Repository navigation
Add PERCENTRANK.INC, MODE.MULT, FREQUENCY and PROB functions (HF-421, HF-424, HF-425, HF-428) - #1818
marcin-kordas-hoc wants to merge 8 commits into
Conversation
…ions Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Deploying with
|
| Status | Name | Latest Commit | Preview URL | Updated (UTC) |
|---|---|---|---|---|
| ✅ Deployment successful! View logs |
hyperformula-docs | 9487276 | Commit Preview URL Branch Preview URL |
Oct 25 2026, 12:48 PM |
Performance comparison of head (9487276) vs base (d143456) |
…centrank-inc-modemult-frequency-prob # Conflicts: # CHANGELOG.md
…ust the stack Spreading every value into Math.max and Array.push threw a RangeError for about 200000 values. Also documents that MODE.MULT and FREQUENCY reserve their result area before evaluation, and restores the spacing of an unrelated error message. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes using default effort and found 3 potential issues.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 2d8f790. Configure here.
| } | ||
| } | ||
| return numbers | ||
| } |
There was a problem hiding this comment.
MODE.MULT skips scalar coercible values
Medium Severity
MODE.MULT collects numbers through numbersOfArguments, which applies manyToExactNumbers to scalars and therefore drops booleans and numeric text instead of coercing them. Excel counts direct TRUE/FALSE and numeric strings and returns #VALUE! for other text. coerceNumbersExactRanges already implements that split.
Additional Locations (1)
Reviewed by Cursor Bugbot for commit 2d8f790. Configure here.
| if (numericBins instanceof CellError) { | ||
| return numericBins | ||
| } | ||
| const bins = numericBins.length === 0 ? [-Infinity] : numericBins |
There was a problem hiding this comment.
FREQUENCY fabricates a dummy empty bin
Medium Severity
When bins_array has no numbers, FREQUENCY inserts a -Infinity bin and returns a leading 0 plus the data count. Excel returns only the count of data_array. An empty numericBins list already sends every value to the overflow row, so the dummy bin changes both the shape and the values.
Reviewed by Cursor Bugbot for commit 2d8f790. Configure here.
| } | ||
| vals.sort((a, b) => a - b) | ||
| if (vals.length === 0 || x < vals[0] || x > vals[vals.length - 1]) { | ||
| return new CellError(ErrorType.NA, ErrorMessage.ValueNotFound) |
There was a problem hiding this comment.
Empty PERCENTRANK.INC returns wrong error
Low Severity
PERCENTRANK.INC returns #N/A when array contains no numbers, because the empty case is grouped with “x is outside the data”. Excel returns #NUM! for an empty array and #N/A only when x lies outside a non-empty set. getSortedValues already surfaces #NUM! for that empty case.
Reviewed by Cursor Bugbot for commit 2d8f790. Configure here.
…ificant digits The 1e-9 tolerance accepted sums that are off by up to 1e-9, which Excel rejects with #NUM!. Rounding the sum to 15 significant digits still accepts floating-point sums such as 0.1 added ten times. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…s with a compensated sum Rounding the plain sum to 15 significant digits accepted 0.9999999999999995, which Excel rejects. The sum is now exact to double-double precision (0.1 added ten times is 1), must be at least 1, and must be 1 when rounded to 15 significant digits. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## develop #1818 +/- ##
===========================================
+ Coverage 97.28% 97.32% +0.04%
===========================================
Files 207 207
Lines 16218 16323 +105
Branches 3556 3610 +54
===========================================
+ Hits 15777 15887 +110
+ Misses 433 428 -5
Partials 8 8
🚀 New features to boost your workflow:
|
Summary
Adds four statistical functions:
PERCENTRANK.INC(with itsPERCENTRANKalias),MODE.MULT,FREQUENCYandPROB(HF-421, HF-424, HF-425, HF-428).PERCENTRANK.INC(array, x, [significance])returns the rank ofxas a fraction from 0 to 1, interpolating between neighbours, truncated tosignificancedigits (3 by default).xoutside the data is#N/A, andsignificanceis truncated to an integer and must be at least 1 (otherwise#NUM!).MODE.MULT(number1, ...)returns a column with every most frequent value, in order of first occurrence, and#N/Awhen no value repeats.FREQUENCY(data_array, bins_array)returns a column with the count of values up to each bin and one more row for the values above the last bin. Text and blank cells