feat: implement UNIQUE dynamic array function (HF-68) - #1708
Merged
Merged
Conversation
✅ Deploy Preview for hyperformula-dev-docs ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
Contributor
|
Task linked: HF-28 Modern dynamic array functions |
marcin-kordas-hoc
force-pushed
the
feature/HF-68-unique
branch
from
July 13, 2026 18:48
728efaf to
994b88e
Compare
7 tasks
Performance comparison of head (a4097a4) vs base (fbb4710) |
Add UNIQUE(array, [by_col], [exactly_once]) as a dynamic array function, mirroring the FILTER machinery for data-dependent result size. Deduplication preserves first-occurrence order and reuses ArithmeticHelper for equality (case-insensitive by default, honoring the caseSensitive config). Supports by_col (unique columns) and exactly_once (values occurring exactly once). Errors in the input range propagate; an empty exactly_once result yields #N/A. Includes i18n for all 17 language packs, docs (built-in-functions, known-limitations), a changelog entry, and the shared ADR. Source: https://app.clickup.com/t/86c89q1tq ADR: docs/adr/2026-07-13-sort-unique-array-functions.md Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
marcin-kordas-hoc
force-pushed
the
feature/HF-68-unique
branch
from
July 18, 2026 15:22
994b88e to
9eb508d
Compare
Without enableArrayArithmeticForArguments, an array expression passed as the first argument (e.g. =UNIQUE(A1:A3*2)) is coerced as a scalar and only the first cell is computed. Matches FILTER/VSTACK/HSTACK/SORT. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…s (HF-68) The compared vectors are always rows (or columns, after transpose) of the same rectangular range, so they never differ in length — the guard was dead code. Removes the last uncovered line in UniquePlugin. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Per review on #1707: keep the ADR but relocate it out of the engine repo into hyperformula-tests/dev_docs, since a per-function implementation ADR belongs with the test suite rather than the shipped docs dir. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Collaborator
Author
|
Moved the shared SORT/UNIQUE ADR out of the engine repo into |
Reword the known-limitations guide and the plugin JSDoc to describe HF's own case-insensitive comparison instead of framing it as matching Excel, keeping published surfaces free of the comparison claim. Internal code comments referencing Excel as the behavioural oracle are kept. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
sequba
pushed a commit
that referenced
this pull request
Jul 23, 2026
## What & why Implements **SORT** (`HF-69`, child of HF-28 "Modern dynamic array functions", sibling of the shipped SEQUENCE and of VSTACK/HSTACK). Adds `SORT(array, [sort_index], [sort_order], [by_col])` as a dynamic array function. Tests: handsontable/hyperformula-tests#24 (paired). Sibling PR: #1708 (UNIQUE / HF-68). ADR: `docs/adr/2026-07-13-sort-unique-array-functions.md`. ## Behavior - Returns an array the **same shape** as the input. - `sort_index` (default 1): 1-based index into the sort dimension. - `sort_order`: `1` ascending (default) or `-1` descending. - `by_col`: `FALSE` (default) reorders rows; `TRUE` reorders columns. - Ordering reuses `ArithmeticHelper` (mixed types: numbers < text < logical; empties; locale collation via `caseSensitive`/`accentSensitive`) and is **stable** — ties keep input order. ## Design Mirrors the SEQUENCE/FILTER machinery: `sizeOfResultArrayMethod` + `vectorizationForbidden: true`, runtime via `runFunction` returning `SimpleRangeValue`/`CellError`, parse-time size method returning a **fresh** `ArraySize` (the input's `isRef` flag is dropped — a ref-flagged size is treated as scalar and would collapse the spill). ## Notes — divergences from Excel (surfaced here + inline + in tests) - **`sort_order` is strictly `{1, -1}`**; any other value → `#VALUE!`. Excel documents only `{1,-1}`; the reported "`sort_order=0` does not error" quirk is undocumented and **could not be re-verified against live Excel in this environment**, so the strict documented contract was chosen (see ADR `dec_2`, `con_1`). Flagged for live-Excel/Kuba confirmation. - **Multi-key array-constant `sort_index`** (e.g. `{1,2}`) is **not supported in v1** (documented in `known-limitations.md`; ADR `dec_6`). - In-range errors propagate (first error found; ADR `dec_7`). ## Error-type map `sort_order ∉ {1,-1}` → `#VALUE!` (BadMode) · `sort_index < 1` → `#VALUE!` (LessThanOne) · `sort_index >` dimension → `#VALUE!` (ValueLarge) · in-range error → propagate · wrong arity → `#N/A`. ## Definition of Done - [x] Production code (`SortPlugin.ts`, registered via `plugin/index.ts`) - [x] i18n — all 17 language packs (authoritative MS Functions Translator names; enUS inherits enGB) - [x] Tests (paired tests PR) — across the standard array-function groups, dual-env safe - [x] Docs — `built-in-functions.md`, `known-limitations.md` - [x] JSDoc on all methods - [x] CHANGELOG entry - [x] ADR with audit-verified citations Source: https://app.clickup.com/t/86c89q1tt <!-- CURSOR_SUMMARY --> --- > [!NOTE] > **Low Risk** > Self-contained new array function following existing SEQUENCE/FILTER patterns; no changes to auth, persistence, or core recalculation beyond registering one plugin. > > **Overview** > Adds the **SORT** dynamic array function: `SORT(Array, [SortIndex], [SortOrder], [ByCol])` returns the input range reordered by row (default) or column, same dimensions as the source. > > Implementation lives in new `SortPlugin.ts`, wired like other array functions (`sizeOfResultArrayMethod`, `vectorizationForbidden`, spill size copied from input without propagating `isRef`). Sort keys use `ArithmeticHelper` (with empty cells forced last); invalid `sort_order` (not `1` or `-1`), bad `sort_index`, in-range errors, and empty ranges get the documented `#VALUE!` / `#N/A` / error propagation behavior. > > Docs and changelog are updated; **known-limitations** documents single-key only, strict sort order, and HF comparison rules. **SORT** is added to all 17 language packs. > > <sup>Reviewed by [Cursor Bugbot](https://cursor.com/bugbot) for commit f081be5. Bugbot is set up for automated code reviews on this repo. Configure [here](https://www.cursor.com/dashboard/bugbot).</sup> <!-- /CURSOR_SUMMARY --> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
sequba
approved these changes
Jul 23, 2026
![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 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit a4097a4. Configure here.
