Sort studies by name and add basic study intersection check. (#1023)

* Sort studies by name and add basic study intersection validation.

* Fix wildcard version comparison, add tests.

* Make sure to properly cover empty channel/platform lists (although this is not allowed).

seed/serialize.py

@@ -12,6 +12,7 @@
 import time
 import proto.variations_seed_pb2 as variations_seed_pb2
 import argparse
+import collections
 
 SEED_BIN_PATH = "./seed.bin"
 SERIALNUMBER_PATH = "./serialnumber"
@@ -26,6 +27,73 @@
     'RELEASE': study_pb2.Study.Channel.STABLE
 }
 
+def version_to_int_array(version_str):
+    version_list = []
+    if version_str is None:
+        return version_list
+
+    parts = version_str.split('.')
+    for part in parts:
+        if part == '*':
+            version_list.append(part)
+            break
+        version_list.append(int(part))
+
+    return version_list
+
+def compare_versions(version1, version2):
+    min_len = None
+    if not version1:
+        min_len = 0
+    elif version1[-1] == '*':
+        version1 = version1[:-1]
+        min_len = len(version1)
+
+    if not version2:
+        min_len = 0
+    elif version2[-1] == '*':
+        version2 = version2[:-1]
+        if min_len is not None:
+            min_len = min(min_len, len(version2))
+        else:
+            min_len = len(version2)
+
+    if min_len is not None:
+        version1 = version1[:min_len]
+        version2 = version2[:min_len]
+
+    if version1 > version2:
+        return 1
+    elif version1 < version2:
+        return -1
+    else:
+        return 0
+
+def test_version_comparison():
+    # //base/version_unittest.cc VersionTest.CompareToWildcardString
+    test_cases = [
+        ["1.0", "1.*", 0],
+        ["1.0", "0.*", 1],
+        ["1.0", "2.*", -1],
+        ["1.2.3", "1.2.3.*", 0],
+        ["10.0", "1.0.*", 1],
+        ["1.0", "3.0.*", -1],
+        ["1.4", "1.3.0.*", 1],
+        ["1.3.9", "1.3.*", 0],
+        ["1.4.1", "1.3.*", 1],
+        ["1.3", "1.4.5.*", -1],
+        ["1.5", "1.4.5.*", 1],
+        ["1.3.9", "1.3.*", 0],
+        ["1.2.0.0.0.0", "1.2.*", 0],
+        [None, None, 0],
+        [None, "1", 0],
+        ["1", None, 0],
+    ]
+    for test_case in test_cases:
+        version1 = version_to_int_array(test_case[0])
+        version2 = version_to_int_array(test_case[1])
+        assert compare_versions(version1, version2) == test_case[2]
+
 def validate(seed):
     for study in seed['studies']:
         total_proba = 0
@@ -44,6 +112,57 @@ def validate(seed):
             print("platform not in ", PLATFORMS)
             return False
 
+    feature_names_to_studies = collections.defaultdict(list)
+    for study in seed['studies']:
+        used_feature_names = set()
+        for experiment in study['experiments']:
+            feature_association = experiment.get('feature_association')
+            if feature_association:
+                for enable_feature in feature_association.get('enable_feature', []):
+                    used_feature_names.add(enable_feature)
+                for disable_feature in feature_association.get('disable_feature', []):
+                    used_feature_names.add(disable_feature)
+
+        for used_feature_names in used_feature_names:
+            feature_names_to_studies[used_feature_names].append(study)
+
+    def get_study_platforms(study):
+        return set(study.get('filter', {}).get('platform', []))
+
+    def get_study_channels(study):
+        return set(study.get('filter', {}).get('channel', []))
+
+    def get_study_version_range(study):
+        return [
+            version_to_int_array(study.get('filter', {}).get('min_version')),
+            version_to_int_array(study.get('filter', {}).get('max_version')),
+        ]
+
+    def is_filter_set_intersect(a, b):
+        return not a or not b or a.intersection(b)
+
+    def is_version_range_intersect(range1, range2):
+        return compare_versions(range1[1], range2[0]) >= 0 and compare_versions(range2[1], range1[0]) >= 0
+
+    test_version_comparison()
+    for studies in feature_names_to_studies.values():
+        for i, study1 in enumerate(studies):
+            study1_platform = get_study_platforms(study1)
+            study1_channel = get_study_channels(study1)
+            study1_version_range = get_study_version_range(study1)
+            for j in range(i + 1, len(studies)):
+                study2 = studies[j]
+                study2_platform = get_study_platforms(study2)
+                study2_channel = get_study_channels(study2)
+                study2_version_range = get_study_version_range(study2)
+                # Check if the studies overlap in platform
+                if is_filter_set_intersect(study1_platform, study2_platform):
+                    # Check if the studies overlap in channel
+                    if is_filter_set_intersect(study1_channel, study2_channel):
+                        # Check if the studies overlap in version
+                        if is_version_range_intersect(study1_version_range, study2_version_range):
+                            raise ValueError(f"Studies overlap:\n{json.dumps(study1, indent=2)}\n\n{json.dumps(study2, indent=2)}")
+
     return True
 
 
@@ -213,6 +332,7 @@ def main():
 
     print("Load", args.seed_path.name)
     seed_data = json.load(args.seed_path)
+    seed_data['studies'].sort(key=lambda study: study['name'])
 
     print("Validate seed data")
     if not validate(seed_data):