Merge pull request #683 from padix-key/compress

Do not try integer packing for huge values

src/biotite/application/viennarna/rnaplot.py

@@ -99,8 +99,12 @@ def run(self):
         self._in_file.write(self._dot_bracket)
         self._in_file.flush()
         self.set_arguments(
-            ["-i", self._in_file.name, "-o", "xrna", "-t", self._layout_type]
-        )
+            [
+                "-i", self._in_file.name,
+                "--output-format", "xrna",
+                "-t", self._layout_type,
+            ]
+        )  # fmt: skip
         super().run()
 
     def evaluate(self):

src/biotite/structure/io/pdbx/compress.py

@@ -170,6 +170,14 @@ def _find_best_integer_compression(array):
                 array_after_rle = array_after_delta
             for packed_byte_count in [None, 1, 2]:
                 if packed_byte_count is not None:
+                    # Quickly check this heuristic
+                    # to avoid computing an exploding packed data array
+                    if (
+                        _estimate_packed_length(array_after_rle, packed_byte_count)
+                        >= array_after_rle.nbytes
+                    ):
+                        # Packing would not reduce the size
+                        continue
                     encoding = IntegerPackingEncoding(packed_byte_count)
                     array_after_packing = encoding.encode(array_after_rle)
                     encodings_after_packing = encodings_after_rle + [encoding]
@@ -194,6 +202,29 @@ def _find_best_integer_compression(array):
     return best_encoding_sequence, smallest_size
 
 
+def _estimate_packed_length(array, packed_byte_count):
+    """
+    Estimate the length of an integer array after packing it with a given number of
+    bytes.
+
+    Parameters
+    ----------
+    array : numpy.ndarray
+        The array to pack.
+    packed_byte_count : int
+        The number of bytes used for packing.
+
+    Returns
+    -------
+    length : int
+        The estimated length of the packed array.
+    """
+    # Use int64 to avoid integer overflow in the following line
+    max_val_per_element = np.int64(2 ** (8 * packed_byte_count))
+    n_bytes_per_element = packed_byte_count * (np.abs(array // max_val_per_element) + 1)
+    return np.sum(n_bytes_per_element, dtype=np.int64)
+
+
 def _to_smallest_integer_type(array):
     """
     Convert an integer array to the smallest possible integer type, that is still able

src/biotite/structure/io/pdbx/encoding.pyx

@@ -696,7 +696,7 @@ class IntegerPackingEncoding(Encoding):
         # Get length of output array
         # by summing up required length of each element
         cdef int number
-        cdef int length = 0
+        cdef long length = 0
         for i in range(data.shape[0]):
             number = data[i]
             if number < 0: