<div dir="ltr"><div>Hi David, all,</div><div><br></div>Making this general for any function of 2 histograms (not just division) is tricky, no?<div><br></div><div>I think the option someone mentioned a few months ago was to only allow access to *booked* histograms in finalize(). That way you are guaranteed to be able to rerun finalize() on the output yoda file(s).<br><br>Chris</div><div class="gmail_extra"><br><div class="gmail_quote">On Wed, Jan 20, 2016 at 5:03 PM, David Bjergaard <span dir="ltr"><<a href="mailto:david.bjergaard@gmail.com" target="_blank">david.bjergaard@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Why can't there be custom 1D booking for ratios that is implemented in the software that writes out name_num and name_den? Then users can yodamerge to their hearts content.<br>
<br>
David<br>
<div><div><br>
> On Jan 20, 2016, at 11:56 AM, Holger Schulz <<a href="mailto:holger.schulz@durham.ac.uk" target="_blank">holger.schulz@durham.ac.uk</a>> wrote:<br>
><br>
>> On 20/01/16 16:51, David Grellscheid wrote:<br>
>> Hi Holger,<br>
>><br>
>> Is your question why the result of dividing two histos is a scatter rather than a histogram?<br>
>> It gets hard to define many of the properties that a histogram / bin should have after the division.<br>
><br>
> Sorry, I was unprecise.<br>
><br>
> Specifically, what people want to do (for observables that are defined as ratios)<br>
> is to do some yodamerge stuff for the denominator and separately some yodamerge<br>
> stuff for the numerator. For instance adding background and signal contributions.<br>
> That requires output in the form of HISTO1D.<br>
><br>
> I.e. do we have/ can we add a switch which disables the divide call in analyses and<br>
> instead writes out the histo1D guys of the numerator and the denominator such<br>
> that one can do arbitrary histogram manipulation afterwards?<br>
><br>
> Cheers,<br>
> Holger<br>
><br>
>> David<br>
>><br>
>><br>
>><br>
>>> On 20 January 2016 13:47:20 GMT+00:00, Holger Schulz<<a href="mailto:holger.schulz@durham.ac.uk" target="_blank">holger.schulz@durham.ac.uk</a>> wrote:<br>
>>>> On 20/01/16 13:44, Andy Buckley wrote:<br>
>>>> Er, it already does. If you call bookHisto1D, a Histo1D will be<br>
>>>> written out. Do you have a counterexample?<br>
>>> Ah sorry I forgot one important bit of information and that is the call<br>
>>><br>
>>> of divide in finalize.<br>
>>><br>
>>> Sorry, overworked already ...<br>
>>><br>
>>> Holger<br>
>>>> A<br>
>>>><br>
>>>> On 20 Jan 2016 13:30, "Holger Schulz"<<a href="mailto:holger.schulz@durham.ac.uk" target="_blank">holger.schulz@durham.ac.uk</a><br>
>>>> <mailto:<a href="mailto:holger.schulz@durham.ac.uk" target="_blank">holger.schulz@durham.ac.uk</a>>> wrote:<br>
>>>><br>
>>>> Hi,<br>
>>>><br>
>>>> is there a way to have rivet write out YODA Histo1D at the end of<br>
>>>> a run<br>
>>>> rather than Scatter2D? (I think the experimental --histo-interval<br>
>>>> kind of does<br>
>>>> that, right?)<br>
>>>><br>
>>>> I am asking because of post processing concerns brought forward<br>
>>> by<br>
>>>> some<br>
>>>> people at IPPP when using yodamerge.<br>
>>>><br>
>>>> Cheers,<br>
>>>> Holger<br>
><br>
</div></div>> _______________________________________________<br>
> Rivet mailing list<br>
> <a href="mailto:Rivet@projects.hepforge.org" target="_blank">Rivet@projects.hepforge.org</a><br>
> <a href="https://www.hepforge.org/lists/listinfo/rivet" rel="noreferrer" target="_blank">https://www.hepforge.org/lists/listinfo/rivet</a><br>
_______________________________________________<br>
Rivet mailing list<br>
<a href="mailto:Rivet@projects.hepforge.org" target="_blank">Rivet@projects.hepforge.org</a><br>
<a href="https://www.hepforge.org/lists/listinfo/rivet" rel="noreferrer" target="_blank">https://www.hepforge.org/lists/listinfo/rivet</a><br>
</blockquote></div><br></div></div>