complex op• Data kinds: cpoints × cpoints → table
• Call: import fullseye as fs; fs.ledger.cplx_laurent_coeffs(z, fz, kmin=-1, kmax=4, method='exact') (to call the implementation directly, import mathops; mathops.cplx_laurent_coeffs(z, fz, kmin=-1, kmax=4, method='exact'); from the registry, opsmath.get("cplx_laurent_coeffs"))
Laurent (and Taylor) coefficients on a uniformly sampled circle — residues included.
For f holomorphic on an annulus around c, f(z) = sum_k c_k (z - c)^k with c_k = 1/(2*pi*i) ∮ f(zeta)/(zeta - c)^(k+1) dzeta. On a circle of radius r sampled at n equally spaced angles this becomes a discrete Fourier sum, c_k = (1/(n r^k)) sum_j f_j exp(-i k theta_j) — the trapezoidal rule in the angle, where it converges geometrically rather than as O(n^-2) (Trefethen & Weideman 2014, "The exponentially convergent trapezoidal rule").
c_-1 is the residue at c (when c is the only singularity inside), c_k for k >= 0 are the Taylor coefficients f^(k)(c)/k!, and a non-zero c_-m for m > 1 reveals a pole of order m. Measured on the unit circle with f = 1/(z - 0.5), n = 64: c_-1 = 1 and c_-2 = 0.5 to 1e-16 (machine precision).
method="exact" (default) preserves the direct Fourier sum. method="fast" uses an FFT only when angular gaps agree within 64 * float64 epsilon; otherwise it routes to the exact sum. On 4,096 samples and 201 orders the fast path measured over 15 times faster, while differing from the direct sum by less than 1e-12.
Returns a dict: k (int64 orders, kmin..kmax) · c (complex128 coefficients) · center · radius. The centre is the sample mean, which is exact for a uniformly sampled circle.
Orientation, and how it differs from the rest of the family: the sum runs over the sample *set*, not the sample *order*, so this op always returns the coefficients of the positively oriented circle — the standard definition — whatever order the points arrive in. Feed a clockwise circle and c_-1 still comes back + the residue, while cplx_contour_integral / (2*pi*i) on the same points returns - it (verified). Both are right; they answer different questions (the intrinsic coefficient vs the integral along *this* traversal). Do not cross-check one against the other without fixing the orientation first.
Honest limitation — aliasing: the discrete sum cannot distinguish c_k from c_{k+n}, so a coefficient carries the alias sum sum_m c_{k+m n} r^{m n}. That is negligible for a rapidly converging series (the 0.5^64 term above) and ruinous near the annulus boundary. Requesting more than n coefficients is refused for the same reason.
Raises ValueError: the samples are not a uniformly spaced circle (unequal radii or unequal angular gaps beyond 1e-8 relative — this op is *not* valid on an arbitrary contour, and silently pretending otherwise would return numbers that mean nothing), kmin > kmax, more than n coefficients requested, non-integer orders, and a coefficient that overflowed (r^-k for a small radius and a large negative order).
HALCON: no operator.
Every mathops op validates its input before computing (nothing slips through silently):
• complex input raises ValueError — coercing to float64 silently discards the imaginary part (numpy only emits a ComplexWarning and returns a plausible-looking wrong real number). State .real/.imag/abs() explicitly, or use complexops, which handles complex data.
• masked arrays with masked elements raise ValueError — the implicit conversion that peels off the mask and uses the raw values underneath is refused. Say explicitly whether to fill or to drop.
• NaN/Inf raises ValueError on every input (refused with the count stated — it propagates through the whole result).
• Shapes are strict: 1-D and 2-D are never implicitly promoted or broadcast (a matrix in a vector slot, or a vector in a matrix slot, raises ValueError; reshape explicitly).
• Size cap: ops that take a matrix, and the stat_histogram bins, raise ValueError beyond mathops.MAX_ELEMENTS (2^26 ≈ 67 million elements).
• Sample-data catalog (download URLs / licences) — 2-D uses skimage.data (BSD/public domain) plus synthetic images; 3-D lists download URLs for real data sources (Stanford, PDS, …).
• Operator provenance and references — the sources of the research/methods this op family came from.
• The canonical algorithm (author, year) and its uses are named in the family usage guide above.
• math_complex — py -3.11 examples/math_complex.py
table as input)dynsys_poincare_section · dwt_inverse · chebyshev_eval_nd
complex)cplx_contour_circle · cplx_poly_eval · cplx_contour_integral · cplx_winding_number · cplx_cauchy_value · cplx_argument_principle · cplx_joukowski · cplx_mobius
*Provenance: mathops.py — MATH operator registry. This per-op note is generated by tools/opdocs.py md (do not hand-edit).*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.