I know someone who might be able to invest the time to do so, but I think
> contributors are going to have to be patient with such difficult-to-review
> patches.

I'm qualified to review the Twisted patch, and happy to do so.

BUT, Esteve works with me, so I'm not fully independent. However, I didn't
have any input into that code - apart from agreeing that it was something
he should do. I have not looked at the code at all, though I've used it
with success.

If I'm considered independent enough that my review would count (and I
meant count in the sense of getting the patch across the line, not just
count as an additional thumbs up), I'll go for it. I don't mean that in
negative sense. I mean that if I'm not considered independent enough, I'd
rather put time to better use, like reviewing some other Python patch where
there'd be no question of conflict of interest, etc.

Regards,
Terry

Re: Thrift and Python love

[Doug Cutting <cu...@apache.org>]

David Reiss wrote:
> I'd be fine with adding more committers, but I'm not convinced that
> too few committers is the problem.

More committers is better since it can spread the binding review load 
around.  While anyone can review a patch, the ability to act on reviews 
does motivate folks.  Put alternately, if someone's a committer then 
they can no longer complain about a slow review process!

The way the Apache Pyramid Scheme works is that, if committers review 
enough of a contributors patches to find that they're committer 
material, then they can make them a committer, to help with reviewing.

The best practice in Jira is to, as soon as someone contributes a patch, 
give them the "Contributor" role in Jira on the project and assign them 
that issue and any others where they contribute patches.  Then you can 
use Jira to see how many patches someone has contributed (since nice 
reports are available by assignee, but not by attacher-of-patches) and 
what percentage of those patches got committed, and whether the dialog 
around them was civil, etc.

I see Dave Engberg has contributed a bunch of patches, but he's not been 
assigned any issues nor is he yet a contributor in Jira so that he can 
be assigned issues...

Doug

Re: Thrift and Python love

[David Reiss <dr...@facebook.com>]

I've reviewed and commented on or committed a lot of the open bugs
with patches.  I'll continue to try to keep this queue short.

Ben Maurer is the other committer with Python expertise.  He is also
the only one who knows C#.

I'd be fine with adding more committers, but I'm not convinced that
too few committers is the problem.  Anyone can read over a submitted
patch and provide feedback (or just comment to say they're read it
and think it is good).  If an issue has a lot of discussion, revision,
and finally approval, it becomes a lot easier for a committer to
rubber-stamp the patch.  I'll try to get more Thrift users at Facebook
to subscribe to the dev list, but this is something that non-committers
can do on their own.

As to the specific issue of the Twisted patch, this is a pretty significant
amount of code that takes some specialized knowledge to review competently.
I know someone who might be able to invest the time to do so, but I think
contributors are going to have to be patient with such difficult-to-review
patches.

--David

Kevin Clark wrote:
> How many committers do we have that know python? Someone must own that
> code, right? I thought the Facebook guys do - dreiss? I can't handle
> them since my knowledge is simple at best.
> 
> I realize I've got a couple of Ruby patches sitting around that I'll
> review tomorrow (156,157).
> 
> 
> --
> Kevin Clark
> http://glu.ttono.us

Re: Thrift and Python love

[Kevin Clark <ke...@gmail.com>]

How many committers do we have that know python? Someone must own that
code, right? I thought the Facebook guys do - dreiss? I can't handle
them since my knowledge is simple at best.

I realize I've got a couple of Ruby patches sitting around that I'll
review tomorrow (156,157).


-- 
Kevin Clark
http://glu.ttono.us

Re: Thrift and Python love

[Terry Jones <te...@jon.es>]

Hi all

> Is there any interest in a commit-fest, where we can work through the
> backlog of patches in an flurry of commits/discards?

Sounds good.  We'd be happy to help.

Terry

Re: Thrift and Python love

[Ian Pye <ia...@gmail.com>]

Is there any interest in a commit-fest, where we can work through the
backlog of patches in an flurry of commits/discards?

On Fri, Oct 10, 2008 at 4:16 AM, Esteve Fernandez <es...@sindominio.net> wrote:
> On Tuesday 07 October 2008 17:15:43 Bryan Duxbury wrote:
>> I agree that we have a problem. I think that we might need more
>> committers, to be honest. There are a lot more libraries than there
>> are people in charge at the moment, so clearly any patch that needs
>> to get committed will have to go through one of these few people. It
>> would also be nice to have at least a clear point person for every
>> library so there's someone to email. I know that we have people in
>> charge of a few libraries, but definitely not all of them.
>
> I completely agree with you. It's obvious that some of the people who report
> issues have a steady interest in improving particular areas of Thrift. For
> example, we at Fluidinfo use Python heavily, and have already contributed
> code that may not be of interest to all those who are subscribed to this
> list, but that might be useful to others who are not deeply involved in the
> Thrift community.
>
> I'm particularly thinking of our Twisted code (sorry if it sounds pushy, but
> it's just an example that describes this situation), which may help those who
> are already using Twisted and cannot use the standard blocking RPC mechanism
> that comes with Thrift. It's a relatively obscure combination
> (Thrift+Twisted), but we believe it could be useful. Also, we want to release
> more stuff that uses this as free software, but we don't want to track Thrift
> ourselves and maintain a forked version, though.
>
> Cheers.
>
>

