go-approx: interp v0.2.1 released, no open gaps

content/en/_index.md

@@ -32,7 +32,7 @@ MATLAB and Mathematica. Wave 1 is under way, in four organisations:
 
 | Organisation | Package | Status | Pages |
 | --- | --- | --- | --- |
-| go-approx | [interp](https://github.com/go-approx/interp) | released v0.2.0 | [go-approx]({{< relref "approx" >}}) |
+| go-approx | [interp](https://github.com/go-approx/interp) | released v0.2.1 | [go-approx]({{< relref "approx" >}}) |
 | go-approx | [poly](https://github.com/go-approx/poly) | released v0.1.0 | [go-approx]({{< relref "approx" >}}) |
 | go-dsp | [signal](https://github.com/go-dsp/signal) | released v0.1.0 | [go-dsp]({{< relref "dsp" >}}) |
 | go-sparse | [sparse](https://github.com/go-sparse/sparse) | released v0.1.0 | [go-sparse]({{< relref "sparse" >}}) |

content/en/approx.md

@@ -1,6 +1,6 @@
 ---
 title: "go-approx"
-description: "Interpolation, splines and polynomials. interp is released v0.2.0 and poly v0.1.0."
+description: "Interpolation, splines and polynomials. interp is released v0.2.1 and poly v0.1.0."
 ---
 
 go-approx is the home of approximation in the parity plan: interpolation,
@@ -9,18 +9,18 @@ data.
 
 | Package | Status | Repository |
 | --- | --- | --- |
-| interp | released v0.2.0 | [go-approx/interp](https://github.com/go-approx/interp) |
+| interp | released v0.2.1 | [go-approx/interp](https://github.com/go-approx/interp) |
 | poly | released v0.1.0 | [go-approx/poly](https://github.com/go-approx/poly) |
 
 ## What is verified
 
 From [STATUS.md](https://github.com/go-scikit/roadmap/blob/main/STATUS.md):
 `numpy.interp`, `interp1d`, PCHIP, Akima and `poly1d` are bit-identical to
-NumPy and SciPy; splines agree within 1e-15 to 1e-10. `interp` v0.2.0 adds `Splrep`
-(FITPACK's legacy `splrep`) and is faster than SciPy on every case measured
-except two still open: `RegularGridInterpolator` linear evaluation at 10^3 to
-10^4 points (0.95 to 0.98x) and shuffled-query cubic evaluation at 10^4 points
-(0.94x). The speed comparison with the Python reference is in
+NumPy and SciPy; splines agree within 1e-15 to 1e-10. `interp` v0.2.0 added `Splrep`
+(FITPACK's legacy `splrep`); v0.2.1 closes the last two gaps, `RegularGridInterpolator`
+linear evaluation (7 to 10x faster than SciPy) and shuffled-query cubic
+evaluation (3 to 9x faster). Every measured case of `interp` and `poly` is now
+at or below the time of SciPy and NumPy. The speed comparison with the Python reference is in
 [interp's docs/perf.md](https://github.com/go-approx/interp/blob/main/docs/perf.md).
 
 ## Where to read more

hugo.yaml

@@ -87,10 +87,11 @@ params:
   # The release of each library these pages describe, for the member
   # organisations that are new in the parity plan. The newest minor version
   # among them is the documentation's version label (v0.1.0 -> 0.1): the CI
-  # reads it here. interp is released v0.2.0, so the documentation is version
-  # 0.2; the 0.1 pages stay published beside it.
+  # reads it here. interp is released v0.2.1, so the documentation is version
+  # 0.2 (a patch release does not make a new version); the 0.1 pages stay
+  # published beside it.
   libraries:
-    interp: "v0.2.0"
+    interp: "v0.2.1"
     poly: "v0.1.0"
     signal: "v0.1.0"
     sparse: "v0.1.0"