No `$ERA` support for `ExtractShapes` and `HistMapping`

[IzaakWN]

I naively tried

cb.cp().backgrounds().ExtractShapes(
  infname, '$BIN_$ERA/$PROCESS', '$BIN_$ERA/$PROCESS_$SYSTEMATIC')
cb.cp().signals().ExtractShapes(
  infname, '$BIN_$ERA/$PROCESS_M$MASS','$BIN_$ERA/$PROCESS_M$MASS_$SYSTEMATIC')

but got the error

$ ./harvestcards_shapes.py
Traceback (most recent call last):
  File "/afs/cern.ch/user/i/ineuteli/analysis/CMSSW_14_1_X_comb/src/datacard-tutorial/./harvestcards_shapes.py", line 49, in <module>
    cb.cp().backgrounds().ExtractShapes(
cppyy.gbl.std.runtime_error: void ch::CombineHarvester::ExtractShapes(const string& file, const string& rule, const string& syst_rule) =>
    runtime_error: 
*******************************************************************************
Context: Function ch::GetClonedTH1 at 
  src/CombineHarvester/CombineTools/src/TFileIO.cc:24
Problem: TH1 SR_$ERA/data_obs not found in input/xmm_inputs.root
*******************************************************************************
Please report issues at
  https://github.com/cms-analysis/CombineHarvester/issues
*******************************************************************************

Digging in the source code, it seems like the $ERA keyword is not supported in CombineHarvester::ExtractShapes:

CombineHarvester/CombineTools/src/CombineHarvester_Creation.cc

Lines 159 to 167 in cfff94b

	void CombineHarvester::ExtractShapes(std::string const& file,
	std::string const& rule,
	std::string const& syst_rule) {
	std::vector<HistMapping> mapping(1);
	mapping[0].process = "*";
	mapping[0].category = "*";
	mapping[0].file = std::make_shared<TFile>(file.c_str());
	mapping[0].pattern = rule;
	mapping[0].syst_pattern = syst_rule;

nor in CombineHarvester::LoadShapes:

CombineHarvester/CombineTools/src/CombineHarvester.cc

Lines 301 to 306 in cfff94b

	HistMapping mapping =
	ResolveMapping(entry->process(), entry->bin(), mappings);
	boost::replace_all(mapping.pattern, "$CHANNEL", entry->bin());
	boost::replace_all(mapping.pattern, "$BIN", entry->bin());
	boost::replace_all(mapping.pattern, "$PROCESS", entry->process());
	boost::replace_all(mapping.pattern, "$MASS", entry->mass());

nor in HistMapping:

CombineHarvester/CombineTools/interface/HistMapping.h

Lines 10 to 13 in cfff94b

	struct HistMapping {
	std::string process;
	std::string category;
	std::string pattern;

It this by design, or could we add it so one does not need to add an additional loop? If it's simple adding a couple of lines in the places above, I could open a PR.

[ajgilbert]

I think this would be fine to add. The original motivation was that we would replace only the special placeholders ($CHANNEL, $BIN, $PROCESS, $MASS) that are interpreted in text2workspace.py, but I don't see any harm in allowing $ERA or $ANALYSIS that exist only in CombineHarvester. So yes, a PR is welcome for this.