Wires final

.github/CHANGELOG.md

@@ -15,12 +15,15 @@
 * Added an Ansatz abstract class and PolyExpAnsatz concrete implementation. This is used in the Bargmann representation.
   [(#295)](https://github.com/XanaduAI/MrMustard/pull/295)
 
-* Added `complex_gaussian_integral` and `real_gaussian_integral` methods.
+* Added `complex_gaussian_integral` method.
   [(#295)](https://github.com/XanaduAI/MrMustard/pull/295)
 
 * Added `Bargmann` representation (parametrized by Abc). Supports all algebraic operations and CV (exact) inner product.
   [(#296)](https://github.com/XanaduAI/MrMustard/pull/296)
 
+* Added a new class `Wires` in `mrmustard.lab` to handle the connectivity of objects in a circuit.
+  [(#330)](https://github.com/XanaduAI/MrMustard/pull/330)
+
 ### Breaking changes
 * Removed circular dependencies by:
   * Removing `graphics.py`--moved `ProgressBar` to `training` and `mikkel_plot` to `lab`.

mrmustard/lab/abstract/measurement.py

@@ -87,12 +87,10 @@ def outcome(self):
         ...
 
     @abstractmethod
-    def _measure_fock(self, other: State) -> Union[State, float]:
-        ...
+    def _measure_fock(self, other: State) -> Union[State, float]: ...
 
     @abstractmethod
-    def _measure_gaussian(self, other: State) -> Union[State, float]:
-        ...
+    def _measure_gaussian(self, other: State) -> Union[State, float]: ...
 
     def primal(self, other: State) -> Union[State, float]:
         """performs the measurement procedure according to the representation of the incoming state"""

mrmustard/lab/abstract/state.py

@@ -591,7 +591,6 @@ def get_modes(self, item) -> State:
         fock_partitioned = fock.trace(self.dm(self.cutoffs), keep=item_idx)
         return State(dm=fock_partitioned, modes=item)
 
-    # TODO: refactor
     def __eq__(self, other) -> bool:  # pylint: disable=too-many-return-statements
         r"""Returns whether the states are equal."""
         if self.num_modes != other.num_modes:

mrmustard/lab/gates.py

@@ -1041,9 +1041,7 @@ def primal(self, state):
 
         coeff = math.cast(
             math.exp(
-                -0.5
-                * self.phase_stdev.value**2
-                * math.arange(-dm.shape[-2] + 1, dm.shape[-1]) ** 2
+                -0.5 * self.phase_stdev.value**2 * math.arange(-dm.shape[-2] + 1, dm.shape[-1]) ** 2
             ),
             dm.dtype,
         )

