cmd: python3 ast_identity.py (bak-r10 vs cert.py)
AST (docstrings stripped) identical: True
string constants differing: 1 removed, 1 added
REMOVED len=6835
ADDED len=6980
--- 
+++ 
@@ -109,11 +109,14 @@
     roughly 3-16 seconds; the mixed walk, which builds a
     Fraction per element, runs at roughly 6-17
     microseconds per element and the same field in
-    roughly 85-160 seconds; the complex walk runs at
-    roughly 22-35 microseconds per element, a 10^6-element
-    complex field in roughly 24-35 seconds. The float
-    path, for contrast, is numpy's at a tenth of a
-    microsecond per element or less. No field in
+    roughly 60-170 seconds -- the per-element range
+    above, carried to the field's size; the complex
+    walk runs at roughly 22-35 microseconds per
+    element, a 10^6-element complex field in roughly
+    24-35 seconds. The float path, for contrast, is
+    numpy's at roughly a tenth of a microsecond per
+    element -- the deposited contrast runs read
+    0.007-0.122 microseconds per element. No field in
     the certificate's own battery comes near those sizes
     -- the real harness's fields run to thousands of
     elements, where the dearest walk is a third of a
OK
exit: 0

--- textual diff (bak-r10 -> cert.py) ---
1018,1022c1018,1025
<     roughly 85-160 seconds; the complex walk runs at
<     roughly 22-35 microseconds per element, a 10^6-element
<     complex field in roughly 24-35 seconds. The float
<     path, for contrast, is numpy's at a tenth of a
<     microsecond per element or less. No field in
---
>     roughly 60-170 seconds -- the per-element range
>     above, carried to the field's size; the complex
>     walk runs at roughly 22-35 microseconds per
>     element, a 10^6-element complex field in roughly
>     24-35 seconds. The float path, for contrast, is
>     numpy's at roughly a tenth of a microsecond per
>     element -- the deposited contrast runs read
>     0.007-0.122 microseconds per element. No field in