A date in a group header is the reader's display string, not a date
When a view is grouped by a Date field, @group.fieldData in groupProps.headerFormatter is the already-formatted text in the reader's locale (05.10.2026 for a Czech reader), not a date value — so date functions are the wrong tool; reshape it with string functions only after checking the shape, and fall back to the raw text otherwise.
Ve view seskupeném podle data je @group.fieldData v groupProps.headerFormatter hotový text v locale čtenáře (u českého čtenáře 05.10.2026), ne datum – datumové funkce na něj nepatří; přeformátuj ho řetězcovými funkcemi až po kontrole tvaru a jinak nech původní text.
Symptom
A grouped "by meeting date" view shows group headers like Schůzka 05.10.2026 — readable, but not the house style (5. 10. 2026). The usual row-formatter recipe (getDate([$Field]) + '. ' + (getMonth([$Field]) + 1) + ...) expects a date value, which @group.fieldData is not.
Cause
Group header data comes from the rendered group, not from an item: @group.fieldData carries the field's display value as text, formatted for whoever is looking at the page (observed: 05.10.2026 for a Czech-locale reader). Readers with other regional settings should be expected to get a different shape — hence the shape check below rather than fixed positions alone.
Fix
Reformat only when the text has the shape you expect, and keep the raw value otherwise (s stands for toString(@group.fieldData)):
"txtContent": "=if(indexOf(s, '.') == 2 && lastIndexOf(s, '.') == 5, toString(Number(substring(s, 0, 2))) + '. ' + toString(Number(substring(s, 3, 5))) + '. ' + substring(s, 6, 10), s)"Number() drops the leading zeros (05 → 5). Do not test the length with length() — it counts array elements, not characters (details).
Notes
- For the group of items with no date, keep a fallback label, but do not rely on
== ''without testing it on your list: on rows an empty Date field is not''(details). @group.countis a number; pluralize with plain comparisons (== 1,> 4).