Use property itemsize to estimate size of elems in arrays (#19)

* Use property `.itemsize` to estimate the size of elements in arrays
* Fix size-tests to be platform-independent

pdsa/cardinality/hyperloglog.pyx

@@ -150,8 +150,7 @@ cdef class HyperLogLog:
             Number of bytes allocated for the counter.
 
         """
-        cdef uint8_t bytes_per_counter = 4  # 'L' unsigned long takes 4 bytes
-        return bytes_per_counter * self.num_of_counters
+        return self.num_of_counters * self._counter.itemsize
 
     def __repr__(self):
         return "<HyperLogLog (length: {}, precision: {})>".format(

pdsa/frequency/count_min_sketch.pyx

@@ -235,7 +235,7 @@ cdef class CountMinSketch:
             Number of bytes allocated for the sketch.
 
         """
-        return self._length * sizeof(uint32_t)
+        return self._length * self._counter.itemsize
 
     def __repr__(self):
         return "<CountMinSketch ({} x {})>".format(

pdsa/frequency/count_sketch.pyx

@@ -261,7 +261,7 @@ cdef class CountSketch:
             Number of bytes allocated for the sketch.
 
         """
-        return self._length * sizeof(int32_t)
+        return self._length * self._counter.itemsize
 
     def __repr__(self):
         return "<CountSketch ({} x {})>".format(

tests/cardinality/test_hyperloglog.py

@@ -1,3 +1,5 @@
+
+import array
 import pytest
 
 from math import sqrt
@@ -6,14 +8,20 @@
 
 def test_init():
     hll = HyperLogLog(10)
-    assert hll.sizeof() == 4096, "Unexpected size in bytes"
 
     with pytest.raises(ValueError) as excinfo:
         hll = HyperLogLog(2)
     assert str(excinfo.value) == (
         "Precision has to be in range 4...16")
 
 
+def test_size():
+    hll = HyperLogLog(10)
+
+    element_size = array.array('L', [1]).itemsize
+    assert hll.sizeof() == element_size * len(hll), "Unexpected size in bytes"
+
+
 def test_repr():
     hll = HyperLogLog(6)
 

tests/frequency/test_count_min_sketch.py

@@ -1,11 +1,12 @@
+
+import array
 import pytest
 
 from pdsa.frequency.count_min_sketch import CountMinSketch
 
 
 def test_init():
     cms = CountMinSketch(2, 4)
-    assert cms.sizeof() == 32, 'Unexpected size in bytes'
 
     with pytest.raises(ValueError) as excinfo:
         cms = CountMinSketch(0, 5)
@@ -18,6 +19,13 @@ def test_init():
     )
 
 
+def test_size():
+    cms = CountMinSketch(2, 4)
+
+    element_size = array.array('I', [1]).itemsize
+    assert cms.sizeof() == element_size * len(cms), "Unexpected size in bytes"
+
+
 def test_create_from_expected_error():
     cms = CountMinSketch.create_from_expected_error(0.000001, 0.01)
     assert repr(cms) == "<CountMinSketch (5 x 2718282)>"

tests/frequency/test_count_sketch.py

@@ -1,3 +1,4 @@
+import array
 import pytest
 
 from pdsa.frequency.count_sketch import CountSketch
@@ -18,6 +19,13 @@ def test_init():
     )
 
 
+def test_size():
+    cs = CountSketch(2, 4)
+
+    element_size = array.array('i', [1]).itemsize
+    assert cs.sizeof() == element_size * len(cs), "Unexpected size in bytes"
+
+
 def test_create_from_expected_error():
     cs = CountSketch.create_from_expected_error(0.0001, 0.01)
     assert repr(cs) == "<CountSketch (5 x 271828209)>"