A single `ix:nonFraction` quietly carries three numbers in tension. The rendered text the reader sees. The canonical XBRL value the consumer parses. The declared accuracy that says how much rounding has been applied. Get the relationship between those three right and the fact is honest. Get it wrong and the file lies, often without any validator noticing.
Start with `format`. A transformation from the Inline XBRL Transformation Registry version 4 (`ixt:numdotdecimal`, `ixt:numcommadecimal`, `ixt:datedaymonthyearen`, and friends) converts the rendered string into a canonical numeric or date value. `ixt:numdotdecimal` reads "1,234.56" as one thousand two hundred thirty-four and fifty-six hundredths; `ixt:numcommadecimal` reads the same string as one and two hundred thirty-four thousand five hundred sixty hundred-thousandths. Pick the wrong one and the value silently flips by orders of magnitude.
Then `scale`. `scale` is an inline-only multiplier: the parsed text is multiplied by 10 to the power of `scale` to produce the canonical value. `scale="3"` on rendered "1,234" yields 1,234,000. This is how reports stated "in thousands" or "in millions" reconcile to whole-unit canonical values without the reader squinting.
Then `sign`. `sign="-"` on the inline tag negates the parsed value to produce the canonical form. It pairs with parentheses formatting in the host XHTML: "(1,234)" plus `sign="-"` yields a negative canonical value. The fact's underlying mathematical sign is what consumers see; presentation is how the reader sees the same thing.
Finally `decimals`. Positive numbers count digits after the decimal point. Negative numbers count rounding to powers of ten: `decimals="-3"` means rounded to thousands, `decimals="0"` means whole units, `decimals="-6"` means rounded to millions. SEC EFM 6.5.16 forbids `decimals="INF"` (it claims exactness that rounded values cannot honour); ESEF discourages it for the same reason. `precision` is the legacy alternative; SEC and SBR forbid it outright.
There is one identity you should burn into memory. Canonical value equals `transform(rendered_text) * 10^scale * (sign == "-" ? -1 : 1)`. If that calculation does not match what the reader sees on the page, the filing has a tagging defect, even when every validator says green.
decimals declares the accuracy in decimal places relative to the decimal point: positive numbers count digits after the point, negative numbers count rounding to tens, hundreds, thousands, and so on.
decimals='-3' means rounded to thousands. decimals='0' means whole units. decimals='INF' claims exact value and is forbidden by SEC EFM 6.5.16 and discouraged in ESEF when rendered text is shorter than INF claims.
scale on ix:nonFraction multiplies the parsed text by 10^scale to produce the canonical fact value. scale='3' on rendered '1,234' yields 1,234,000.
sign='-' on the inline tag negates the parsed value. It pairs with parentheses formatting in the host XHTML; it does not change the underlying fact's mathematical sign.
format names a transformation from the Transformation Registry (TRR 4) such as ixt:numdotdecimal that converts rendered text into a canonical numeric lexical value.
Audit identity: canonical_value = transform(rendered_text) * 10^scale * (sign == '-' ? -1 : 1). If that does not match what the reader sees, it is a defect.
<ix:nonFraction name="ifrs-full:ProfitLoss"
contextRef="c-Group-2025"
unitRef="u-eur"
decimals="-3"
scale="3"
sign="-"
format="ixt:numdotdecimal">(1,234)</ix:nonFraction>4 questions. Druk op 1 tot 4 om te kiezen, Enter om te bevestigen.
Door deze quiz af te ronden, markeer je deze sectie als voltooid.
Tip: kies met 1 tot en met 4 en bevestig met Enter.