Bond coupon BPS and ATM rate

These additive helpers expose existing BondFunctions analytics on ordinary fixed-rate, zero-coupon and floating-rate bonds through Python and C/Go. They do not change core pricing mathematics or existing constructors.

Python method	Go method	Meaning
bps_from_curve(curve, settlement=...)	BPSFromCurve(curve, settlement)	Coupon-rate BPS per 100 with curve discounting
bps_from_yield(yield_rate, dc, compounding, frequency, settlement=...)	BPSFromYield(rate, config)	Coupon-rate BPS per 100 with flat-yield discounting
atm_rate_from_curve(curve, settlement=...)	ATMRateFromCurve(curve, settlement)	Coupon rate repricing the bond's own curve NPV

Units and target

BPS means a uniform one-basis-point change in coupon rates, not a one-basis-point change in discount yield. The value is normalized per 100 of current notional. For otherwise identical ordinary bullet bonds, changing face amount does not change these per 100 results. In contrast, basis_point_value / BasisPointValue is the existing signed, unscaled currency yield sensitivity.

ATM uses the existing core default target: the bond's own curve-discounted NPV. It is not a par-price target and has no caller-supplied price overload. With a single fixed coupon rate, the result recovers that coupon rate. With differing coupon rates, it is their discounted-accrual-weighted rate. Redemptions are not coupons and do not create BPS. A zero-coupon bond returns zero for all three queries, preserving native behavior.

Dates, ownership and pricing lifecycle

• Explicit settlement is an actual date, not a reference date to advance by the bond's settlement lag. Omitted settlement uses the existing checked bond default and requires an evaluation date. Nontradable settlement is an error.
• Settlement-date flows and ex-coupon coupons are excluded, preserving core laws.
• BPS reads coupon notionals/accrual periods, not coupon amounts. A floating bond can therefore have BPS even when a required historical fixing is unavailable. ATM needs coupon amounts and propagates fixing/pricer errors.
• No pricing engine is needed. An attached or frozen engine does not supply the analytical answer. Queries use current explicit curve/coupon data without rewriting cached instrument results.
• The bond retains its own cashflows/settings and floating dependencies. Go still requires the supplied curve/day counter wrappers to be live and from the same session. Close a session or object explicitly; no new finalizers are introduced.
• Callable bonds are rejected. These methods do not supply callable effective sensitivity, a new pricing engine, all BondFunctions overloads or full QuantLib parity. The separate existing callback-cycle limitation remains unchanged.

Examples

Python: example/python/bond_coupon_analytics.py.

Go: sdk/go/examples/bond-coupon-analytics/main.go.

Validation and provenance

The shared 55-row binary64 table exercises all five existing compounding modes and four frequencies plus zero-coupon and ex-coupon cases. It is generated by independent calls to the separately compiled QuantLib 1.43 distribution using scripts/fixtures/bond_coupon_analytics/generate.py. This distribution is not claimed to be built from the local QuantLib source checkout 9863b578. Public regression tests additionally exercise analytical annuity laws, units, errors, live inputs, retained dependencies and missing fixings.