Initilize instruction does not write a QASM definition

[omarcostahamido]

Information

when building a quantum circuit from qasm: if the qasm code includes initialize it gets ignored (or throws an error).

• Qiskit Terra version:
  Qiskit | 0.24.1
  -- | --
  Terra | 0.16.4
  Aer | 0.7.6
  Ignis | 0.5.2
  Aqua | 0.8.2
  IBM Q Provider | 0.12.2
• Python version:
  Python | 3.8.8

Steps to reproduce the problem

from qiskit import QuantumCircuit
qc = QuantumCircuit(2)
qc.initialize([0,1,0,0],range(2))
qc.qasm()
qc2 = QuantumCircuit().from_qasm_str(qc.qasm())

[quantumjim]

A point to note about this is that the from_qasm_str() method does not accept a qasm string created by the qasm() method.

[1ucian0]

Is initialize supported by OpenQASM2?

[omarcostahamido]

I don't see a reason why it should not be.

…

On Sat, Apr 17, 2021, 2:14 AM Luciano Bello ***@***.***> wrote: Is initialized supported by OpenQASM2? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#6169 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEL4NAFMX2SXU62K62SIWRLTJFGNRANCNFSM42PB6BWA> .

[1ucian0]

I mean, they could as opaque instructions (one per each size of idlist)

OPENQASM 2.0;
include "qelib1.inc";

opaque initialize4_2(x0, x1, x2, x3) q0, q1;

qreg q[2];
initialize4_2(0,1,0,0) q[0],q[1];

However this wont make it loadable per se (an opaque instruction wont have a definition). Am I missing something?

[jwoehr]

I think the big problem with supporting initialize in OpenQASM 2 is that QuantumCircuit.initialize() can take a vector of complex amplitudes as in the example

circuit.initialize([1/np.sqrt(2), -1/np.sqrt(2)], 0)

in the Qiskit Terra API Documentation. This sort of thing cannot easily be represented within the limited parameter expression syntax of OpenQASM 2.0

[omarcostahamido]

Thank you @jwoehr and @1ucian0 .
I understand qasm is really simple and won't understand np or that it is possibly more work to allow it to understand a vector of complex numbers. But something is better than nothing at all...
In my work to rebuild the qsphere visualization using javascript and opengl inside Max, I just added a feature to my microqiskit object to pass the statevector as a simple list of numbers. Assuming that all states are complex numbers, and that all complex numbers are just a group of 2 values (for real and imaginary components), it is simply waiting to receive a list of ((2^nqubits)*2) values.
Thoughts?

[jwoehr]

Well, I've got a working OpenQASM 2 translator based on the ANTLR4 grammar for OpenQASM 3. It's pretty easy to extend, but I'm not sure how much I can get away with on the current PR..

Currently I'm working on the test suite.

The next thing I was thinking to add is parameterization of translation so .qasm files can be re-used.

Somewhere along the line I should be able to add initialize.

[omarcostahamido]

@jwoehr what would I do without you?
this sounds promising. And yes, I'm all in to become early adopter of OpenQASM 3; we should all move that direction sooner or later. Keep me posted please :)

[jwoehr]

We'll have to see if this PR flies. OpenQASM 3 is a profoundly more complex environment; I'm just using a grammar file authored by the Qiskit team on the basis of OpenQASM 3: A broader and deeper quantum assembly language. As the language evolves I may have to subset that grammar to keep this OpenQASM 2 development manageable. However, it's possible that instead we will move towards an "OpenQASM 2.1, 2.2 ..." which will migrate in stages towards OpenQASM 3.