union — Return unions of input arrays with MATLAB-compatible ordering and optional index outputs.
union(A, B) returns distinct values (or rows) appearing in either input and supports MATLAB-compatible ordering modes and optional index outputs that map back to A and B.
Syntax
C = union(A, B)
C = union(A, B, option...)
[C, ia] = union(A, B)
[C, ia] = union(A, B, option...)
[C, ia, ib] = union(A, B)
[C, ia, ib] = union(A, B, option...)Inputs
A- Type
Any- Required
- Yes
- Description
- First input array.
B- Type
Any- Required
- Yes
- Description
- Second input array.
option- Type
StringScalar- Required
- Variadic
- Description
- Option tokens: 'rows'|'sorted'|'stable'.
Returns
C- Type
Any- Description
- Union values or rows.
ia- Type
NumericArray- Description
- Indices selecting contributions from A.
ib- Type
NumericArray- Description
- Indices selecting contributions from B.
Returned values from union depend on how many outputs the caller requests.
Errors
RunMat:union:LegacyOptionUnsupported- When
- Legacy compatibility options are requested.
- Message
union: the 'legacy' behaviour is not supported
RunMat:union:ConflictingOrderOptions- When
- Both 'sorted' and 'stable' options are provided.
- Message
union: cannot combine 'sorted' with 'stable'
RunMat:union:UnknownOption- When
- An unsupported option token is provided.
- Message
union: unrecognised option
Behavior
union(A, B) flattens the inputs column-major, removes duplicates, and returns the sorted union of their values.
[C, IA, IB] = union(A, B) also returns index vectors where IA points to the elements in A that contribute to C and IB refers to the elements in B that only appear in B.
union(A, B, 'stable') preserves the first-seen order: unique values from A appear first, followed by new values appearing in B.
union(A, B, 'rows') treats each row as a record. Inputs must share the same number of columns.
Character arrays and string arrays are fully supported in both element and row modes.
Complex values follow MATLAB's ordering rules (magnitude first, then real/imaginary parts).
Legacy compatibility flags ('legacy' or 'R2012a') are not supported in RunMat.
Types, placement and limitations
Element results are rows when both inputs are row vectors, and columns otherwise. Index outputs remain columns. Matching numeric classes are preserved. Combining double with single or one integer class preserves the nondouble class; unlike nondouble numeric classes are rejected. Logical inputs currently produce double values. Each NaN occurrence is distinct, including repeated NaNs in the same input. Sorted output places them last; stable output preserves their occurrence order.
Resident execution
These operations are eager fusion boundaries, not elementwise kernels or reductions, and do not broadcast. Provider availability does not imply a device-native sorting or set kernel: current providers may stage host work. An unsupported input can use exact host fallback before provider work starts. Errors after an attempt, including cancellation, are returned without repeating the operation on the host. Supported resident outputs return to their owning provider; predicates and all-indices cells have the host placement described above. Distributed execution is not supported.
Examples
Combining Two Numeric Vectors
A = [5 7 1];
B = [3 1 1];
[C, IA, IB] = union(A, B)Preserving Input Order with 'stable'
A = [5 7 1];
B = [3 2 4];
C = union(A, B, 'stable')Union of Matrix Rows
A = [1 2; 3 4; 1 2];
B = [3 4; 5 6];
[C, IA, IB] = union(A, B, 'rows')Expected output:
C =
1 2
3 4
5 6
IA =
1
2
IB =
2Working with Character Arrays
A = ['m','z'; 'm','a'];
B = ['a','x'; 'm','a'];
union(A, B)Expected output:
ans =
4x1 char array
a
m
x
zUnion on GPU Data
G = gpuArray([1 4 2 5]);
H = gpuArray([2 6]);
[C, IA, IB] = union(G, H)Preserve distinct NaNs and row orientation
A = [NaN NaN 1]; B = [1 NaN]; [C,IA,IB] = union(A,B); [S,SA,SB] = union(A,B,'stable');Using union with coding agents
Open a RunMat example with live inputs, then ask the agent to explain how union changes the result.
Run a small union example, explain the result, then change one input and compare the output.
