The track event model dates back to ~1999. And over time its design has been adapted on several occasions and for several reasons:
In the meantime the need for a new event model has been recognized; the main focus being the convergence/combination of the online and offline track models.
The line of thought is the combination of good ideas from both the online and offline codes having in mind the different requirements - offline (online) emphasis on resolution and precise covariance matrices (speed).
Many discussions were held and mails exchanged in the course of 2004 (mainly contributers: Jeroen, Jose, Marcel, Marco, Matthew, Olivier). But no concrete implementations/new proposals were made until ...
Jose Hernando made a first concrete proposal for base classes for tracks, states, nodes and clusters (= measurements). Two new ideas/concepts were here introduced:
All the details in the "Doxygen documentation" style can be found here.
After many discussions and mails exchanged, a very first common proposal was made. This proposal is rather preliminary and incomplete. But it serves for ongoing discussions and further developments.
The first version of the proposed classes - only "TrTrack" and "TrState", so far - can be found here.
This version is still being developed.
The details of the implementation of the classes as of this moment can be found here.
At this stage, and given the feedback and ongoing discussions, the proposal has been sub-divided in 2 main parts:
The general ideas, goals, proceeding steps and timescale were presented at the reconstruction meeting of the 2nd November.
The details of the implementation of the "end-user" classes now proposed can be found here. Also available is the corresponding XML file.
The details of the implementation of the "end-user" classes proposed for our first step of the proposal can be found here. Also available is the corresponding XML file.
At this point the status and planning have been reviewed. The base classes for tracks and states are essentially ready (apart from details that do not affect the user interface), as well as converters to "jump" from the old model to the new model.
We should now proceed towards a final agreement on these and towards first real tests and implementation/update of the several tracking packages concerned.
The details of the implementation of the "end-user" classes proposed for our first step of the proposal can be found here. Also available are the corresponding XML files for the State and the Track classes.
The first version of these "basic" classes has been committed to CVS and is now available/released:
Compared to the extrapolator interface for the "old" event model, this version is at a different place to avoid clashes in the code. Indeed all new and old event models will coexist at least for the time being, so that comparisons can easily be made.
As mentioned earlier, the plan of the proposal has been separated in two parts, the "end-user" and the "pattern recognition" classes.
The "end-user" classes proposed some time ago have been changed and updated according to needs (minor changes). The tools for extrapolation (inheriting from the interface) have also been adapted to the new model.
This new proposal presents the first version of some of the classes and tools needed for the "pattern recognition" software. The main ideas have been presented at past talks.
The details of the proposed implementation can be found here. Note that some files have suffered from a slight change in name to avoid clashes with the present code.
This quick update reflects mainly the new classes introduced in the TrackFitEvent package and some minor changes/improvements/updates to other classes such as TrackEvent and TrackTools. The details are here.
Compared to April this last version of the proposal introduces many of the classes and algorithms needed for the track fitting software, in particular the implementations of the XxxMeasurements and the corresponding projectors. Note that some of the changes to the track and state base classes discussed at the last meeting have not yet been incorporated. The details are here.
Some substantial changes have been made to the actual implementation of the Track and State classes (e.g. all Track.h methods related to the "physics state", and the enums in State.h). The details are here.
The set of classes presented at this moment reflect the status of the implementation after the discussions (and subsequent agreements) at the LHCb tracking workshop in Orsay and the software week at CERN. Note that the bulk of these packages are now part of the official software releases, in LHCb and Rec packages. The details are here.
The status of the implementation of the new Track Event Model is rather stable at this point; only small changes might still come up. The bulk of these packages are part of the official software releases, in LHCb and Rec packages. The details are here.
|
|