705899 - Study correlation between number of votes and number of CC

Status: RESOLVED FIXED

(bugzilla.mozilla.org :: API, defect)

Description

[David Bruant]

I have been told once on IRC that voting for a bug was pretty much meaningless.
One idea could be to remove (or at least disable) the voting feature if the Mozilla project doesn't care about this. But before doing this, I noticed by hanging out on BMO that when bugs have a high number of votes, it also has a high number of CC and I wonder whether there was a correlation between these 2 variables (number of votes and number of cc).

But I can't easily make this study. First, apparently, the Bugzilla API does not provide the number of voters (At least, I can't find it here: https://wiki.mozilla.org/Bugzilla:REST_API:Objects#Flag).
Even if it was present, I'm not sure I could conviniently do my study. Do I need to do a request for each bug? If so, it's 700000 requests and for sure, I won't do it.

So, I'm here to ask if a solution can be found so that I can pull out of the database the data I need for my study. Either by adding things to the Bugzilla API or if someone is kind enough to do it for me direclty on the database.

What I would need is for each bug:
- date of creation
- date of RESOLVED (or status if not RESOLVED)
- number of vote
- number of CC

I'll do the analysis and will get back to you as soon as I have the info.

Comment 1

[David Bruant]

I filed this bug, because I don't see how to do my study myself. I promise that as soon as I have a way to do it myself, I won't bother you anymore :-p

Comment 2

[David Lawrence [:dkl]]

Gerv, is there any objection to providing the data being requested? I can provide a flat text file with the bits of bug data he needs to do his report. It will be from a sanitized copy of the BMO database (no private bugs, etc.).

dkl

Comment 3

[David Bruant]

(In reply to David Lawrence [:dkl] from comment #2)
> Gerv, is there any objection to providing the data being requested?
I don't think there can be, all I'm asking is public information. I could write a crawler to retrieve it, but it would be obviously ridiculously long and painful for me and BMO server/database, I could use the bugzilla API, but it does not seem to provide the vote number (and would probably be painful for the server/database as well). That's the only reason why I'm asking.

> It will be from a sanitized copy of the BMO database (no private
> bugs, etc.)
Oh yeah of course, I really don't want anything private. Plus, the study of the correlation between votes and number of cc on private bugs does not make much sense.

(In reply to David Bruant from comment #0)
> What I would need is for each bug:
> - date of creation
> - date of RESOLVED (or status if not RESOLVED)
Just to clarify, if the bug has been through RESOLVED several times, I would like the last date of RESOLVED.

> - number of vote
> - number of CC

Thank you :-)

Comment 4

[Gervase Markham [:gerv]]

I have no objection to providing this data from a sanitized database.

Indeed, the Bugzilla API does not provide vote numbers, or support voting information in any other way.

Gerv

Comment 5

[David Bruant]

(In reply to Gervase Markham [:gerv] from comment #4)
> Indeed, the Bugzilla API does not provide vote numbers, or support voting
> information in any other way.
Is it on purpose?
Is there any intention to enhance the bugzilla API to support voting information?

(In reply to David Bruant from comment #0)
> What I would need is for each bug:
> - date of creation
> - date of RESOLVED (or status if not RESOLVED)
> - number of vote
> - number of CC
May I add Product and Component, please?


(In reply to Gervase Markham [:gerv] from comment #4)
> I have no objection to providing this data from a sanitized database.
Thanks :-)

Comment 6

[Gervase Markham [:gerv]]

No intention of enhancing the API. The future direction of voting, if it goes anywhere, is into a Bugzilla extension - so votes are not supported in an API designed for the core.

Gerv

Comment 7

[David Lawrence [:dkl]]

Ok, I have the data ready to go. You can access the file at 

https://people.mozilla.com/~dkl/david_bruant_data.gz

Please close this when you have the file and I will remove it.

dkl

Comment 8

[David Bruant]

Thank you very much.
I have a question, just to be sure about the legend.
Data are of this format:
35, 1998-04-07 01:37:03, 1998-12-12 09:06:46, 0, 2

I suppose it's (in the same order):
Bug#, date of creation, date of resolution, #vote, #cc
?

You can remove the file. I'll study the data and blog about it as soon as I can. I'll add a comment on this blog. Thanks a lot.

Comment 9

[Gervase Markham [:gerv]]

Checking bug 35, that seems like a fairly sane interpretation to me :-)

Gerv

Comment 10

[David Lawrence [:dkl]]

(In reply to David Bruant from comment #8)

My bad. I should have added in the legend as the first line out the output.

> I suppose it's (in the same order):
> Bug#, date of creation, date of resolution, #vote, #cc

That is correct.

> You can remove the file. I'll study the data and blog about it as soon as I
> can. I'll add a comment on this blog. Thanks a lot.

Thanks

Comment 11

[David Lawrence [:dkl]]

(In reply to David Bruant from comment #5)
> May I add Product and Component, please?

Not doing too well this morning. Do you need a new dump with the product/component information added? I can do a new dump and put it up for you today if so.

dkl

Comment 12

[David Bruant]

(In reply to David Lawrence [:dkl] from comment #10 - 11)
> My bad. I should have added in the legend as the first line out the output.
On the other hand, gerv was right, I could have verified myself since all data is public :-p

> (In reply to David Bruant from comment #5)
> > May I add Product and Component, please?
> 
> Not doing too well this morning.
No worries :-)

> Do you need a new dump with the
> product/component information added? I can do a new dump and put it up for
> you today if so.
That would be awesome. I intuit that product/component influence whether people vote or not and how they vote. I'd like to check whether or not it's the case and how it influences votes if it's the case.

Comment 13

[David Lawrence [:dkl]]

Ok, new file for you. Close this bug when you have it. Thanks.

https://people.mozilla.com/~dkl/david_bruant_data_2.gz

dkl

Comment 14

[David Bruant]

Well received, thanks very much.