More explicit documentation of endianness (#1500)

qualtran/_infra/data_types.py

@@ -16,7 +16,7 @@
 We often wish to write algorithms which operate on quantum data. One can think
 of quantum data types, similar to classical data types, where a collection of
 qubits can be used to represent a specific quantum data type (eg: a quantum
-integer of width 32 would comprise of 32 qubits, similar to a classical uint32
+integer of width 32 would comprise 32 qubits, similar to a classical uint32
 type). More generally, many current primitives and algorithms in qualtran
 implicitly expect registers which represent signed or unsigned integers,
 fixed-point (fp) numbers , or “classical registers” which store some classical
@@ -40,7 +40,7 @@
 type assuming the bitsizes match. QInt(32) == QAny(32), QInt(32) !=
 QFxp(32, 16). QInt(32) != QUInt(32).
 5. We assume a big endian convention for addressing QBits in registers
-throughout qualtran. Recall that in a big endian convention the most signficant
+throughout qualtran. Recall that in a big endian convention the most significant
 bit is at index 0. If you iterate through the bits in a register they will be
 yielded from most significant to least significant.
 6. Ones' complement integers are used extensively in quantum algorithms. We have
@@ -181,7 +181,7 @@ def __str__(self):
 
 @attrs.frozen
 class QAny(QDType):
-    """Opaque bag-of-qbits type."""
+    """Opaque bag-of-qubits type."""
 
     bitsize: SymbolicInt
 
@@ -214,7 +214,10 @@ def assert_valid_classical_val_array(self, val_array, debug_str: str = 'val'):
 class QInt(QDType):
     """Signed Integer of a given width bitsize.
 
-    A two's complement representation is assumed for negative integers.
+    A two's complement representation is used for negative integers.
+
+    Here (and throughout Qualtran), we use a big-endian bit convention. The most significant
+    bit is at index 0.
 
     Attributes:
         bitsize: The number of qubits used to represent the integer.
@@ -275,7 +278,11 @@ def __str__(self):
 class QIntOnesComp(QDType):
     """Signed Integer of a given width bitsize.
 
-    A ones' complement representation is assumed for negative integers.
+    In contrast to `QInt`, this data type uses the ones' complement representation for negative
+    integers.
+
+    Here (and throughout Qualtran), we use a big-endian bit convention. The most significant
+    bit is at index 0.
 
     Attributes:
         bitsize: The number of qubits used to represent the integer.
@@ -323,8 +330,11 @@ def assert_valid_classical_val(self, val, debug_str: str = 'val'):
 class QUInt(QDType):
     """Unsigned integer of a given width bitsize which wraps around upon overflow.
 
-    Similar to unsigned integer types in C. Any intended wrap around effect is
-    expected to be handled by the developer.
+    Any intended wrap around effect is expected to be handled by the developer, similar
+    to an unsigned integer type in C.
+
+    Here (and throughout Qualtran), we use a big-endian bit convention. The most significant
+    bit is at index 0.
 
     Attributes:
         bitsize: The number of qubits used to represent the integer.

qualtran/_infra/data_types_test.py

@@ -295,6 +295,9 @@ def test_qint_to_and_from_bits():
         assert qint4.from_bits(qint4.to_bits(x)) == x
     assert list(qint4.to_bits(-2)) == [1, 1, 1, 0]
     assert list(QInt(4).to_bits(2)) == [0, 0, 1, 0]
+    # MSB at lowest index -- big-endian
+    assert qint4.from_bits([0, 0, 0, 1]) == 1
+    assert qint4.from_bits([0, 0, 0, 1]) < qint4.from_bits([0, 1, 0, 0])
     assert qint4.from_bits(qint4.to_bits(-2)) == -2
     assert qint4.from_bits(qint4.to_bits(2)) == 2
     with pytest.raises(ValueError):
@@ -325,6 +328,10 @@ def test_quint_to_and_from_bits():
     assert [*quint4.get_classical_domain()] == [*range(0, 16)]
     assert list(quint4.to_bits(10)) == [1, 0, 1, 0]
     assert quint4.from_bits(quint4.to_bits(10)) == 10
+    # MSB at lowest index -- big-endian
+    assert quint4.from_bits([0, 0, 0, 1]) == 1
+    assert quint4.from_bits([0, 0, 0, 1]) < quint4.from_bits([1, 0, 0, 0])
+
     for x in range(16):
         assert quint4.from_bits(quint4.to_bits(x)) == x
     with pytest.raises(ValueError):