Override Numpy's copy method to SegyArray to ensure uninterpreted b…

…ytes are preserved. (#78)

* Add special copy method to SegyArray class

A 'copy' method, which maintains raw binary data and padding bytes, has been added to the SegyArray class in arrays.py. This is essential for SEG-Y data processing where unparsed fields are present but preserving raw binary data integrity is crucial.

* Update literal values and copy method static type checking

* Simplify array copy process

* make mypy happy

* Add test for SegyArray copy method

The test has been added to confirm that the copy method of SegyArray can successfully copy a segy array with the exact underlying buffer. The test also verifies that the memory between the original segy array and its copy are not shared.

---------

Co-authored-by: Altay Sansal <altay.sansal@tgs.com>

src/segy/arrays.py

@@ -18,8 +18,12 @@
 from pandas import DataFrame
 
 if TYPE_CHECKING:
+    from typing import Literal
+
     from numpy.typing import NDArray
 
+    OrderKACF = Literal[None, "K", "A", "C", "F"]
+
 
 class SegyArray(np.ndarray):  # type: ignore[type-arg]
     """Base class for array interface. Like ndarray but extensible."""
@@ -33,6 +37,18 @@ def __array_finalize__(self, obj: NDArray[Any] | None) -> None:
         if obj is None:
             return
 
+    def copy(self, order: OrderKACF = "K") -> SegyArray:
+        """Copy structured array preserving the padded bytes as is.
+
+        This method ensures that the copy includes raw binary data and any padding
+        bytes, preserving the entire memory layout of the array. This is necessary
+        for working with SEG-Y data where not all fields are parsed, but raw binary
+        data preservation is crucial.
+        """
+        void_view = self.view("V")
+        void_copy = np.copy(void_view, order=order, subok=True)
+        return void_copy.view(self.dtype)  # type: ignore[no-any-return]
+
 
 class HeaderArray(SegyArray):
     """Header ndarray with convenience features."""

tests/test_arrays.py

@@ -0,0 +1,24 @@
+"""Test Numpy array subclasses."""
+
+import numpy as np
+
+from segy.arrays import SegyArray
+
+
+def test_segy_array_copy() -> None:
+    """Test copying a segy array with exact underlying buffer."""
+    buffer_expected = np.asarray([0, 1, 2, 3, 4], dtype="uint16").tobytes()
+
+    dtype = np.dtype(
+        {
+            "names": ["f1", "f2"],
+            "offsets": [0, 4],
+            "formats": ["uint16", "uint16"],
+            "itemsize": 10,
+        }
+    )
+    segy_array = SegyArray(np.frombuffer(buffer_expected, dtype=dtype))
+    segy_array_copy = segy_array.copy()
+
+    assert segy_array_copy.tobytes() == buffer_expected
+    assert np.may_share_memory(segy_array_copy, segy_array) is False