blog: add formula, range-value and NOTNULL-constraint caveats to the special missings post

Verified on a real SAS estate:
- a physical NOT NULL (or primary key) constraint rejects a special missing,
  and getdata merges that constraint into a frontend NOTNULL rule - which
  passes a special missing, so the editor is more lenient than the constraint
- a special missing as a MINVAL rule value fails every cell; as a MAXVAL rule
  value it fails every real number
- a HARDFORMULA/SOFTFORMULA reading a special-missing cell returns #VALUE!
- PRX and the JS engine agree on the value; SAS pads the numeric-to-character
  conversion, so an anchored pattern re-used in SAS needs strip()

content/feed/sas-special-missings/index.md

@@ -50,6 +50,11 @@ These are Data Controller's own rules, configured per column in the `MPE_VALIDAT
 - `MAXVAL` - passes, for the same reason it is below any maximum
 - `HARDREGEX` - checked against the pattern like any other value; unlike blanks and the plain `.`, special missings are **not** exempt, so a numeric column that carries them needs a pattern which allows for a single letter
 - `SOFTREGEX` - the same check, but a failure is only a warning rather than a block, and it is ignored entirely if the column also has a `HARDREGEX` rule
+- `HARDFORMULA` / `SOFTFORMULA` - a formula that reads a special-missing cell does not compute; the grid's spreadsheet engine returns `#VALUE!` for that row, where the plain `.` contributes 0
+
+One mismatch to be aware of: the `NOTNULL` rule above is the editor's, and it is more forgiving than the SAS constraint it mirrors. A real NOT NULL - or primary key - constraint on the target table rejects a special missing, so on a column carrying one the editor will let the value through and the load will then fail on the constraint.
+
+Two other sharp edges. A range rule's *value* has to be a number: put a special missing in `MINVAL` and every cell in the column fails, put one in `MAXVAL` and every real number fails. And although the regex pattern is written in SAS PRX syntax, the check itself runs in the browser - SAS only parses the pattern (`PRXPARSE`) when the rule is saved - so the two engines can disagree on exotic patterns, and SAS pads a numeric-to-character conversion, so a pattern re-used in SAS needs `strip()` for an anchored match.
 
 `CASE` (`UPCASE` / `LOWCASE`) is a character rule and does not come into it: SAS hands the browser a special missing as an uppercase letter, so there is no case to enforce, and a `CASE` rule on a numeric column would reject the column's numbers rather than the missing.
 
@@ -86,10 +91,11 @@ We solved that in the open source SASjs Adapter, and it has been in Data Control
 
 And they behave like values in Data Controller's validation rules, which is usually what you want:
 
-- NOTNULL passes - it is not null
+- NOTNULL passes - it is not null (though a real SAS NOT NULL constraint is stricter, and does reject a special missing)
 - MINVAL fails - it sorts below every number; MAXVAL passes for the same reason
 - HARDREGEX checks it against the pattern, unlike blanks and plain .
 - SOFTREGEX warns instead of blocking (and is ignored if the column also has a HARDREGEX)
+- a HARDFORMULA/SOFTFORMULA that reads a special-missing cell returns #VALUE! rather than a number
 - CASE is a character rule, so it has no place on a numeric column
 
 So a value recorded as "not collected" cannot quietly satisfy a completeness or range rule.