A day whose food was all unlabelled read "Fri, Sep 18 — No kind 300 g · 300 g
in total": the number twice, under a label with nothing to distinguish it from.
The guard was on the wrong quantity. It asked whether the *chart* was split,
when what decides this is how many kinds *that day* holds — and a chart split
across other days can still land on a day of one kind. Keying off the day's own
parts fixes the reported case and a second one nobody had hit yet, where a day
of a single named kind read "Dry 300 g · 300 g in total".
So: several kinds keep the total beside them, being what the bar's height shows
and what you would otherwise add up. One kind does not, because it already is
the total. And "No kind" alone drops its label, which was only ever there to
tell it apart from something else.
The new assertions were checked against the old guard, where five of them fail.
The original ones passed throughout, which is the point — they only ever
exercised a chart with one series, and this bug lives on the other axis.
Alexander Heldt
cd13efdbed