| } | ||
|
|
||
| const result = byCol ? UniquePlugin.transpose(kept) : kept | ||
| return SimpleRangeValue.onlyValues(result) |
There was a problem hiding this comment.
Missing empty-range crash guard
Medium Severity
UNIQUE can still produce a zero-width result when the input has height but no width (for example after addRows on an empty sheet, then a whole-row reference). That path builds [[]] and later hits ArrayValue, which rejects non-positive dimensions and throws. Sibling SORT already guards this case before building a range value.
Reviewed by Cursor Bugbot for commit a4097a4. Configure here.
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## develop #1708 +/- ##
===========================================
- Coverage 97.20% 96.86% -0.35%
===========================================
Files 177 178 +1
Lines 15546 15611 +65
Branches 3444 3460 +16
===========================================
+ Hits 15112 15122 +10
- Misses 434 489 +55
🚀 New features to boost your workflow:
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What & why
Implements UNIQUE (
HF-68, child of HF-28 "Modern dynamic array functions", sibling of the shipped SEQUENCE and of VSTACK/HSTACK). AddsUNIQUE(array, [by_col], [exactly_once])as a dynamic array function.Tests: handsontable/hyperformula-tests#25 (paired).
Sibling PR: #1707 (SORT / HF-69).
ADR:
docs/adr/2026-07-13-sort-unique-array-functions.md.Behavior
by_colisTRUE) of the input, preserving first-occurrence order.by_col:FALSE(default) compares rows;TRUEcompares columns.exactly_once:TRUEreturns only rows/columns occurring exactly once;FALSE(default) returns all distinct.ArithmeticHelper.eq→ case-insensitive by default (honorscaseSensitive), matching Excel's UNIQUE.Design
Mirrors the FILTER machinery for dynamic-size results:
sizeOfResultArrayMethod+vectorizationForbidden: true, runtime viarunFunction, parse-time size method returning a freshArraySize(drops the input'sisRefflag). Deduplication is O(n²) in the number of vectors because locale-aware equality is not trivially hashable — noted in code; acceptable for v1.Notes — divergences from Excel (surfaced here + inline + in tests)
exactly_oncewhen nothing occurs exactly once) →#N/A. Excel returns#CALC!, which HyperFormula has no type for; mirrors FILTER's empty-result mapping (ADRdec_8).con_1).dec_7).Definition of Done
UniquePlugin.ts, registered viaplugin/index.ts)built-in-functions.md,known-limitations.mdSource: https://app.clickup.com/t/86c89q1tq
Note
Low Risk
Additive array function behind existing dynamic-array machinery; no changes to auth, persistence, or core evaluation paths beyond new plugin registration.
Overview
Adds the Excel-style
UNIQUE(array, [ByCol], [ExactlyOnce])dynamic array function so formulas can return distinct rows or columns with first-occurrence order preserved.UniquePluginimplements deduplication viaArithmeticHelper.eq(honorscaseSensitive/accentSensitive), optional column-wise mode and “exactly once” filtering, propagates the first in-range error, and returns#N/AwhenExactlyOncewould yield an empty result (aligned with FILTER). Spill sizing follows FILTER: parse-time upper bound from input dimensions,vectorizationForbidden: true, and a freshArraySizesoisRefis not carried through.Also registers the plugin, adds
UNIQUEto all language packs, documents the function and known limitations, and records the change in the changelog.Reviewed by Cursor Bugbot for commit a4097a4. Bugbot is set up for automated code reviews on this repo. Configure here.