[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#565219: qa.debian.org: bug history graphs are incorrect



On Sat, 30 Oct 2010 09:21:45 +0200 Mike Hommey wrote:

> On Sat, Oct 30, 2010 at 12:10:43AM +0200, Francesco Poli wrote:
> > On Thu, 28 Oct 2010 16:49:00 +0200 Mike Hommey wrote:
[...]
> > > Tried to set all of the above on the new graphs on
> > > http://qa.debian.org/data/bts/graphs
> > 
> > They look fairly nice, IMHO.
> > 
> > I am still wondering why the old graphs (the ones currently on
> > http://people.debian.org/~glandium/bts ) seem to be still smarter with
> > respect to the y-axis grid.
> > 
> > I mean: the old graphs seem to almost always guess optimal y-axis grid
> > lines and labels, while the new graphs look nice but not optimal...
> 
> Would you have examples of such optimal vs. suboptimal graphs ?

Here we go: let the old graphs (the one currently linked from the PTS)
be named "old", the previous experimental graphs be named "test-0", and
the current experimental graphs be named "test-1".

Hence, we have the following general naming scheme:

old     http://people.debian.org/~glandium/bts/
test-0  http://merkel.debian.org/~glandium/bts/
test-1  http://qa.debian.org/data/bts/

With this naming convention, I will comment on some examples, with
respect to one feature *only*, that is to say, number and placement of
horizontal grid lines and y-axis numeric labels.
Needless to say, as far as *all* other features are concerned, I think
that test-1 graphs are better than the other two, as I've already
stated previously...

Examples:

old     http://people.debian.org/~glandium/bts/a/apt-listbugs.png
test-0  http://merkel.debian.org/~glandium/bts/a/apt-listbugs.png
test-1  http://qa.debian.org/data/bts/graphs/a/apt-listbugs.png

 * old numeric labels are well chosen and old grid lines are good
 * test-0 numeric labels are well chosen, but test-0 grid lines are too
   many
 * test-1 numeric labels and grid lines are too few

old     http://people.debian.org/~glandium/bts/a/aptitude.png
test-0  http://merkel.debian.org/~glandium/bts/a/aptitude.png
test-1  http://qa.debian.org/data/bts/graphs/a/aptitude.png

 * numeric labels are well chosen for all of the three versions
 * old and test-0 grid lines are OK, but test-1 grid lines are too few

old     http://people.debian.org/~glandium/bts/d/dput.png
test-0  http://merkel.debian.org/~glandium/bts/d/dput.png
test-1  http://qa.debian.org/data/bts/graphs/d/dput.png

 * old and test-1 numeric labels and grid lines seem to be reasonable
 * test-0 has maybe a bit too many labels and lines...

old     http://people.debian.org/~glandium/bts/r/rpncalc.png
test-0  http://merkel.debian.org/~glandium/bts/r/rpncalc.png
test-1  http://qa.debian.org/data/bts/graphs/r/rpncalc.png

 * old and test-0 make not much sense, since a package cannot have,
say, 1.6 bugs (hence it's not really useful to have labels or lines for
fractions of a bug)
 * test-1 is better in this case, since it has meaningful labels, even
though grid lines are too many

old     http://people.debian.org/~glandium/bts/w/wnpp.png
test-0  http://merkel.debian.org/~glandium/bts/w/wnpp.png
test-1  http://qa.debian.org/data/bts/graphs/w/wnpp.png

 * test-0 is hard to read (too many labels and grid lines), even though
it makes perfectly sense
 * old and test-1 are similar (even though I slightly prefer test-1)
and are easier to read

Well, I could go on, but I think the above examples are enough to get
the big picture.
After all, "test-1" is not bad and compares well with "old".
The only version I don't like is "test-0".

In summary, I think that test-1 could be considered OK, unless we
manage to find a way to further enhance the result...

> 
> > Have you tried to remove the "-Y also known as --alt-y-grid" option
> > and see the default result?
> 
> The graphs on merkel are without -Y.
> (http://merkel.debian.org/~glandium/bts/)

OK, as I said above, I don't like the result, so let's forget about
removing the -Y option, unless we find some other, better, option!
;-)


-- 
 http://www.inventati.org/frx/progs/scripts/pdebuild-hooks.html
 Need some pdebuild hook scripts?
..................................................... Francesco Poli .
 GnuPG key fpr == C979 F34B 27CE 5CD8 DC12  31B5 78F4 279B DD6D FCF4

Attachment: pgpUmuvekmCQq.pgp
Description: PGP signature


Reply to: