In this post I am looking for advices, sorry if this is not the right place to do so but I don’t know wich other Forum is best.
I am working on a project where C++, HDF5 and Python are used. Persistency with HDF5 cannot be changed since it involves many differents other related projects. There is no way I can change that, and besides HDF5 is pretty good in our field since we have no need for TTrees-like object.
However I believe we can gain a lot using CINT, as an automatic dictionnary generator (to be used with PyROOT instead of our not-working SWIG) and streamer generators for objects. Right now we have to generate everything by hand, it is very painful, error prone and we don’t have automatic shema evolution. Basically I am trying to get people using ROOT in the project with HDF5 contraints. I have already convinced some people that Root classes through PyROOT can be great for data post-processing and with the help of the PyROOT Forum I was also able to generate Python dictionnaries for our own C++ classes. I still have the persistency problem at hand.
So here are my 2 questions:
Do you think if it is feasible to create a “THDF5File” object in a reasonable amount of time (If I use a trainee working with me for 3 months for instance) and if yes, where should I start? For instance there is a TXMLFile that does what I want but in XML. Is it a good start for me to study TXMLFile source files eventhough TXMLFile doesn’t support directories (and I need them)?
I read on ROOT discussions that HDF5 is used with ROOT in AstroRoot. Those people also have HDF5 constraints. Do you know if they have already some solutions and should I contact them? It is hard to tell from their homepage what is their FITS format.
ROOT could be a BIG help for Persistencey in our project. I have made some research and there is nothing comparable to TFILE when it comes to C++ object persistency, even BOOST::Serialization is far behind.
Thank for your help