New text circuit drawer

[albi3ro]

This method can hopefully replace CircuitGraph.draw() and associated objects like CircuitDrawer, RepresentationResolver and Charset.

This function already:

• is much faster (about 8x)
• has more features
  • max_length Issue Preventing printing overflow for long circuits in terminal with qml.draw_circuit() #1841
  • decimals for printing scalar parameters to arbitrary precision
• has fewer bugs and peculiarities
  -naturally supports templates
  -broadcasts qml.state across all wires

Downsides: This function does not display matrix-valued parameters below the circuit diagram. Circuit drawings are for inspecting whether or not everything is generally in the correct place. Users can inspect specific parameters via other means.

with qml.tape.QuantumTape() as tape:
    qml.QFT(wires=(0,1,2))
    qml.RX(1.234, wires=0)
    qml.RY(1.234, wires=1)
    qml.RZ(1.234, wires=2)
    qml.Toffoli(wires=(0,1,"aux"))
    
    qml.expval(qml.PauliZ("aux"))
    qml.state()


>>> print(draw_text(tape))
  0: ─╭QFT──RX─╭C─┤     State
  1: ─├QFT──RY─├C─┤     State
  2: ─╰QFT──RZ─│──┤     State
aux: ──────────╰X─┤ <Z> State

[github-actions]

Hello. You may have forgotten to update the changelog!
Please edit doc/releases/changelog-dev.md with:

• A one-to-two sentence description of the change. You may include a small working example for new features.
• A link back to this PR.
• Your name (or GitHub username) in the contributors section.

[josh146]

Users can inspect specific parameters via other means.

@albi3ro does the new circuit drawer preserve the ability to print scalar parameters, e.g.,

>>> print(draw_text(tape))
  0: ─╭QFT──RX(0.1)─╭C─┤     State
  1: ─├QFT──RY(0.2)─├C─┤     State
  2: ─╰QFT──RZ(0.3)─│──┤     State
aux: ───────────────╰X─┤ <Z> State

I would say that this is a very important feature, since we use this to visualize circuit diagrams with rotation parameters across the docs + blog + demos!

[albi3ro]

@josh146 Yes it can print scaler parameters via something like draw_text(tape, decimals=2). It just won't print out matrices. Though we could add that in if people really want it.

[josh146]

Hey all, just curious on the timeframe for this PR?

[albi3ro]

@josh146 I'm going to make it my first priority next quarter. Mostly just test writing at this point, but those are going to be extensive. It could probably use a round of pre-review if anyone has the time.

[dime10]

Something I noticed while comparing the old drawer to the new, is the change in the qml.PauliRot representation deliberate? :)

0: ─╭QFT──X─╭C─╭X─────╭R(XYY)─┤ ╭<X@Y>
1: ─├QFT────│──├C──RX─├R(XYY)─┤ ╰<X@Y>
2: ─╰QFT────╰X─╰C─────╰R(XYY)─┤  Probs
 0: ──╭QFT──X──╭C──╭X───────────────╭RX(0.1)──╭┤ ⟨X ⊗ Y⟩ 
 1: ──├QFT─────│───├C──RX(0.00335)──├RY(0.1)──╰┤ ⟨X ⊗ Y⟩ 
 2: ──╰QFT─────╰X──╰C───────────────╰RY(0.1)───┤ Probs   

Another thing I'm curious about is the capability of drawing multiple measurement layers. Was this added with complex measurement processes in mind or for practical reasons?

[albi3ro]

[sc-12939]

[Jaybsoni]

Just a few small things but other than that it looks really good! 8x speed up is insane!

[Jaybsoni]

Suggested change

	if isinstance(op, qml.tape.QuantumTape): # isa tape
	if isinstance(op, qml.tape.QuantumTape): # is a tape

[Jaybsoni]

Would it be worth it to add a usage case example for the tape_offset kwarg?

[albi3ro]

adding.

[Jaybsoni]

Just another thing I came across when playing around with it. There are two features of the old text drawer that don't pass over: the numbering of QubitUnitary operations don't show up and the basis with respect to which we sample our circuit isn't specified.

I think the first feature is nice to have since we now omit showing the unitary matrices; if they are labeled the user can at least keep track of them and see if / where they repeat in the circuit.

The second feature would also be a nice to have as it just provides more info in general, but maybe not fully required.

Here is an example of what the current behaviour is:

[Jaybsoni]

Looks great!