fix typo in soft_compare (#12054) (#12073)

* fix logic in soft_compare

* black

* add release note

* Update releasenotes/notes/fix_soft_compare-3f4148aab3a4606b.yaml

---------

Co-authored-by: Luciano Bello <bel@zurich.ibm.com>
(cherry picked from commit 5e7a6d0)

Co-authored-by: ewinston <ewinston@us.ibm.com>

qiskit/circuit/instruction.py

@@ -241,8 +241,8 @@ def soft_compare(self, other: "Instruction") -> bool:
         """
         if (
             self.name != other.name
-            or other.num_qubits != other.num_qubits
-            or other.num_clbits != other.num_clbits
+            or self.num_qubits != other.num_qubits
+            or self.num_clbits != other.num_clbits
             or len(self.params) != len(other.params)
         ):
             return False

releasenotes/notes/fix_soft_compare-3f4148aab3a4606b.yaml

@@ -0,0 +1,4 @@
+---
+fixes:
+  - |
+    The method :meth:`qiskit.instruction.Instruction.soft_compare` is meant to compare whether two gates match in their name, number of qubits, number of clbits, and the number of parameters. However, there was a typo where it would not check the number of qubits and number of clbits for a match. This resolves the apparent typo.

test/python/circuit/test_instructions.py

@@ -118,6 +118,12 @@ def test_instructions_soft_compare(self):
             Instruction("u", 1, 0, [0.4, phi]).soft_compare(Instruction("v", 1, 0, [theta, phi]))
         )
 
+        # Test that when names are the same but number of qubits differ we get False
+        self.assertFalse(Instruction("u", 1, 0, []).soft_compare(Instruction("u", 2, 0, [])))
+
+        # Test that when names are the same but number of clbits differ we get False
+        self.assertFalse(Instruction("u", 1, 0, []).soft_compare(Instruction("u", 1, 1, [])))
+
         # Test cutoff precision.
         self.assertFalse(
             Instruction("v", 1, 0, [0.401, phi]).soft_compare(Instruction("v", 1, 0, [0.4, phi]))