Added an option to generate pseudo-CHI numbers as UUIDs

exhibit/core/generate/tests/test_uuid.py

@@ -79,6 +79,42 @@ def test_uuid_column_generation_with_more_uuids_than_requested(self):
 
         self.assertEqual(len(result), 100)
 
+    def test_uuid_column_generation_with_pseudo_chis_standalone(self):
+        '''
+        Dummy CHIs are only similar to real CHIs in that they have 10 digits
+        and that they can start from zero. Neither DOB, nor Gender is encoded,
+        plus the month of the DOB part is always set to 13 to avoid even 
+        accidental matches. The algorithm is completely deterministic.
+        '''
+
+        n = 1000
+        pseudo_chis = tm._generate_pseudo_chis(n=n)
+
+        self.assertEqual(len(set(pseudo_chis)), n)
+
+    def test_uuid_column_generation_with_pseudo_chis(self):
+        '''
+        Pseudi CHIs are only similar to real CHIs in that they have 10 digits
+        and that they can start from zero. Neither DOB, nor Gender is encoded,
+        plus the month of the DOB part is always set to 13 to avoid even 
+        accidental matches. The algorithm is completely deterministic.
+        '''
+
+        n = 1000
+        expected = 900 # 800 CHIs to appear once and 100 CHIs appear twice.
+
+        freq_dist = [
+            "frequency | probability_vector",
+            "1         | 0.800",
+            "2         | 0.200"
+        ]
+
+        result = tm.generate_uuid_column(
+            "pseudo_chi", n, 0, freq_dist, 0, uuid_type="pseudo_chi")
+
+
+        self.assertEqual(len(set(result)), expected)
+
 if __name__ == "__main__" and __package__ is None:
     #overwrite __package__ builtin as per PEP 366
     __package__ = "exhibit"

exhibit/core/generate/uuids.py

@@ -40,6 +40,18 @@ def generate_uuid_column(
     uuids = []
     range_max = 0
 
+    # we need to know the total number of pseudo-chis ahead of generation time
+    # to ensure consistency with random seed and to avoid duplicates so that the first
+    # N of pseudo-chis for a given seed will always be the same.
+
+    if uuid_type == "pseudo_chi":
+        pseudo_chi_total = 0
+        for row in freq_df.itertuples():
+            _num_rows = int(np.ceil(
+            num_rows * float(row.probability_vector) / int(row.frequency)))
+            pseudo_chi_total = pseudo_chi_total + _num_rows
+            pseudo_chis = _generate_pseudo_chis(n=pseudo_chi_total, seed=seed)
+
     for row in freq_df.itertuples():
 
         # always round up the number of generated rows before casting to int
@@ -53,6 +65,12 @@ def generate_uuid_column(
             range_max = range_max + _num_rows
             continue
 
+        if uuid_type == "pseudo_chi":
+            _uuids = pseudo_chis[range_max: range_max + _num_rows] * int(row.frequency)
+            uuids.extend(_uuids)
+            range_max = range_max + _num_rows
+            continue
+
         _uuids = []
 
         for _ in range(_num_rows):
@@ -87,3 +105,36 @@ def generate_uuid_column(
     )
 
     return uuid_series
+
+def _generate_pseudo_chis(n, seed=0):
+    '''
+    Generate pseudo CHI numbers that consist of 10 digits, including
+    a possible zero as the first digit.
+
+    The logic of CHIs is not preserved to avoid potential collisions with the real CHIs.
+    In addition, the month part of the CHI is fixed at the impossible 13.
+
+    Parameters
+    ----------
+    n : int
+        the number of dummy CHIs to generate.
+
+    Returns
+    -------
+    A sorted list with unique dummy CHI numbers
+    '''
+
+    random.seed = seed
+    result = set()
+
+    while len(result) < n:
+        pseudo_chi = (
+            str(random.randint(0,31)) + # day will be zero padded if total length < 10
+            '13' +                      # ensure no accidental collissions
+            str(random.randint(20, 99)) +
+            str(random.randint(0,9999)).zfill(4) # no specific logic for 9th digit
+        ).zfill(10)
+
+        result.add(pseudo_chi)
+
+    return sorted(list(result))

exhibit/core/generate/yaml.py

@@ -126,7 +126,7 @@ def generate_YAML_string(spec_dict):
     #  miss_probability: 0.0
     #  anonymising_set: uuid
     #
-    # You can choose between uuid and range anonymising sets.
+    # You can choose between uuid, range and pseudo_chi anonymising sets.
     # ----------------------------------------------------------
     """)