Zero-row datasets¶
A dataset can be present, carry every column the rule names, and hold no rows. This page is about what your rule does then.
It is already reported — do not report it again¶
An empty dataset is an error case, and every non-draft rule package already carries exactly one rule that reports it: the "domain table has no records" rule, scoped to every domain, counting rows and firing on zero.
⛔ Your rule should not also report it. If every rule flagged an empty dataset, one missing table would produce a finding from every rule in the package, burying the one finding that says what actually happened.
Write the check for the question it is about, and let the empty dataset be somebody else's finding.
What the engine does¶
A rule that must read a row has nothing to judge. It runs, finds nothing, and reports as executed with zero findings.
⚠ Not skipped. A skip means "a dataset or a provider was absent" — a different fact, read differently downstream. An empty table was present; it simply had nothing in it.
A rule that can decide without reading a row — one that asks about metadata, about Define-XML, about the library — fires exactly as it would on a populated dataset. The answer does not depend on rows, so the absence of rows changes nothing:
Check:
expression: 'var_label("DATA") != var_label("LIBRARY")'
A grouped rule sees zero rows, so it forms zero groups, so it reports nothing.
What this means when you author¶
You do not need a row-count guard. A check that reads rows is already inert on an empty
dataset; adding $records > 0 and … buys nothing and adds a binding to maintain.
Be deliberate about metadata rules. A rule that compares a label, checks a declared length or reads Define-XML will fire on an empty dataset, because it does not need rows to be right. That is normally correct — a mis-declared variable is mis-declared whether or not anyone populated it — but it is worth knowing that these are the rules an empty table does not silence.
Scenario coverage is where this gets pinned. If your rule's behaviour on an empty dataset is part of what it means, write a zero-row scenario for it. A rule with no such scenario has an untested answer for a case that happens in real submissions.
Next: back to Values and missing values, or on to Scenarios to pin any of this.