Nest PF candidates inside jets (only reco so far)

coffea/nanoevents/methods/nanoaod.py

@@ -298,6 +298,12 @@ def matched_muons(self):
     def matched_gen(self):
         return self._events().GenJet._apply_global_index(self.genJetIdxG)
 
+    @property
+    def constituents(self):
+        if "pFCandsIdxG" not in self.fields:
+            raise RuntimeError("PF candidates are only available for PFNano")
+        return self._events().JetPFCands._apply_global_index(self.pFCandsIdxG)
+
 
 _set_repr_name("Jet")
 
@@ -336,6 +342,12 @@ def subjets(self):
     def matched_gen(self):
         return self._events().GenJetAK8._apply_global_index(self.genJetAK8IdxG)
 
+    @property
+    def constituents(self):
+        if "pFCandsIdxG" not in self.fields:
+            raise RuntimeError("PF candidates are only available for PFNano")
+        return self._events().FatJetPFCands._apply_global_index(self.pFCandsIdxG)
+
 
 _set_repr_name("FatJet")
 

coffea/nanoevents/schemas/nanoaod.py

@@ -138,6 +138,13 @@ class NanoAODSchema(BaseSchema):
         "Jet_electronIdxG": ["Jet_electronIdx1G", "Jet_electronIdx2G"],
     }
     """Nested collections, where nesting is accomplished by a fixed-length set of indexers"""
+    nested_index_items = {
+        "Jet_pFCandsIdxG": ("Jet_nConstituents", "JetPFCands"),
+        "FatJet_pFCandsIdxG": ("FatJet_nConstituents", "FatJetPFCands"),
+        "GenJet_pFCandsIdxG": ("GenJet_nConstituents", "GenJetCands"),
+        "GenFatJet_pFCandsIdxG": ("GenJetAK8_nConstituents", "GenFatJetCands"),
+    }
+    """Nested collections, where nesting is accomplished by assuming the target can be unflattened according to a source counts"""
     special_items = {
         "GenPart_distinctParentIdxG": (
             transforms.distinctParent_form,
@@ -207,6 +214,13 @@ def _build_collections(self, branch_forms):
                     [branch_forms[idx] for idx in indexers]
                 )
 
+        # Create nested indexer from n* counts arrays
+        for name, (local_counts, target) in self.nested_index_items.items():
+            if local_counts in branch_forms and "o" + target in branch_forms:
+                branch_forms[name] = transforms.counts2nestedindex_form(
+                    branch_forms[local_counts], branch_forms["o" + target]
+                )
+
         # Create any special arrays
         for name, (fcn, args) in self.special_items.items():
             if all(k in branch_forms for k in args):

coffea/nanoevents/transforms.py

@@ -125,6 +125,43 @@ def local2global(stack):
     stack.append(out)
 
 
+def counts2nestedindex_form(local_counts, target_offsets):
+    if not local_counts["class"].startswith("ListOffset"):
+        raise RuntimeError
+    if not target_offsets["class"] == "NumpyArray":
+        raise RuntimeError
+    form = {
+        "class": "ListOffsetArray64",
+        "offsets": "i64",
+        "content": copy.deepcopy(local_counts),
+    }
+    form["content"]["content"]["itemsize"] = 8
+    form["content"]["content"]["primitive"] = "int64"
+    form["content"]["content"]["parameters"] = {}
+    key = concat(
+        local_counts["form_key"], target_offsets["form_key"], "!counts2nestedindex"
+    )
+    form["form_key"] = local_counts["form_key"]
+    form["content"]["form_key"] = key
+    form["content"]["content"]["form_key"] = concat(key, "!content")
+    return form
+
+
+def counts2nestedindex(stack):
+    """Turn jagged local counts into doubly-kagged global index into a target
+
+    Signature: local_counts,target_offsets,!counts2nestedindex
+    Outputs a jagged array with same axis-0 shape as counts axis-1
+    """
+    target_offsets = stack.pop()
+    local_counts = stack.pop()
+    out = awkward.unflatten(
+        numpy.arange(target_offsets[-1], dtype=numpy.int64),
+        awkward.flatten(local_counts),
+    )
+    stack.append(out)
+
+
 @numba.njit
 def _distinctParent_kernel(allpart_parent, allpart_pdg):
     out = numpy.empty(len(allpart_pdg), dtype=numpy.int64)