Leave a dotted key out of the runs export header

_leaves named a column for a key like `-1.0` that _dig can never read
back, so the export showed an always-empty column. Skip such a key.

backend/fluksio/api/routes/runs.py

@@ -441,9 +441,15 @@ def _leaves(value: Any, prefix: str = "") -> Iterator[tuple[str, Any]]:
     the numbers arrive inside `final_metrics` — and a record in one cell is
     not a column anybody can compare. Lists are left whole: a curve belongs in
     the long table, not in a cell of this one.
+
+    A key with a dot in its own name — a frequency written `-1.0` — is left
+    out, and whatever sits under it with it: `_dig` cannot read such a path
+    back, so naming it would promise a column that is always empty.
     """
     if isinstance(value, dict):
         for key, inner in value.items():
+            if "." in str(key):
+                continue
             yield from _leaves(inner, f"{prefix}.{key}" if prefix else str(key))
     else:
         yield prefix, value

backend/tests/api/routes/test_runs.py

@@ -1121,6 +1121,9 @@ def exported():
                         "note": "n/a",
                         "final_metrics": {"train_loss": acc * 2},
                         "test_metrics": {"known": {"perfect": 1.0}},
+                        # Keyed by a number written with its decimal point,
+                        # which no dotted path can address.
+                        "spectrum": {"-1.0": acc, "0.25": {"mean": acc}},
                     },
                     created_at=made + timedelta(seconds=index),
                 )
@@ -1199,6 +1202,8 @@ def test_an_exported_run_row_carries_every_recorded_input(
     assert rows[0]["metric.final_metrics.train_loss"] == 0.5
     assert rows[0]["metric.test_metrics.known.perfect"] == 1.0
     assert "metric.note" not in rows[0]
+    # A key holding the separator cannot be read back, so it is not a column.
+    assert not any(key.startswith("metric.spectrum") for key in rows[0])
 
     named = _lines(export(format="jsonl", metrics="test_metrics.known.perfect"))
     assert list(named[0])[-1] == "metric.test_metrics.known.perfect"

docs/code/api.md

@@ -134,7 +134,8 @@ the inputs that *vary* across the selection in `param.` columns (the sweep
 axis) unless `?params=` names them, and the run's numbers in `metric.`
 columns. Both are dotted paths into whatever a node returned, so
 `?metrics=final_metrics.train_loss` selects a field of a record and the
-defaults reach every number inside one. The run id is on every row of both, which is what makes an exported
+defaults reach every number inside one. A key with a dot in its own name, such
+as `-1.0`, cannot be addressed that way and gets no column. The run id is on every row of both, which is what makes an exported
 file a join back to the run rather than a loose table.
 
 `compare` answers in the same shape a chart widget draws, so three training