Add nptest.py to a new demos dir, update some text

Author: msullivan

demos/nptest.py

@@ -0,0 +1,77 @@
+# TODO: Integrate this as a runtime test
+
+# Simulate array broadcasting
+
+from typing import Literal as L
+
+from typing import (
+    Literal,
+    Never,
+    GetArg,
+    Bool,
+    Matches,
+    Iter,
+    Slice,
+    Length,
+    IsSub,
+    RaiseError,
+)
+
+
+class Array[DType, *Shape]:
+    def __add__[*Shape2](
+        self,
+        other: Array[DType, *Shape2]
+    ) -> Array[DType, *Merge[tuple[*Shape], tuple[*Shape2]]]:
+        raise BaseException
+
+
+type AppendTuple[A, B] = tuple[
+    *[x for x in Iter[A]],
+    B,
+]
+
+type MergeOne[T, S] = (
+    T
+    if Matches[T, S] or Matches[S, Literal[1]]
+    else S if Matches[T, Literal[1]]
+    else RaiseError[Literal["Broadcast mismatch"], T, S]
+)
+
+type DropLast[T] = Slice[T, Literal[0], Literal[-1]]
+type Last[T] = GetArg[T, tuple, Literal[-1]]
+
+# Matching on Never here is intentional; it prevents stupid
+# infinite recursions.
+type Empty[T] = IsSub[Length[T], Literal[0]]
+
+type Merge[T, S] = (
+    S if Bool[Empty[T]] else T if Bool[Empty[S]]
+    else
+    AppendTuple[
+        Merge[DropLast[T], DropLast[S]],
+        MergeOne[Last[T], Last[S]]
+    ]
+)
+
+a1: Array[float, L[4], L[1]]
+a2: Array[float, L[3]]
+ar = a1 + a2
+reveal_type(ar)  # N: Revealed type is "__main__.Array[builtins.float, Literal[4], Literal[3]]"
+checkr: Array[float, L[4], L[3]] = ar
+
+
+b1: Array[float, int, int]
+b2: Array[float, int]
+reveal_type(b1 + b2)  # N: Revealed type is "__main__.Array[builtins.float, builtins.int, builtins.int]"
+
+
+c1: Array[float, L[4], L[1], L[5]]
+c2: Array[float, L[4], L[3], L[1]]
+reveal_type(c1 + c2)  # N: Revealed type is "__main__.Array[builtins.float, Literal[4], Literal[3], Literal[5]]"
+
+#
+
+err1: Array[float, L[4], L[2]]
+err2: Array[float, L[3]]
+err1 + err2  # E: Broadcast mismatch: Literal[2], Literal[3]

pep.rst

@@ -262,12 +262,12 @@ typing. The situation was substantially improved by the introducing of
 unsupported:
 
  * Adding/removing/modifying a keyword parameter
- * Modifying a variable number of parameters -- XXX: check how well TypeVarTuple does this
+ * Adding/removing/modifying a variable number of parameters. (Though
+   there are potential small extensions to ``TypeVarTuple`` that would
+   allow adding and removing.)
 
 This proposal will cover those cases.
 
-XXX: Ehhhhhh the generic situation could be bad for some of it? For
-partial certainly, I think, which otherwise we can almost do.
 
 NumPy-style broadcasting
 ------------------------