The Compute-Action-Uncompute Pattern
A recurring quantum idiom runs, in circuit-execution order, compute → action → compute† (first U, then V, then U†). If you write it with the standard matrix convention where operators act on a state vector with the rightmost applied first, the overall operator is W = U† V U: U first maps the state into a working basis convenient for the action, V does the real work, then U† restores it. qalgora-Q captures this directly so the compiler can both insert the inverse for you and optimise the controlled version.
The pattern
qalgora.compute_action(compute, action) runs compute (U), then
action (V), then automatically appends U†. You never write the uncompute step — it is
synthesised from compute.
import qalgora
@qalgora.kernel
def kernel():
q = qalgora.qvector(2)
theta = 0.5
def compute(): # U
h(q[0])
x.ctrl(q[0], q[1])
def action(): # V
rz(theta, q[1])
# runs compute (U), then action (V), then compute-dagger (U†) automatically
qalgora.compute_action(compute, action)Why it pays off
The real win shows up under control. To build a controlled-W you do not need to control all of U, V and U† — controlling V alone suffices, because the U and U† halves cancel on the branch where the control is |0⟩. The compiler knows this and only controls the action, cutting gate count and circuit depth substantially.
This optimisation applies only when compute is a pure quantum subroutine that is
reversible — no measurement, no reset, no randomness, and no I/O or host-side side effects.
计算-操作-逆计算模式
量子编程里有个反复出现的惯用写法：按电路执行顺序为 compute → action → compute†（先 U、再 V、最后 U†）。 若以"算符作用在态矢量上、最右者最先作用"的标准矩阵约定书写，整体算符为 W = U† V U。 先用 U 把态变换到便于 action 作用的工作基底，执行 V，再用 U† 还原。qalgora-Q 能直接表达这种结构， 于是编译器既能替你自动补上逆操作，又能优化它的受控版本。
该模式
qalgora.compute_action(compute, action) 先跑 compute（U），再跑
action（V），然后自动接上 U†。逆计算这一步不用你写——它会从
compute 自动综合出来。
import qalgora
@qalgora.kernel
def kernel():
q = qalgora.qvector(2)
theta = 0.5
def compute(): # U
h(q[0])
x.ctrl(q[0], q[1])
def action(): # V
rz(theta, q[1])
# 先 compute(U)，再 action(V)，最后自动接上 compute 的共轭(U†)
qalgora.compute_action(compute, action)它的价值所在
真正的好处在受控操作上。要构造受控 W，你不用对 U、V、U† 全都加控制—— 只控制 V 就够了，因为在控制位为 |0⟩ 的分支里，U 和 U† 这两半会相互抵消。编译器 懂得这一点，只给 action 部分加控制，从而大幅减少门的数量和电路深度。
该优化仅适用于 compute 是可逆、无测量、无 reset、无随机数、无 I/O 或主机端副作用的纯量子子程序。