Keep engineering labels crisp

01 / TextCompose scripts without baseline drift

There is no TeX runtime here, and the labels still read like engineering. Grouped _ and ^ scripts nest, and bounded commands such as \Omega, \pi, \theta, \mu, and \infty resolve to glyphs. Each generated tspan carries an explicit baseline correction, so text following a nested script lands back on the parent baseline rather than drifting. Rotation stops at the vector: the label sits outside the rotated group and stays upright.

source:V4 "I_{load}^{DC}" at (100, 150) #blue [type=current-dc]
capacitor:C4 "47 \mu F" at (330, 150) #cyan [orientation=down]
qgate:Q4 "R_x" at (570, 150) #purple [parameter="\theta/3" phase="e^{-i\pi/4}"]
port:O4 "\omega = 2\pi f" at (800, 150) #emerald

Width is estimated, never measured — no font is loaded at any point. The estimator is deterministic and deliberately conservative, and a long label grows the component's bounds before placement is validated. That ordering is what lets the whole thing work without an SSR-incompatible getBBox().

02 / QuantumUse polished custom gate rows

qgate is the escape hatch, and it shares hadamard's exact shell — same stub lengths, corner radius, focus treatment, and ports. Its parameter, phase, and matrix rows grow the body deterministically, so a custom operator can wear its own definition without becoming a different kind of object.

prepare:PD "|+\rangle" at (80, 130) #blue
rz:RD "R_z" at (270, 130) #cyan [parameter="\varphi"]
qgate:UD "U" at (480, 130) #purple [parameter="\beta" matrix="[e,f;g,h]"]
measure:MD "M_z" at (700, 130) #emerald

PD.out -> RD.in #blue [quantum line]
RD.out -> UD.in #cyan [quantum line]
UD.out -> MD.in #purple [quantum line]