VecWire

VecWire draws electrical circuit schematics as SVG, with circuit-QED figures as the first use case. The schematic is generated from code and stays editable afterwards: each component becomes a named SVG group, the same script writes byte-identical SVG, and the output opens in Inkscape.

_{A circuit-QED schematic with labels typeset by VecTeX, from examples/fig1a_cqed.py with --tex.}

The symbol set covers wires, capacitors, inductors, Josephson junctions (the bare X, boxed harmonic junctions, and multiharmonic junctions), ground symbols, connection nodes, dashed enclosures, and labels. Components connect named, positioned nodes and draw their own leads.

VecWire is alpha: an early version, with an API that will still change.

Install

VecWire needs Python 3.12 or newer:

python -m pip install vecwire

The only runtime dependency is lxml. For labels typeset by TeX, install the optional VecTeX extra; VecTeX also needs pdflatex and dvisvgm on PATH:

python -m pip install "vecwire[tex]"

Quick start

A capacitively shunted Josephson junction, grounded at the bottom:

from vecwire import Circuit

circuit = Circuit(70, 96)
circuit.row(14, top_left=18, phi=35, top_right=52, dots={"phi"})
circuit.row(78, bottom_left=18, gnd=35, bottom_right=52)

circuit.wire("top_left", "phi", "top_right")
circuit.wire("bottom_left", "gnd", "bottom_right")
circuit.capacitor("top_left", "bottom_left", id="C")
circuit.junction("top_right", "bottom_right", id="JJ")
circuit.ground("gnd", id="ground")

circuit.label("C", (11, 46), anchor="rc")
circuit.label("φ", "phi", offset=(0, -3), anchor="cb")
circuit.write_svg("transmon.svg")

Coordinates are SVG user units, points by default (Circuit(..., unit="pt"); "px" and "mm" are also accepted). to_svg_document() returns the same document as a string.

Examples

python examples/fig1a_cqed.py              # writes examples/output/fig1a.svg
python examples/fig1a_cqed.py --tex        # editable TeX labels through Vectex
python examples/harmonic_junctions.py      # the junction symbol legend

fig1a_cqed.py draws a shunt capacitance C across an inductor L in series with a parallel junction branch of V_J and C_J, marking the node phases φ and δ. harmonic_junctions.py draws the bare X beside the boxed m = 1, 2, 3 and sin symbols, and an inductor in series with a sinusoidal junction next to the multiharmonic junction it acts like. Both previews are checked in: fig1a.svg and harmonic_junctions.svg. Without --tex the examples use plain SVG text and need no TeX installation.

Nodes and wires

A Node is a placed connection point, and wires join nodes into an electrical net. row() and column() declare several nodes sharing one coordinate; nodes() takes arbitrary positions. None of them draws anything. Methods accept node names or the Node values these calls return.

A wire visits its nodes in the order given and may pass through several of them; a route that doubles back is rejected. Crossing or overlapping in the drawing does not make a connection. Optional via points shape a route without adding connections:

circuit.wire("left", "right", via=((18, 30), (52, 30)), corner_radius=2)

Each routed wire is one SVG path, so linejoin="round" softens a bend and corner_radius curves the centerline; both can be set on the Circuit or per wire. A route still passes exactly through every named node it visits, even with a corner radius. Two-terminal symbols are placed horizontally or vertically.

Labels take plain text or a VecTeX fragment; a fragment's SVG group is embedded as is, so it stays editable through TexText.

The Circuit records drawing geometry only. It does not infer electrical topology or compute circuit parameters; those belong to the scientific model that produces the figure.

Plain-text labels accept font_family and size (in the circuit's user units):

circuit.label("Capacitance", (40, 20), size=8, font_family="TeX Gyre Heros")

Omitting font_family preserves inherited SVG typography. For fragment labels, choose typography when rendering the fragment; these text options do not rescale or replace its glyphs.