Re: Thrift and Python love

[Esteve Fernandez <es...@sindominio.net>]

On Tuesday 07 October 2008 17:15:43 Bryan Duxbury wrote:
> I agree that we have a problem. I think that we might need more
> committers, to be honest. There are a lot more libraries than there
> are people in charge at the moment, so clearly any patch that needs
> to get committed will have to go through one of these few people. It
> would also be nice to have at least a clear point person for every
> library so there's someone to email. I know that we have people in
> charge of a few libraries, but definitely not all of them.

I completely agree with you. It's obvious that some of the people who report 
issues have a steady interest in improving particular areas of Thrift. For 
example, we at Fluidinfo use Python heavily, and have already contributed 
code that may not be of interest to all those who are subscribed to this 
list, but that might be useful to others who are not deeply involved in the 
Thrift community.

I'm particularly thinking of our Twisted code (sorry if it sounds pushy, but 
it's just an example that describes this situation), which may help those who 
are already using Twisted and cannot use the standard blocking RPC mechanism 
that comes with Thrift. It's a relatively obscure combination 
(Thrift+Twisted), but we believe it could be useful. Also, we want to release 
more stuff that uses this as free software, but we don't want to track Thrift 
ourselves and maintain a forked version, though.

Cheers.

Re: Thrift and Python love

[Bryan Duxbury <br...@rapleaf.com>]

I agree that we have a problem. I think that we might need more  
committers, to be honest. There are a lot more libraries than there  
are people in charge at the moment, so clearly any patch that needs  
to get committed will have to go through one of these few people. It  
would also be nice to have at least a clear point person for every  
library so there's someone to email. I know that we have people in  
charge of a few libraries, but definitely not all of them.

Additionally, for those of you who watch the list, if you see someone  
post a new patch, please try and take some time to review it once and  
a while. It's a lot easier on the committers if the rest of the  
community has already given the +1 on the patch.

-Bryan

On Oct 7, 2008, at 7:56 AM, Michael Greene wrote:

> Hi,
>
> Unfortunately this is typical of most of the issues, and not  
> specific to
> Python.  Pulling up a list of the issues that have been closed  
> fixed in the
> past two months in JIRA [
> https://issues.apache.org/jira/secure/IssueNavigator.jspa? 
> reset=true&&pid=12310800&updated% 
> 3Aprevious=-8w&status=5&status=6&resolution=1&sorter/ 
> field=updated&sorter/order=DESC]
> is particularly revealing.
>
> Only two individuals have had issues resolved fixed in the past two  
> months.
> Excluding an issue committed and resolved without any voting or  
> mailing list
> discussion, only one individual has had patches committed in the  
> last two
> months.
>
> Is the project slowing down?
>
> Michael
>
> On Tue, Oct 7, 2008 at 6:00 AM, Terry Jones <te...@jon.es> wrote:
>
>> Hi all!
>>
>> There are at least half a dozen issues in Jira that relate to  
>> Python, all
>> with patches, but these don't seem to get to the top of the  
>> attention stack
>> for the Thrift team. Is that right? My apologies if not, it's just  
>> that a
>> couple of those patches are things I'd like to be using without  
>> needing to
>> keep a patched local version of Thrift.
>>
>> There are people who'd be happy to help out, if help is needed/ 
>> wanted.
>>
>> Regards & thanks,
>> Terry
>>

Re: Thrift and Python love

[Michael Greene <mi...@gmail.com>]

Hi,

Unfortunately this is typical of most of the issues, and not specific to
Python.  Pulling up a list of the issues that have been closed fixed in the
past two months in JIRA [
https://issues.apache.org/jira/secure/IssueNavigator.jspa?reset=true&&pid=12310800&updated%3Aprevious=-8w&status=5&status=6&resolution=1&sorter/field=updated&sorter/order=DESC]
is particularly revealing.

Only two individuals have had issues resolved fixed in the past two months.
Excluding an issue committed and resolved without any voting or mailing list
discussion, only one individual has had patches committed in the last two
months.

Is the project slowing down?

Michael

On Tue, Oct 7, 2008 at 6:00 AM, Terry Jones <te...@jon.es> wrote:

> Hi all!
>
> There are at least half a dozen issues in Jira that relate to Python, all
> with patches, but these don't seem to get to the top of the attention stack
> for the Thrift team. Is that right? My apologies if not, it's just that a
> couple of those patches are things I'd like to be using without needing to
> keep a patched local version of Thrift.
>
> There are people who'd be happy to help out, if help is needed/wanted.
>
> Regards & thanks,
> Terry
>