mrmustard/lab_dev/wires.py

@@ -0,0 +1,327 @@
+# Copyright 2023 Xanadu Quantum Technologies Inc.
+
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+
+#     http://www.apache.org/licenses/LICENSE-2.0
+
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""`Wires` class for handling the connectivity of an object in a circuit."""
+
+from __future__ import annotations
+
+from typing import Iterable, Optional
+
+import numpy as np
+
+from mrmustard import settings
+
+# pylint: disable=protected-access
+# pylint: disable=import-outside-toplevel
+
+
+class Wires:
+    r"""`Wires` class for handling the connectivity of an object in a circuit.
+    In MrMustard, ``CircuitComponent``s have a ``Wires`` object as attribute
+    to handle the wires of the component and to connect components together.
+
+    Wires are arranged into four groups, and each of the groups can
+    span multiple modes:
+                        _______________
+    input bra modes --->|   circuit   |---> output bra modes
+    input ket modes --->|  component  |---> output ket modes
+                        ---------------
+    Each of the four groups can be empty. In particular, the wires of a state have
+    no inputs and the wires of a measurement have no outputs. Similarly,
+    a standard unitary transformation has no bra wires.
+
+    We refer to these four groups in a "standard order":
+        0. output_bra for all modes
+        1. input_bra for all modes
+        2. output_ket for all modes
+        3. input_ket for all modes
+
+    A ``Wires`` object can return subsets (views) of itself. Available subsets are:
+    - input/output  (wires on input/output side)
+    - bra/ket       (wires on bra/ket side)
+    - modes         (wires on the given modes)
+
+    For example, ``wires.input`` returns a ``Wires`` object with only the input wires
+    (on bra and ket sides and on all the modes).
+    Note that these can be combined together: ``wires.input.bra[(1,2)] returns a
+    ``Wires`` object with only the input bra wires on modes 1 and 2.
+    Note these are views of the original ``Wires`` object, i.e. we can set the ``ids`` on the
+    views and it will be set on the original, e.g. this is a valid way to connect two sets
+    of wires by setting their ids to be equal: ``wires1.output.ids = wires2.input.ids``.
+
+    A very useful feature of the ``Wires`` class is the support for the right shift
+    operator ``>>``. This allows us to connect two ``Wires`` objects together as
+    ``wires1 >> wires2``. This will return the ``Wires`` object of the two
+    wires objects connected together as if they were in a circuit:
+               ____________                             ____________
+    in bra --->|  wires1  |---> out bra ---> in bra --->|  wires2  |---> out bra
+    in ket --->|          |---> out ket ---> in ket --->|          |---> out ket
+               ------------                             ------------
+    The returned ``Wires`` object will contain the surviving wires of the two
+    ``Wires`` objects and it will raise an error if there are overlaps between
+    the surviving wires. This is especially useful for handling the ordering of the
+    wires when connecting components together: we are always guaranteed that a
+    ``Wires`` object will provide the wires in the standard order.
+
+    Args:
+        modes_out_bra (Iterable[int]): The output modes on the bra side.
+        modes_in_bra (Iterable[int]): The input modes on the bra side.
+        modes_out_ket (Iterable[int]): The output modes on the ket side.
+        modes_in_ket (Iterable[int]): The input modes on the ket side.
+
+    Note that the order of the modes passed to initialize the object doesn't matter,
+    as they get sorted at init time.
+    """
+
+    def __init__(
+        self,
+        modes_out_bra: Optional[Iterable[int]] = None,
+        modes_in_bra: Optional[Iterable[int]] = None,
+        modes_out_ket: Optional[Iterable[int]] = None,
+        modes_in_ket: Optional[Iterable[int]] = None,
+    ) -> None:
+        modes_out_bra = modes_out_bra or []
+        modes_in_bra = modes_in_bra or []
+        modes_out_ket = modes_out_ket or []
+        modes_in_ket = modes_in_ket or []
+
+        self._modes = sorted(
+            set(modes_out_bra) | set(modes_in_bra) | set(modes_out_ket) | set(modes_in_ket)
+        )
+        randint = settings.rng.integers  # MM random number generator
+        outbra = {m: randint(1, 2**62) if m in modes_out_bra else 0 for m in self._modes}
+        inbra = {m: randint(1, 2**62) if m in modes_in_bra else 0 for m in self._modes}
+        outket = {m: randint(1, 2**62) if m in modes_out_ket else 0 for m in self._modes}
+        inket = {m: randint(1, 2**62) if m in modes_in_ket else 0 for m in self._modes}
+        self._id_array = np.array([[outbra[m], inbra[m], outket[m], inket[m]] for m in self._modes])
+        self._mask = np.ones_like(self._id_array)  # multiplicative mask
