[Rivet] PyPI package for YODA?

[Andy Buckley]

On 18/08/15 13:44, Graeme Watt wrote:
> Dear Andy,
>
> Do you have any comments on my message from 6th August?

I replied back then, but just noticed that you were somehow left off the 
recipient list. I've forwarded the message. Apologies.

> I noticed that
> you have released some of your other projects (PySLHA, PyPDT, PyFeyn) as
> PyPI packages, but I guess the difference is that YODA is mostly C++
> with a Python wrapper, so it is not so appropriate to release it as a
> PyPI package.

Exactly.

> I guess we can just export to the YODA format by hand (as
> in the current HepData system) rather than using the YODA package.
>
> More generally, we should discuss the interaction between the new
> HEPData and Rivet/YODA.  Do we actually need to provide conversion
> to/from the YODA format, or would this be better handled as part of the
> YODA package?  Since you already provide conversion scripts like
> "aida2yoda" and "flat2yoda", could you also provide (with our help) a
> "hepdatayaml2yoda" script?

We're trying not to proliferate these scripts, and I think it is very 
useful to have YODA directly downloadable from HepData -- it is 
certainly not the most obscure format currently supported! But yes we 
should definitely discuss, particularly in the context of the HD rewrite.

> That way it is easier for you to make0
> changes if you extend the YODA format (for example, to provide a
> breakdown of uncertainties instead of just using the total
> uncertainty).  It might also be good to provide a "rivet2hepdatayaml"
> script to prepare a HEPData submission given a Rivet analysis since
> there seems to be a lot of overlap in the necessary metadata (and
> reference data).

Yes, this could work for both of us.

> As I've mentioned before, some people seem to be departing from the
> intended workflow of first submitting data to HepData, then using the
> exported YODA format and automatic histogram names for inclusion in a
> Rivet analysis.  Instead, they just prepare the Rivet analysis directly,
> including the reference data, and don't bother submitting to HepData (or
> do it much later, with a mismatch in the automatic histogram names).
> Could you maybe check whether reference data included in future
> submitted Rivet analyses is already available in HepData?  Ideally you
> would make HepData submission of reference data to be a mandatory
> requirement for a Rivet analysis to be part of the released package.

At some level we need the experiments to be the standard enforcers on 
their own submissions, but we would like to be in a situation where we 
can e.g. re-dump ~all of our experiment reference data from HD on a 
regular basis, so we pick up any upstream improvements. This would also 
be good for all of us, but we need some development & testing work.

Andy



> On 08/06/15 14:21, Graeme Watt wrote:
>> Dear YODA authors,
>>
>> Michal, a CERN summer student, is working on a Python tool for
>> converting the new HEPData YAML format to the YODA format for use with
>> Rivet:
>>
>>    https://github.com/HEPData/hepdata-converter/issues/5
>>
>> The inverse conversion (i.e. YODA to HEPData YAML) would also be useful
>> to prepare a HEPData submission given an existing Rivet analysis:
>>
>>    https://github.com/HEPData/hepdata-converter/issues/10
>>
>> Would it be possible to release the YODA Python interface as a PyPI package?
>>
>> Thanks and best regards,
>> Graeme
>>
>


-- 
Dr Andy Buckley, Lecturer / Royal Society University Research Fellow
Particle Physics Expt Group, University of Glasgow