Understand LHE files despite wrong XML

[johannbrehmer]

The LHE files that MadGraph produces ostensibly conform to the XML standard. But sometimes MadGraph produces headers that crudely violate this standard, for instance in this part about the weight groups:

<weightgroup name="PDF4LHC15_nlo_30" combine="symmhessian"> # 90900: PDF4LHC15_nlo_30. mem=0 ; alphas(MZ)=0.118 central value; mem=1-30 ; PDF symmetric eigenvectors
<weight id="5" MUR="1.0" MUF="1.0" PDF="90900" >  </weight>
[...]
<weight id="35" MUR="1.0" MUF="1.0" PDF="90930" > PDF=90900 MemberID=30 </weight>
<weightgroup name="Central scale variation" combine="envelope">
</weightgroup> PDF
<weight id="36" MUR="1.0" MUF="1.0" DYN_SCALE="1" PDF="90900" > dyn_scale_choice=sum pt  </weight>
[...]
<weight id="39" MUR="1.0" MUF="1.0" DYN_SCALE="4" PDF="90900" > dyn_scale_choice=sqrts  </weight>
</weightgroup>

(two lines in the middle are swapped).

I'm not sure what exactly triggers this MadGraph behavior (I've seen it when calculating PDF uncertainties with PDF4LHC, but it was fine with CTEQ14nlo and identical settings otherwise).

But this can be a problem for MadMiner, which relies on an XML parser to read the weight setup and the events. In this particular example, the weights related to the PDF uncertainties are not identified out properly.

We should

• talk to MadGraph people to understand this behavior, and
• make sure it doesn't happen, or repair the files before reading them.

[johannbrehmer]

Actually, it seems that the parsing is pretty robust to some errors in the XML. My isssue came from something else (now fixed). I'll leave this open, but it might not be super important.