+
+    def _args(self):
+        r"""Returns the same args one needs to initialize this object."""
+        ob_modes = np.array(self._modes)[self._id_array[:, 0] > 0].tolist()
+        ib_modes = np.array(self._modes)[self._id_array[:, 1] > 0].tolist()
+        ok_modes = np.array(self._modes)[self._id_array[:, 2] > 0].tolist()
+        ik_modes = np.array(self._modes)[self._id_array[:, 3] > 0].tolist()
+        return tuple(ob_modes), tuple(ib_modes), tuple(ok_modes), tuple(ik_modes)
+
+    @classmethod
+    def _from_data(cls, id_array: np.ndarray, modes: list[int], mask=None):
+        r"""Private class method to initialize Wires object from the given data."""
+        w = cls()
+        w._id_array = id_array
+        w._modes = modes
+        w._mask = mask if mask is not None else np.ones_like(w._id_array)
+        return w
+
+    def _view(self, masked_rows: tuple[int, ...] = (), masked_cols: tuple[int, ...] = ()) -> Wires:
+        r"""A masked view of this Wires object."""
+        w = self._from_data(self._id_array, self._modes, self._mask.copy())
+        w._mask[masked_rows, :] = -1
+        w._mask[:, masked_cols] = -1
+        return w
+
+    def _mode(self, mode: int) -> np.ndarray:
+        "A slice of the id_array matrix at the given mode."
+        return np.maximum(0, self.id_array[[self._modes.index(mode)]])[0]
+
+    @property
+    def id_array(self) -> np.ndarray:
+        "The id_array of the available wires in the standard order (bra/ket x out/in x mode)."
+        return self._id_array * self._mask
+
+    @property
+    def ids(self) -> list[int]:
+        "The list of ids of the available wires in the standard order."
+        flat = self.id_array.T.ravel()
+        return flat[flat > 0].tolist()
+
+    @ids.setter
+    def ids(self, ids: list[int]):
+        "Sets the ids of the available wires."
+        if len(ids) != len(self.ids):
+            raise ValueError(f"wrong number of ids (expected {len(self.ids)}, got {len(ids)})")
+        self._id_array.flat[self.id_array.flatten() > 0] = ids
+
+    @property
+    def modes(self) -> list[int]:
+        "The set of modes spanned by the populated wires."
+        return [m for m in self._modes if any(self.id_array[self._modes.index(m)] > 0)]
+
+    @property
+    def indices(self) -> list[int]:
+        r"""Returns the array of indices of this subset in the standard order.
+        (bra/ket x out/in x mode). Use this to get the indices for bargmann contractions.
+        """
+        flat = self.id_array.T.ravel()
+        flat = flat[flat != 0]
+        return np.where(flat > 0)[0].tolist()
+
+    @property
+    def input(self) -> Wires:
+        "A view of self without output wires"
+        return self._view(masked_cols=(0, 2))
+
+    @property
+    def output(self) -> Wires:
+        "A view of self without input wires"
+        return self._view(masked_cols=(1, 3))
+
+    @property
+    def ket(self) -> Wires:
+        "A view of self without bra wires"
+        return self._view(masked_cols=(0, 1))
+
+    @property
+    def bra(self) -> Wires:
+        "A view of self without ket wires"
+        return self._view(masked_cols=(2, 3))
+
+    @property
+    def adjoint(self) -> Wires:
+        r"""
+        The adjoint (ket <-> bra) of this wires object.
+        """
+        return self._from_data(self._id_array[:, [2, 3, 0, 1]], self._modes, self._mask)
+
+    @property
+    def dual(self) -> Wires:
+        r"""
+        The dual (in <-> out) of this wires object.
+        """
+        return self._from_data(self._id_array[:, [1, 0, 3, 2]], self._modes, self._mask)
+
+    def copy(self) -> Wires:
+        r"""A copy of this Wires object with new ids."""
+        w = Wires(*self._args())
+        w._mask = self._mask.copy()
+        return w
+
+    def __bool__(self) -> bool:
+        return len(self.ids) > 0
+
+    def __getitem__(self, modes: Iterable[int] | int) -> Wires:
+        "A view of this Wires object with wires only on the given modes."
+        modes = [modes] if isinstance(modes, int) else modes
+        idxs = tuple(list(self._modes).index(m) for m in set(self._modes).difference(modes))
+        return self._view(masked_rows=idxs)
+
+    def __lshift__(self, other: Wires) -> Wires:
+        return (other.dual >> self.dual).dual  # how cool is this
+
+    @staticmethod
+    def _outin(si, so, oi, oo):
+        r"""Returns the output and input wires of the composite object made by connecting
+        two single-mode (ket or bra) objects like --|self|-- and --|other|--
+        At this stage we are guaranteed that the configurations `|self|--  |other|--`  and
+        `--|self|  --|other|` (which would be invalid) have already been excluded.
+        """
+        if bool(so) == bool(oi):  # if the inner wires are either both there or both not there
+            return np.array([oo, si], dtype=np.int64)
+        elif not si and not so:  # no wires on self