Josephson junctions

junction() draws the bare X, the standard Josephson-junction symbol, which leaves the current-phase relation unspecified; its leads meet at the centre of the X, so the wire runs through it. harmonic_junction() draws the boxed symbol of one tunnelling harmonic: its square holds order crossed diagonals each way for a cos(mφ) term, in which m Cooper pairs tunnel together, and kind="sin" adds two dots for the odd sin(mφ) term of a φ0 junction. The m = 1 box is the bare X inside its square, so the two symbols line up. In both, color sets the glyph and lead_color keeps the leads in the circuit ink.

circuit.junction("top", "bottom", id="JJ")  # bare X
circuit.harmonic_junction("top", "bottom", 2, color="#8075c1", lead_color="#262626")
circuit.harmonic_junction("top", "bottom", kind="sin")  # sin φ, m = 1

Both groups carry data-component="josephson-junction"; the boxed one also records data-harmonic and data-kind, so a stylesheet or selector can tell them apart.

harmonic_sum() draws a multiharmonic junction: a dashed box holding ⊠ + ⊠⊠ + ⊠⊠⊠ + ⋯ between two terminals. The harmonics of one junction add as energies, so the terms hang in parallel between two rails, which the branch leads join at the centre of the box. rails=False ends each term in terminal dots instead, a shorthand notation. The sum reads left to right on a vertical branch and top to bottom on a horizontal one. colors and kinds set each term, and ellipsis=False drops the trailing + ⋯.

box() draws a dashed enclosure around a subcircuit that acts as one element; it is decoration only and creates no connection. Both return a Box with the bounds: Box.at("rb") uses the label anchors, and Box.terms holds the centres of the summed glyphs for per-term labels:

box = circuit.harmonic_sum("top", "bottom", (1, 2, 3), colors=shades)
for order, centre in zip((1, 2, 3), box.terms):
    circuit.label(f"E{order}", (centre[0], box.bottom + 2), anchor="ct")
circuit.box((10, 18), (60, 95), id="effective-junction")

Architecture

A symbol is a plain function whose first argument is the circuit, and it draws only through the public drawing API: circuit.branch(start, end) resolves the terminals into a Branch, circuit.group(kind, id, branch=...) opens a semantic group with data-from/data-to, and line, rect, dot, and path add strokes in the circuit ink. Symbols are written in branch coordinates (branch.at(along, across)), so one definition serves every orientation. A new symbol is a new function, in vecwire.elements or your own script, with no change to the core:

def resistor(circuit, start, end, *, id=None):
    branch = circuit.branch(start, end)
    group = circuit.group("resistor", id, branch=branch)
    lead, step = branch.length / 4, branch.length / 12
    points = [branch.at(lead + (i + 0.5) * step, 3 * (-1) ** i) for i in range(6)]
    circuit.path(
        group,
        [
            ("M", *branch.start.position),
            ("L", *branch.at(lead)),
            *(("L", *p) for p in points),
            ("L", *branch.at(3 * lead)),
            ("L", *branch.end.position),
        ],
    )


resistor(circuit, "top", "bottom", id="R")

The built-in symbols are also Circuit methods, bound by one class attribute each (capacitor = elements.capacitor), so circuit.capacitor(a, b) is the same call as elements.capacitor(circuit, a, b), with the same signature and docstring. There is no registry or plugin hook. The core imports the library for these bindings, and the library never imports the core at run time (tests/test_elements.py checks this), so the dependency points one way.

The core owns number formatting (.12g) and attribute order, so every symbol serializes coordinates the same way and output is byte-identical from run to run.

Why not svg.py. svg.py was evaluated for the symbols and is not used. It writes floats with repr (0.30000000000000004, 1e-07) and fixes its own attribute order, so adopting it would change every byte of existing figures, or mean passing preformatted strings and losing its typed fields. A symbol's size depends on the circuit's ink, stroke width, and line caps, which svg.py has no notion of, so element code would get longer, not shorter. The canvas needs lxml anyway, to embed VecTeX fragments and reorder connection dots. (Its svg.A is also the <a> element, not an arc.)

Development

python -m pip install -e '.[dev]'
ruff format --check .
ruff check .
ty check
pytest

After touching drawing code, rerun both examples and diff their output against examples/fig1a.svg and examples/harmonic_junctions.svg; every changed byte should be one the change intended. Releases are described in docs/development.md.

Related projects

VecWire is developed alongside FigWorks, which composes multi-panel figures, and two other producers of editable SVG: VecTeX (TeX equations) and VecView (layered 3D schematics). All four share one premise: figures generated from code, with stable ids and byte-identical output, that stay editable in Inkscape.

VecWire does not depend on FigWorks. FigWorks places a Circuit in a panel through to_svg_document() alone; see its README.

License

MIT