[rntuple] Read a column's value range as IEEE 754 doubles

src/rootfilespec/rntuple/schema.py

@@ -129,7 +129,9 @@ class FieldDescription(RecordFrame):
         - Bit:   Meaning
         - 0x01:  Repetitive field, i.e. for every entry `n` copies of the field are stored
         - 0x02:  Projected field
-        - 0x04:  Has ROOT type checksum as reported by TClass"""
+        - 0x04:  Has ROOT type checksum as reported by TClass
+        - 0x08:  The field is a collection that was stored using a SoA layout
+                 (no effect on the on-disk representation; adds no optional field)"""
     fFieldName: RString
     """The name of the field."""
     fTypeName: RString
@@ -178,13 +180,15 @@ class ColumnDescription(RecordFrame):
         - 0x02:  Column with a range of possible values"""
     fRepresentationIndex: Annotated[int, Fmt("<H")]
     """The index of the representation of the column in the list of representations for the field."""
-    # abbott TODO: verify the below are signed. make PR updating ROOT documentation if so (indicate signed bit in table)
     fFirstElementIndex: Annotated[int | None, OptionalField("<q", "fFlags", "&", 0x01)]
-    """The index of the first element in the column. Present only if flag 0x01 is set (deferred column)."""
-    fMinValue: Annotated[int | None, OptionalField("<q", "fFlags", "&", 0x02)]
-    """The minimum value of the column. Present only if flag 0x02 is set (column with range of values)."""
-    fMaxValue: Annotated[int | None, OptionalField("<q", "fFlags", "&", 0x02)]
-    """The maximum value of the column. Present only if flag 0x02 is set (column with range of values)."""
+    """The index of the first element in the column. Present only if flag 0x01 is set (deferred column).
+    Signed: a negative value means the column is deferred and suppressed."""
+    fMinValue: Annotated[float | None, OptionalField("<d", "fFlags", "&", 0x02)]
+    """The minimum value of the column, an IEEE 754 double.
+    Present only if flag 0x02 is set (column with range of values)."""
+    fMaxValue: Annotated[float | None, OptionalField("<d", "fFlags", "&", 0x02)]
+    """The maximum value of the column, an IEEE 754 double.
+    Present only if flag 0x02 is set (column with range of values)."""
 
 
 @serializable

tests/test_column_value_range.py

@@ -0,0 +1,34 @@
+from skhep_testdata import data_path  # type: ignore[import-not-found]
+
+from rootfilespec.bootstrap import ROOT3a3aRNTuple
+from rootfilespec.reader import open_path
+from rootfilespec.rntuple.RNTuple import RNTuple
+from rootfilespec.rntuple.schema import ColumnType
+
+
+def test_column_value_range_is_double():
+    """Issue #75: a column's value range is a pair of IEEE 754 doubles
+
+    The file's quantized columns (4 to 10, fields quant1 to quant32) all have
+    the range [-2.0, 3.0]. Read as int64, the same bytes are the bit patterns
+    -4611686018427387904 and 4613937818241073152.
+    """
+    filename = "test_float_types_rntuple_v1-0-0-0.root"
+    with open_path(data_path(filename)) as reader:
+        anchor = reader.fetch(reader.keylist()["ntuple"])
+        assert isinstance(anchor, ROOT3a3aRNTuple)
+        rntuple = RNTuple.from_anchor(anchor, reader.fetch.buffer)
+
+    columns = rntuple.schemaDescription.columnDescriptions
+    with_range = [i for i, c in enumerate(columns) if c.fFlags & 0x02]
+    assert with_range == list(range(4, 11))
+    for i in with_range:
+        column = columns[i]
+        assert column.fColumnType == ColumnType.kReal32Quant
+        assert column.fMinValue == -2.0
+        assert column.fMaxValue == 3.0
+        assert isinstance(column.fMinValue, float)
+    for i, column in enumerate(columns):
+        if i not in with_range:
+            assert column.fMinValue is None
+            assert column.fMaxValue is None