+            return np.array([oo, oi], dtype=np.int64)
+        else:  # no wires on other
+            return np.array([so, si], dtype=np.int64)
+
+    def __rshift__(self, other: Wires) -> Wires:
+        all_modes = sorted(set(self.modes) | set(other.modes))
+        new_id_array = np.zeros((len(all_modes), 4), dtype=np.int64)
+
+        for m in set(self.modes) & set(other.modes):
+            sob, sib, sok, sik = self._mode(m)  # m-th row of self
+            oob, oib, ook, oik = other._mode(m)  # m-th row of other
+
+            out_bra_issue = sob and oob and not oib
+            out_ket_issue = sok and ook and not oik
+            if out_bra_issue or out_ket_issue:
+                raise ValueError(f"Output wire overlap at mode {m}")
+            in_bra_issue = oib and sib and not sob
+            in_ket_issue = oik and sik and not sok
+            if in_bra_issue or in_ket_issue:
+                raise ValueError(f"Input wire overlap at mode {m}")
+
+            new_id_array[all_modes.index(m)] = np.hstack(
+                [self._outin(sib, sob, oib, oob), self._outin(sik, sok, oik, ook)]
+            )
+        for m in set(self.modes) - set(other.modes):
+            new_id_array[all_modes.index(m)] = self._mode(m)
+        for m in set(other.modes) - set(self.modes):
+            new_id_array[all_modes.index(m)] = other._mode(m)
+
+        return self._from_data(new_id_array, all_modes)
+
+    def __repr__(self) -> str:
+        ob_modes, ib_modes, ok_modes, ik_modes = self._args()
+        return f"Wires({ob_modes}, {ib_modes}, {ok_modes}, {ik_modes})"
+
+    def _repr_html_(self):  # pragma: no cover
+        "A matrix plot of the id_array."
+        row_labels = map(str, self._modes)
+        col_labels = ["bra-out", "bra-in", "ket-out", "ket-in"]
+        array = np.abs(self.id_array) / (self.id_array + 1e-15)
+        idxs = (i for i in self.indices)
+        box_size = "60px"  # Set the size of the squares
+        html = '<table style="border-collapse: collapse; border: 1px solid black;">'
+        # colors
+        active = "#5b9bd5"
+        inactive = "#d6e8f7"
+
+        # Add column headers
+        html += "<tr>"
+        for label in [""] + col_labels:  # Add an empty string for the top-left cell
+            html += f'<th style="border: 1px solid black; padding: 5px;">{label}</th>'
+        html += "</tr>"
+
+        # Initialize rows with row labels
+        rows_html = [
+            f'<tr><td style="border: 1px solid black; padding: 5px;">{label}</td>'
+            for label in row_labels
+        ]
+
+        # Add table cells (column by column)
+        for label, col in zip(col_labels, array.T):
+            for row_idx, value in enumerate(col):
+                color = (
+                    "white"
+                    if np.isclose(value, 0)
+                    else (active if np.isclose(value, 1) else inactive)
+                )
+                cell_html = f'<td style="border: 1px solid black; padding: 5px;\
+                      width: {box_size}px; height: {box_size}px; background-color:\
+                          {color}; box-sizing: border-box;'
+                if color == active:
+                    cell_html += (
+                        f' text-align: center; vertical-align: middle;">{str(next(idxs))}</td>'
+                    )
+                else:
+                    cell_html += '"></td>'
+                rows_html[row_idx] += cell_html
+
+        # Close the rows and add them to the HTML table
+        for row_html in rows_html:
+            row_html += "</tr>"
+            html += row_html
+
+        html += "</table>"
+
+        try:
+            from IPython.core.display import (
+                HTML,
+                display,
+            )
+
+            display(HTML(html))
+        except ImportError as e:
+            raise ImportError(
+                "To display the wires in a jupyter notebook you need to `pip install IPython` first."
+            ) from e

mrmustard/training/callbacks.py

@@ -254,9 +254,9 @@ def call(
             orig_cost = np.array(optimizer.callback_history["orig_cost"][-1]).item()
             obj_scalars[f"{obj_tag}/orig_cost"] = orig_cost
             if self.cost_converter is not None:
-                obj_scalars[
-                    f"{obj_tag}/{self.cost_converter.__name__}(orig_cost)"
-                ] = self.cost_converter(orig_cost)
+                obj_scalars[f"{obj_tag}/{self.cost_converter.__name__}(orig_cost)"] = (
+                    self.cost_converter(orig_cost)
+                )
 
         for k, v in obj_scalars.items():
             tf.summary.scalar(k, data=v, step=self.optimizer_step)

mrmustard/utils/typing.py

@@ -95,5 +95,4 @@
 class Batch(Protocol[T_co]):
     r"""Anything that can iterate over objects of type T_co."""
 
-    def __iter__(self) -> Iterator[T_co]:
-        ...
+    def __iter__(self) -> Iterator[T_co]: ...