158140 - google-like importance calculation

Status: NEW

(Bugzilla :: Bugzilla-General, enhancement)

Description

[Knud Sørensen]

In bugzilla i miss a unified importance rating 
which take into account the bugs dependers and there votes.
I would like to suggest the following calculation.

Let V[n] be the votes for bug number n.

Let D be a dependency matrix with the form
if bug n blocks bug m let D[n][m]=V[m]+1 else D[n][m]=0 and D[n][n]=V[n]+1 
let I={I_0, I_1 , ... I_n} be a sequence of importance vectors
where I_n[m] is the importance of bug m in the n'te iteration.

Initialize I_0 to I_0[m]= 1.0 for all bugs m,

calculate I by 


I_(n+1)=D*(I_n/|I_n|)

This calculation convergate exponential fast to the primary eigenvector for D.

normally I_3 or I_2 is okay.

If bugzilla recalculates this every night it might be enough to make one
iteration if it starts with the vector from the last calculation.

Knud

Comment 1

[Gervase Markham [:gerv]]

Problem is, different sites would differ as to how to rate a bug as "important".
I can't see this working out...

Gerv

Comment 2

[Knud Sørensen]

Different search engines use to differ as to how to rate a page as "important".
Now everybody is using google.

This rating is very simmilar to google's PageRank just 
adapted to the present problem.

You little script which get data from the database could show you 
that this works.

When you are conviced you could add to bugzilla.

Knud 

Comment 3

[Jouni Heikniemi]

This is essentially the same concept as bug 77472.

*** This bug has been marked as a duplicate of 77472 ***

Comment 4

[Knud Sørensen]

The only thing similar between the 2 bugs is
that they both try to make a importance rating.

the approach in bug 77472 is similar
to the approach used in the old type of search engines
and this bug use a approach similar to the new type like google.  

And in my experience the later type is much better 
therefore I reopen this bug.

I could try to make a prove of concept script, 
is there a place where i can get a dump of the relevant 
parts of the database ?

Knud

Comment 5

[Jouni Heikniemi]

>the approach in bug 77472 is similar
>to the approach used in the old type of search engines
>and this bug use a approach similar to the new type like google.  

Yes, but the algorithm is not at all relevant in this level. We don't need two
competing bugs with two algoritms, we need people discussing implementation
issues in one place and then deciding which way is the best one to implement.
There is plenty of space in bug 77472 for your algorithm, too. 

Is this ok with you? I still consider this to be a duplicate. Don't get me
wrong: new ideas are highly welcome, but two bugs of the same issue is bound to
lead us to a lot of wasted work. Keeping things in one bug helps us coordinate
the patches.

>I could try to make a prove of concept script, 
>is there a place where i can get a dump of the relevant 
>parts of the database ?

I don't think you'll have an easy way to get that, but try contacting
[e-mail]. He has the ways to dump the b.m.o database.

Comment 6

[Aleksey Nogin]

I like it, I agree that this might be a good idea. As to the dups issue, IMHO it
might make more sense to close bug 92440 as a dup of this one.

Comment 7

[Knud Sørensen]

Attachment: consept script

This is a script which show the concept in this 
importance calculation.
I am not a perl programmer and I did not have a database to test it on.
So, your better consider this code as pseudo code.

The script calculate the Importance for every bug and print the top 50.

Knud Sørensen

Comment 8

[Knud Sørensen]

Hi.

Have anyone tried my script ??

Or is it posible to get a dump of mozilla's bugzilla database 
so I can test it myself ??

Knud 

Comment 9

[Dave Miller [:justdave]]

Reassigning bugs that I'm not actively working on to the default component owner
in order to try to make some sanity out of my personal buglist.  This doesn't
mean the bug isn't being dealt with, just that I'm not the one doing it.  If you
are dealing with this bug, please assign it to yourself.