[Rivet] Query/fix on behaviour of plotparser

[Frank Siegert]

Hi Jon,

I think this is fine.

If I understand correctly you don't remove the looping over the
"addfiles". Those are necessary for rivet-mkhtml -c (and
rivet-cmphistos -c) and should indeed be allowed to overwrite any
previous settings like currently implemented.
For the plotpaths on the other hand I don't see a good reason for the
current behaviour and have not used that functionality to combine
multiple plot files myself.

Cheers,
Frank



On 18 August 2017 at 15:48, Jonathan Butterworth
<J.Butterworth at ucl.ac.uk> wrote:
> Hi all,
>
> A while ago I mentioned that I was having problems with rivet apparently not
> respecting the users' choice of paths for plotinfo files.
> Specifically, ATLAS_2012_I1203852.plot is buggy in the rivet release and
> this causes problems for Contur, in that the plotparser we use doesn't
> appear to pick up our fixed version of our file. (A fix has also been
> committed to rivet.)
>
> The relevant file in rivet is plotinfo.py
>
> The problem is not that the path isn't known to rivet, but that the loop in
> the getSection method continues and reads all versions of the plotinfo file
> that it finds. The problem then is;
>
> - it reads the user version first, which is then potentially overwritten by
> the official (buggy) version.
>
> I tried reversing the loop direction (there's even a comment already there
> to this effect) but then there is still the case that
>
> - if the problem with the official version is actually an unwanted
> attribute, this won't get overwritten/removed by the fixed version anyway.
>
> ... which is actually the problem here.
>
> I have attached a version of plotinfo.py which works for contur now because
> it reads only the first valid plotinfo file it finds which is the users
> version. (The user also can still feed extra files to this method, which
> will be also searched for plot attributes.)
>
> However, this isn't a simple bug fix, it is a change in behaviour, which
> might have implications elsewhere so I thought I should consult here before
> committing.
>
> If no one care enough to comment, I'll commit it anyway, since this seems
> like the optimal behaviour to me.
> Cheers,
> Jon
>
>
>
> Cheers,
> Jon
>