9. Belle II File Format¶
The official standard Belle II file format is the mini data-summary table (mdst).
There is a basf2 package (the mdst package) which defines this file format. An mdst file contains a curated list of post-reconstruction dataobjects which are provided for analysis use. The dataobjects are particularly important as they are the (only) information that is provided for high-level analysis.
These mdst dataobjects are optimised for minimal disk size per event. This is important because the Belle II experiment will collect around \(5\times10^{10}\) events, so data file size considerations will become rather critical.
9.1. Python interface¶
- mdst.add_mdst_dump(path, print_untested=False)[source]¶
Add a PrintObjectsModule to a path for printing the mDST content.
- Parameters
path (basf2.Path) – Path to add module to
print_untested (bool) – If True print the names of all methods which are not explicitly printed to make sure we don’t miss addition of new members
- mdst.add_mdst_output(path, mc=True, filename='mdst.root', additionalBranches=[], dataDescription=None)[source]¶
Add the mDST output module to a path. This function defines the mDST data format.
- Parameters
path (basf2.Path) – Path to add module to
mc (bool) – Save Monte Carlo quantities? (MCParticles and corresponding relations)
filename (str) – Output file name.
additionalBranches (list) – Additional objects/arrays of event durability to save
dataDescription (dict or None) – Additional key->value pairs to be added as data description fields to the output FileMetaData
9.2. C++ dataobjects¶
The post-reconstruction dataobjects are C++ classes found in the mdst dataobjects directory:
ls $BELLE2_RELASE_DIR/mdst/dataobjects/include
Or in the C++ doxygen documentation here.
9.3. ROOT and compatibility guarantee¶
Mdst files are written by the RootOutput module and are based on the ROOT file format.
However it is important to note that analysis of mdst with any software other than basf2 is not supported or permitted.
I.e. use of the basf2 framework and the Analysis package is mandatory.
Warning
A common misconception: Opening an mdst file with standard ROOT tools (e.g. with a TBrowser) may initially “work”, but the results are not reproducible.
Many dataobject member accessors require the basf2 environment to return meaningful values. Segmentation faults and double-counting of tracks and/or clusters are very likely to be encountered.
However! Backward-compatibility is guaranteed for two major basf2 releases and the supported light releases.
See also
Backward compatibility on confluence.