FAQ
How are the index outputs defined?⌄
IA lists the positions in A that contribute to C. IB lists the positions in B for union values that do not already appear in A. Both use MATLAB's one-based indexing.
What happens if I request 'stable' ordering?⌄
The output keeps the first occurrence order: all unique values from A appear first, followed by new values from B in the order they are encountered.
Can I use 'rows' with string or character arrays?⌄
Yes. union accepts string arrays and character arrays with the 'rows' option, provided the inputs share the same number of columns.
How does union treat NaN values?⌄
Element results are rows when both inputs are row vectors, and columns otherwise. Index outputs remain columns. Matching numeric classes are preserved. Combining double with single or one integer class preserves the nondouble class; unlike nondouble numeric classes are rejected. Logical inputs currently produce double values. Each NaN occurrence is distinct, including repeated NaNs in the same input. Sorted output places them last; stable output preserves their occurrence order.
Does GPU execution change the behaviour?⌄
Element results are rows when both inputs are row vectors, and columns otherwise. Index outputs remain columns. Matching numeric classes are preserved. Combining double with single or one integer class preserves the nondouble class; unlike nondouble numeric classes are rejected. Logical inputs currently produce double values. Each NaN occurrence is distinct, including repeated NaNs in the same input. Sorted output places them last; stable output preserves their occurrence order. These operations are eager fusion boundaries, not elementwise kernels or reductions, and do not broadcast. Provider availability does not imply a device-native sorting or set kernel: current providers may stage host work. An unsupported input can use exact host fallback before provider work starts. Errors after an attempt, including cancellation, are returned without repeating the operation on the host. Supported resident outputs return to their owning provider; predicates and all-indices cells have the host placement described above. Distributed execution is not supported.
What if the inputs have different shapes?⌄
Element-wise unions accept any shapes; both inputs are flattened column-major. Row-wise unions require matching column counts and two-dimensional inputs.
Can I mix data types?⌄
No. Inputs must belong to the same class (numeric/logical, complex, char, or string). Mixing classes produces a descriptive error.
Are scalar outputs returned as vectors?⌄
Scalars follow MATLAB's behaviour: a 1×1 union result is represented as a scalar double.
Do trailing spaces matter for character data?⌄
Yes. Character arrays treat every character, including trailing spaces, as significant—matching MATLAB's behaviour.
Is the 'legacy' flag supported?⌄
No. RunMat only implements the modern MATLAB semantics. Passing 'legacy' or 'R2012a' raises an error.
Related Array functions
Sorting Sets
argsort · intersect · ismember · ismembertol · issorted · issortedrows · setdiff · setxor · sort · sortrows · unique
Accumulation
Shape
blkdiag · cat · circshift · diag · flip · fliplr · flipud · horzcat · ipermute · kron · permute · repelem · repmat · reshape · rot90 · shiftdim · squeeze · toeplitz · tril · triu · vertcat
Creation
colon · createArray · empty · eye · false · full · inf · int16.empty · int32.empty · int64.empty · int8.empty · linspace · logspace · magic · meshgrid · nan · ndgrid · nonzeros · ones · peaks · rand · randi · randn · randperm · range · sparse · spdiags · speye · spones · sprand · true · uint16.empty · uint32.empty · uint64.empty · uint8.empty · zeros
Combinatorics
Binning
Grouping
Source you can read
Every RunMat function's source is public. Read exactly how union is executed, line by line, in Rust.
- View the source for union in Rust on GitHub
- Learn how the RunMat runtime works
- Found a bug? Open an issue with a minimal reproduction.
About RunMat
RunMat is a runtime that executes MATLAB-syntax code fast on any GPU. The runtime is free for individuals, including freelancers; students and faculty at an academic institution, for teaching or research there; contributors to an open-source project, for work on that project; organizations with 250 or fewer people and US$10M or less in annual revenue; and anyone, for non-production use. See runmat.com/license.
- RunMat automatically optimizes your math for GPU execution on Apple, Nvidia, and AMD hardware. No code changes needed. Simulations that took hours now take minutes.
- Start running code in seconds. RunMat runs in the browser, on the desktop, or from the CLI. No license server, no IT ticket.