[huuan26]

Summary & Context

Fixes #191

In issue #191, an inconsistency was identified regarding how day count conventions are handled when computing discount factors from curves. While the core DiscountCurve.df(self, dt) was refactored in commit 310bab06 to consistently use self.curve_dc_type, CDSCurve in financepy/market/curves/cds_curve.py retained several day-count inconsistencies and broken vectorization:

1. Day Count Inconsistency in CDSCurve.df(self, dt):
   When passed a Date, CDSCurve.df manually calculated t = (dt - self.anchor_dt) / G_DAYS_IN_YEAR (hardcoded ACT/365), completely bypassing self.libor_curve.curve_dc_type.
2. Broken Vectorization on Date Lists:
   Both CDSCurve.df(self, dt) and CDSCurve.survival_prob(self, dt) advertise vectorization support in their docstrings. However, when passed a list[Date]:
   • CDSCurve.df ran t = np.array(dt), creating an object array of Date instances and passing it to libor_curve.df_t(t), which crashed with TypeError: float() argument must be a string or a real number, not 'Date'.
   • CDSCurve.survival_prob similarly passed an object array of Date instances to _uinterpolate, causing an AttributeError / TypeError.

---

Changes

1. CDSCurve.df(self, dt):
   • Delegates directly to self.libor_curve.df(dt) when dt is a Date or list[Date]. This ensures the underlying discount curve's curve_dc_type is respected and returns vectorized results (np.ndarray) seamlessly.
   • Delegates to self.libor_curve.df_t when dt is numeric (float, int, or numeric np.ndarray).
2. CDSCurve.survival_prob(self, dt):
   • Converts Date or list[Date] via times_from_dates(self.anchor_dt, dt, self.dc_type) into year fractions before calling _uinterpolate.
   • Properly handles scalar dates, date lists, floats, and numeric numpy arrays.
3. CDSCurve.build_curve:
   • Uses times_from_dates(self.anchor_dt, maturity_dt, self.dc_type) for maturity conversion, ensuring consistency with survival_prob.
4. tests/unit/test_FinCDSCurve.py:
   • Re-enabled previously commented-out test_fin_cds_curve() updated for the modern API (removed deprecated use_cache, updated cds.value return tuple).
   • Added explicit assertions verifying:
     • Scalar Date evaluations for df and survival_prob.
     • Vectorized list[Date] evaluations matching scalar calls.
     • Numeric float and array evaluations.
     • Day-count consistency when the underlying discount curve uses an alternative day-count convention (e.g. ACT_360).

---

Validation

• Verified with pytest tests/unit/test_FinCDSCurve.py (passed).
• Verified full test suite across the repository (pytest tests/unit/): 1366 passed, 0 failures, 0 errors.