From 358cd791c9d661617e5a83c813c95cc1682683c5 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Wed, 18 Feb 2026 17:17:46 +0000
Subject: [PATCH 1/2] Initial plan


From 2746a76674d28cc6c4f59fe6a11f5129c3039efa Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Wed, 18 Feb 2026 17:20:21 +0000
Subject: [PATCH 2/2] Document tobytes(order="C") behavior in _hash_numpy_array
 docstring

Co-authored-by: shaypal5 <917954+shaypal5@users.noreply.github.com>
---
 src/cachier/config.py | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/src/cachier/config.py b/src/cachier/config.py
index aa1ae0d5..1732e117 100644
--- a/src/cachier/config.py
+++ b/src/cachier/config.py
@@ -29,6 +29,12 @@ def _is_numpy_array(value: Any) -> bool:
 def _hash_numpy_array(hasher: "hashlib._Hash", value: Any) -> None:
     """Update hasher with NumPy array metadata and buffer content.
 
+    The array content is converted to bytes using C-order (row-major) layout
+    to ensure consistent hashing regardless of memory layout. This operation
+    may create a copy if the array is not already C-contiguous (e.g., for
+    transposed arrays, sliced views, or Fortran-ordered arrays), which has
+    performance implications for large arrays.
+
     Parameters
     ----------
     hasher : hashlib._Hash
@@ -36,6 +42,13 @@ def _hash_numpy_array(hasher: "hashlib._Hash", value: Any) -> None:
     value : Any
         A NumPy ndarray instance.
 
+    Notes
+    -----
+    The ``tobytes(order="C")`` call ensures deterministic hash values by
+    normalizing the memory layout, but may incur a memory copy for
+    non-contiguous arrays. For optimal performance with large arrays,
+    consider using C-contiguous arrays when possible.
+
     """
     hasher.update(b"numpy.ndarray")
     hasher.update(value.dtype.str.encode("utf-8"))