are ignored.PROB(x_range, prob_range, lower_limit, [upper_limit])returns the probability that a value lies between the limits. The probabilities must be greater than 0, and their sum, added without rounding error, must be at least 1 and equal 1 when rounded to 15 significant digits (otherwise#NUM!): 0.1 added ten times is accepted, sums of 0.9999999999999995 and 1.000000000000006 are rejected. Ranges of different sizes give#N/A.PercentilePluginandStatisticalAggregationPlugin, with metadata, translations, the license capability table entries, a line indocs/guide/known-limitations.mdand a CHANGELOG entry.MODE.MULTandFREQUENCYreserve their result area before evaluation: half as many rows as values passed toMODE.MULT(rounded down, at least one), and one row per cell ofbins_arrayplus one forFREQUENCY. Rows the result does not fill stay empty, and a value in any reserved cell gives#SPILL!. This is the line added to the known limitations.Behaviour check against Excel
122 formulas were evaluated in Excel Online (MS Graph live session) and in HyperFormula: 118 identical, 0 different, 4 not evaluable (calls with a wrong number of arguments, which Excel rejects when the formula is entered). They cover typed numbers (percentage, date and currency cells), booleans, text and numeric text, blank cells, errors, ranges and scalars, ties, bin ordering, limit ordering and the
significanceargument. Array results are read throughINDEX,ROWS,COLUMNSandSUM, because they spill.Later runs measured 26 more formulas in Excel.
MODE.MULT(1,1,2,3,4,5)andFREQUENCY({1,2,3},{"a";2;"b"})spill only the rows they fill (1 and 2), which is the reserved-area difference above.MODE.MULTwith direct numeric text ("7","7") is#N/A, as in HyperFormula. ForPROB, 19 probability sums were measured: Excel rejects every measured sum below 1 (the closest is 0.9999999999999995) and every one from 1.000000000000006 up, and accepts the sums from 1 to 1.000000000000004, as well as 0.1 added ten times, 0.2 added five times, {0.1, 0.2, 0.7}, {0.2, 0.3, 0.5} and {0.3, 0.3, 0.4}.PROBnow gives the same result for all 19 (it used to accept sums off by up to 1e-9).Translations
Each localized name was checked on Microsoft's page for the function in that language (
support.microsoft.com/<locale>/excel/functions/<function>): the name used here is in the page's title or heading and in its syntax line, with one exception. The Dutch page namesFREQUENCYINTERVAL, which HyperFormula's ownINTERVALfunction already uses innlNL, so the English name is kept there. On the Swedish pages ofPERCENTRANK.INCandPERCENTRANK, one of title and heading shows the English name, while the other and the syntax showPROCENTRANG.INKandPROCENTRANG. The Czech pages keep the English names forPERCENTRANK.INC,PERCENTRANK,MODE.MULTandPROB, and so doescsCZ.enGBandidIDkeep the English names, which their Microsoft pages also use.Tests
Paired tests branch:
feat/hf-stats2-percentrank-inc-modemult-frequency-probin the tests repository (185 tests, includingMODE.MULTover 300000 values, 18 measured probability sums and one German-name test per function). All 185 fail againstdevelop. Full local run on this head: 524 suites, 6927 passed, 3 skipped.🤖 Generated with Claude Code
Note
Low Risk
Additive statistical builtins with localized names and spill sizing; no changes to auth, persistence, or core evaluation paths beyond new function handlers.
Overview
Adds four Excel-style statistical functions:
PERCENTRANK.INC(aliasPERCENTRANK) inPercentilePlugin, andMODE.MULT,FREQUENCY, andPROBinStatisticalAggregationPlugin.PERCENTRANK.INCranks a value within a range as a 0–1 fraction with inclusive interpolation, truncates to optionalsignificancedigits, and returns#N/A/#NUM!for out-of-range data or invalid significance.MODE.MULTspills a column of all tied modes (or#N/Aif nothing repeats).FREQUENCYreturns bin counts inbins_arrayorder plus an overflow row.PROBsums probabilities over inclusive limits using compensated double-double summation and Excel-like 15-digit “sum equals 1” validation, with a newProbabilityBoundserror message.MODE.MULTandFREQUENCYregister parse-timesizeOfResultArrayMethodhandlers so spill areas are reserved up front (documented in known limitations). Metadata, license capability lists, CHANGELOG, and locale translations are updated across supported languages.Reviewed by Cursor Bugbot for commit 9487276. Bugbot is set up for automated code reviews on this repo. Configure here.