setdiff — Return values present in the first input but absent from the second, with MATLAB-compatible ordering modes.
setdiff(A, B) returns unique values (or rows) that appear in A but not in B. It supports MATLAB-compatible 'sorted' and 'stable' ordering modes plus row-wise set differences.
Syntax
C = setdiff(A, B)
C = setdiff(A, B, option...)
[C, ia] = setdiff(A, B)
[C, ia] = setdiff(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
- Values that appear in A but not in B.
ia- Type
NumericArray- Description
- Indices selecting retained values/rows from A.
Returned values from setdiff depend on how many outputs the caller requests.
Errors
RunMat:setdiff:LegacyOptionUnsupported- When
- Legacy compatibility options are requested.
- Message
setdiff: the 'legacy' behaviour is not supported
RunMat:setdiff:ConflictingOrderOptions- When
- Both 'sorted' and 'stable' options are provided.
- Message
setdiff: cannot combine 'sorted' with 'stable'
RunMat:setdiff:UnknownOption- When
- An unsupported option token is provided.
- Message
setdiff: unrecognised option
Behavior
setdiff(A, B) flattens inputs column-major, removes duplicates, subtracts the values of B from A, and returns the remaining elements sorted ascending by default.
[C, IA] = setdiff(A, B) also returns indices so that C = A(IA).
setdiff(A, B, 'stable') preserves the first appearance order from A instead of sorting.
setdiff(A, B, 'rows') treats each row as an element. Inputs must share the same number of columns.
Character arrays, string arrays, logical arrays, numeric types, and complex values are all supported.
Legacy flags ('legacy', 'R2012a') are not supported; RunMat does not implement legacy flag semantics.
Types, placement and limitations
Element results follow A: a row-vector A gives a row result, otherwise a column. IA remains a column. Numeric and logical outputs retain A's class. A is never narrowed through the single comparison domain; supported mixed-class controls convert the B comparison operand instead. Unlike nondouble numeric classes are rejected. NaNs never match B or each other. Every NaN occurrence in A remains in C, including complex values or rows containing a NaN component.
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.
RunMat exact mixed-integer comparison
With double A and native integer B, RunMat compares their original numeric values exactly, including int64 and uint64 values beyond the exact double range. B is not rounded through double storage. C still contains selected original A values in double class; IA is a double column. This is a RunMat exactness policy for this mixed form, separate from output-class conversion.
The route uses indexed native comparisons for elements or rows and preserves the ordinary sorted/stable and distinct-NaN rules. Managed finite host allocation is not supported for this route; those calls fail before provider work or gathering because complete selector scratch and retained-output ownership is not available.
Examples
Finding values exclusive to the first numeric vector
A = [5 7 5 1];
B = [7 1 3];
[C, IA] = setdiff(A, B)Expected output:
C =
5
IA =
1Preserving input order with 'stable'
A = [4 2 4 1 3];
B = [3 4 5 1];
[C, IA] = setdiff(A, B, 'stable')Expected output:
C =
2
IA =
2Working with rows of numeric matrices
A = [1 2; 3 4; 1 2];
B = [3 4; 5 6];
[C, IA] = setdiff(A, B, 'rows')Expected output:
C =
1 2
IA =
1Computing set difference for character data
A = ['m','z'; 'm','a'];
B = ['a','x'; 'm','a'];
[C, IA] = setdiff(A, B)Expected output:
C = 'z'; IA = 3Subtracting string arrays by row
A = ["alpha" "beta"; "gamma" "beta"];
B = ["gamma" "beta"; "delta" "beta"];
[C, IA] = setdiff(A, B, 'rows', 'stable')Expected output:
C =
1x2 string array
"alpha" "beta"
IA =
1Using setdiff with GPU arrays
G = gpuArray([10 4 6 4]);
H = gpuArray([6 4 2]);
C = setdiff(G, H)Expected output:
C =
10Retain A precision, row orientation and every NaN
A = [1+eps(1) 2 NaN NaN]; B = single([2 NaN]); [C,IA] = setdiff(A,B); L = setdiff(logical([1 0 1]),false);Compare a wide integer without rounding it into A
A = [9007199254740992 1.5 NaN]; B = [0x0020000000000001u64 0u64]; [C,IA] = setdiff(A,B,'stable');Using setdiff with coding agents
Open a RunMat example with live inputs, then ask the agent to explain how setdiff changes the result.
Run a small setdiff example, explain the result, then change one input and compare the output.
FAQ
What ordering does setdiff use by default?⌄
Results are sorted ascending. Specify 'stable' to preserve the first appearance order from the first input.
How are the index outputs defined?⌄
IA points to the positions in A that correspond to each element (or row) returned in C, using MATLAB's one-based indexing.
Can I combine 'rows' with 'stable'?⌄
Yes. 'rows' can be paired with either 'sorted' (default) or 'stable'. Other option combinations that conflict (e.g. 'sorted' with 'stable') are rejected.
Does setdiff remove NaN values from A when they exist in B?⌄
Element results follow A: a row-vector A gives a row result, otherwise a column. IA remains a column. Numeric and logical outputs retain A's class. A is never narrowed through the single comparison domain; supported mixed-class controls convert the B comparison operand instead. Unlike nondouble numeric classes are rejected. NaNs never match B or each other. Every NaN occurrence in A remains in C, including complex values or rows containing a NaN component.
Are complex numbers supported?⌄
Absolutely. Complex values use MATLAB's ordering rules (magnitude, then real part, then imaginary part) for the sorted output.
Does GPU execution change the results?⌄
Element results follow A: a row-vector A gives a row result, otherwise a column. IA remains a column. Numeric and logical outputs retain A's class. A is never narrowed through the single comparison domain; supported mixed-class controls convert the B comparison operand instead. Unlike nondouble numeric classes are rejected. NaNs never match B or each other. Every NaN occurrence in A remains in C, including complex values or rows containing a NaN component. 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 happens if the inputs have different classes?⌄
Supported numeric inputs retain A's class, including when double is combined with single or one integer class. Unlike nondouble numeric classes are rejected.
Can I request 'legacy' behaviour?⌄
No. RunMat does not implement legacy flag semantics. Passing 'legacy' or 'R2012a' results in an error.
Related Array functions
Sorting Sets
argsort · intersect · ismember · ismembertol · issorted · issortedrows · setxor · sort · sortrows · union · 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 setdiff is executed, line by line, in Rust.
- View the source for setdiff 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.