GSoC 2022: Multiweight integration

[kfan326]

Context:
GSoc 2022 - Multiweight integration

Description of the Change:

This update extends the Cutflow and Histogram classes to hold data for n-weights and outputs the multidimensional histogram and cutflow data to SQLite3 format.

This draft is currently fully functional for Both Cutflows and histograms, currently still working on building an interface for SQlite3 output format.

Benefits:

This implementation will output Cutflow and Histogram data for all weights in a single SQLite3 file which is long term stable(updates in software version will not change accessibility) and easily transferable. SQlite3 supports the core query commands of the SQL standard and users will be able to process the result data however they wish, the database file can be accessed independently of MADAnalysis5 and plotted in Excel or whichever plotting package or the user's choosing.

Possible Drawbacks:

Currently still working on an interface to detect if SQLite3 is available on the system and if not, give user the option to install it. MacOS includes SQLite3 by default, this issue will only apply to Linux users.

Existing analysis files need to update the SampleAnalyzer.cpp file:

in the Execute function: user needs to pass in the weigh map as shown in the example below. the WeightCollection object supports *, /, +, - operators.

Example: user wants to multiply all weights by a double, say 1.2345, the user can do so by the following.

EvMultiweight*=1.2345;

MAdouble64 EvWeight;
WeightCollection EvMultiweight;  // create a WeightCollection object to hold multi weights
if(Configuration().IsNoEventWeight()) {
	  EvWeight=1.;
	  EvMultiweight=1.;
} else if(event.mc()->weight()!=0.) {
	  EvWeight=event.mc()->weight();
	  EvMultiweight=event.mc()->multiweights();
} else { return false;}
 
Manager()->InitializeForNewEvent(EvWeight, EvMultiweight.GetWeights()); //pass in weight map object in addition to single weight. 

The current implementation will run both the single and multi weight simultaneously and output independently (Existing SAF format is preserved and untouched, multi weight will output Outflow and Histogram SQLite3 files in their respective directories). The plan is to deprecate the single weight implementation once numerical validation is complete, the current implementation is a little cumbersome but is fully functional and easier to debug should their be discrepancies between the existing and new implementation.

Related GitHub Issues:

[jackaraz]

Here are some comments that need resolving.

[jackaraz]

This file should not be here!

[jackaraz]

Database Manager is not a good naming here indicates extreme general structure. Please create two dedicated classes for cutflow and histogramming. Current structure does not look expandable i.e. we might want to use sqlite for completely different thing than just writing histograms and curflows so dedicated code structures would be ideal. Additionally this piece of code needs bit of documentation so please add a README.md file here explaining how to read curflow and histo file generated by this file through python.

[jackaraz]

please also add
if importlib.util.find_spec("sqlite3") is None: ...
since one might have weird specification with cpp headers defined but python files missing etc.

[jackaraz]

This is not backwards compatible; it breaks all the codes developed in the past! Please split this into two as follows;

/// Initializing the new event with a single weight
void InitializeForNewEvent(const MAfloat64 &weight);

/// Initializing the new event with multi-weight
void InitializeForNewEvent(const std::map<MAuint32, MAfloat64> &multiweight);

This way, we won't need to modify old implementations individually. You can test old implementations by using the install PADForSFS command, which will download a set of implementation codes in tools/PADForSFS/Build/SampleAnalyzer/User/Analyzer, and you can see which functions can not be changed. All those codes have to work.