StringConcat¶

StringConcat - 20¶

Version¶

name: StringConcat (GitHub)
domain: main
since_version: 20
function: False
support_level: SupportType.COMMON
shape inference: True

This version of the operator has been available since version 20.

Summary¶

StringConcat concatenates string tensors elementwise (with NumPy-style broadcasting support)

Inputs¶

X (heterogeneous) - T:

Tensor to prepend in concatenation
Y (heterogeneous) - T:

Tensor to append in concatenation

Outputs¶

Z (heterogeneous) - T:

Concatenated string tensor

Type Constraints¶

T in ( tensor(string) ):

Inputs and outputs must be UTF-8 strings

Examples¶

default¶

import numpy as np
import onnx

node = onnx.helper.make_node(
    "StringConcat",
    inputs=["x", "y"],
    outputs=["result"],
)
x = np.array(["abc", "def"]).astype("object")
y = np.array([".com", ".net"]).astype("object")
result = np.array(["abc.com", "def.net"]).astype("object")

expect(node, inputs=[x, y], outputs=[result], name="test_string_concat")

x = np.array(["cat", "dog", "snake"]).astype("object")
y = np.array(["s"]).astype("object")
result = np.array(["cats", "dogs", "snakes"]).astype("object")

expect(
    node,
    inputs=[x, y],
    outputs=[result],
    name="test_string_concat_broadcasting",
)

x = np.array("cat").astype("object")
y = np.array("s").astype("object")
result = np.array("cats").astype("object")

expect(
    node,
    inputs=[x, y],
    outputs=[result],
    name="test_string_concat_zero_dimensional",
)

x = np.array(["abc", ""]).astype("object")
y = np.array(["", "abc"]).astype("object")
result = np.array(["abc", "abc"]).astype("object")

expect(
    node,
    inputs=[x, y],
    outputs=[result],
    name="test_string_concat_empty_string",
)

x = np.array(["的", "中"]).astype("object")
y = np.array(["的", "中"]).astype("object")
result = np.array(["的的", "中中"]).astype("object")

expect(
    node,
    inputs=[x, y],
    outputs=[result],
    name="test_string_concat_utf8",
)