Collabora Office ROUNDSIG: Significant Digits and Magnitude
Round numbers by significant figures, distinguish decimal places and verify source-specific digit handling with small numerical examples. Scoped to the new Linux desktop source build.
Verification status: See the evidence and test scope described in this guide. How formula examples are checked
ROUNDSIG rounds a number to a chosen count of significant figures. It follows the magnitude of the value: five significant figures may mean two decimal places for one number and eight decimal places for a much smaller one.
Syntax and the help’s wording
=ROUNDSIG(Value; Digits)
Both arguments are required. The help’s opening definition and examples describe significant digits, although its argument description loosely calls Digits a number of decimal places. The evaluator resolves that ambiguity: it uses the base-10 magnitude of the absolute value to choose a rounding scale. Do not use ROUNDSIG as a fixed-decimal-place substitute for ROUND.
The documented input is a positive integer digit count. The exact source applies approximate floor to Digits and rejects the result if it is below 1. Thus a fractional count such as 2.9 follows a source-derived two-significant-figure path. Prefer explicit integer counts in ordinary worksheets rather than relying on coercion.
Step by step: distinguish significant figures from decimals
Enter A2 123.456789 and B2 5, then calculate:
=ROUNDSIG(A2;B2)
The independent expected value is 123.46. Its first three significant figures are the integer digits 1, 2 and 3, leaving two decimal positions. The literal equivalent is =ROUNDSIG(123.456789;5).
For =ROUNDSIG(0.000123456789;5), the expected value is 0.00012346. Leading zeros locate the decimal magnitude; they do not count as significant figures. Both examples appear in the vendor help and are independently reproduced here.
Understand the scale calculation
For a nonzero value x and digit count d, the checked evaluator computes the decimal scale exponent:
floor(log10(abs(x))) + 1 - d
It rescales x, rounds the rescaled number with its numerical rounding helper, then restores the scale. For 123.456789 and d=5, the exponent is -2, so the value is rounded at hundredths. For 123456789 and d=4, the exponent is 5, so the value is rounded at hundred-thousands.
The independent result of =ROUNDSIG(123456789;4) is 123500000, also shown by the help. A display such as 123.5E6 represents the same numeric value; it does not change the underlying rounding operation.
Signs, zero and a carry into the next magnitude
| Formula | Independent expected result | Reason |
|---|---|---|
=ROUNDSIG(-123.456;4) |
-123.5 | Magnitude chooses the scale; sign is preserved |
=ROUNDSIG(0;4) |
0 | Explicit zero branch in the evaluator |
=ROUNDSIG(9.996;3) |
10 | Rounding carries across the power-of-ten boundary |
=ROUNDSIG(9876;2.9) |
9900 | Source floors the count to 2 |
The numeric value 10 does not retain a distinct “10.0” precision annotation. If a report must show three significant figures, numeric formatting or explanatory metadata is a separate requirement. ROUNDSIG changes the value, not a general measurement-uncertainty record.
Advanced: keep one reporting precision across scales
Suppose a report combines a large count 123456789 and a small ratio 0.000123456789. Four significant figures give 123500000 and 0.0001235, respectively. Their required decimal places differ dramatically, which is why one fixed decimal-place count is unsuitable for this reporting goal.
Retain original measurements for subsequent calculations where possible and round a presentation result deliberately. Repeatedly rounding intermediate stages can change a later total or ratio. Applying the same significant-figure count also does not establish that the original measurements have equal uncertainty.
Errors and numerical boundaries
=ROUNDSIG(12.3;0) reaches the source’s internal IllegalArgument error path because the digit count is below 1. This is a source prediction, not an observed localized error display. The same validation occurs before the zero-value return, so a zero input does not make an invalid digit count valid.
The implementation uses floating-point logarithms, powers and a rounding helper. The examples avoid exact-half ambiguities, extreme exponents and overflow. No unlimited precision, safe arbitrary digit count or exhaustive tie behavior is inferred. Use ROUND when a fixed decimal position is the actual requirement, and keep formatting separate from numerical rounding.
Product scope, version and evidence
This guide concerns the new Collabora Office desktop for Linux, mapped by its Flathub manifest to source build coda-26.04.4.3-1, commit 81339d176148d0d5c81ff05d82540f58f8aaa117. It uses Collabora Office 26.04 ROUNDSIG help and the exact evaluator source, checked October 18, 2026. The vendor’s desktop release announcement and the Linux package manifest provide product context; the latter is a mutable publication, not an installed-binary attestation.
Examples are independent calculations and source traces. No Collabora runtime or calculation engine was executed, and no native screenshot or array-entry workflow is claimed. This narrow evidence does not certify Classic, Online, CODE, mobile, Windows or macOS behavior. The original Collabora introduction release is unknown; upstream release history is not substituted for it.
The help identifies ORG.LIBREOFFICE.ROUNDSIG and states that the function is outside OpenFormula 1.3. This is a format-compatibility caution, not a deprecation or retirement notice. File export, import and recalculation in another product were not tested.