Extract the logic to identify bug types into individual feature classes (#3907)

bugbug/bug_features.py

@@ -16,7 +16,7 @@ from libmozdata import versions
 from libmozdata.bugzilla import Bugzilla
 from sklearn.base import BaseEstimator, TransformerMixin
 
-from bugbug import bug_snapshot, repository
+from bugbug import bug_snapshot, bugzilla, repository
 
 
 def field(bug, field):
@@ -687,3 +687,192 @@ class BugExtractor(BaseEstimator, TransformerMixin):
             bugs_iter = apply_rollback(bugs_iter)
 
         return pd.DataFrame(apply_transform(bug) for bug in bugs_iter)
+
+
+class IsPerformanceBug(SingleBugFeature):
+    """Determine if the bug is related to performance based on given bug data."""
+
+    name = "Is Performance Bug"
+    type_name = "performance"
+    keyword_prefixes = ("perf", "topperf", "main-thread-io")
+    whiteboard_prefixes = (
+        "[fxperf",
+        "[fxperfsize",
+        "[snappy",
+        "[pdfjs-c-performance",
+        "[pdfjs-performance",
+        "[sp3",
+    )
+
+    def __call__(
+        self,
+        bug: bugzilla.BugDict,
+        bug_map: dict[int, bugzilla.BugDict] | None = None,
+    ) -> bool:
+        if bug.get("cf_performance_impact") in ("low", "medium", "high"):
+            return True
+
+        if any(
+            keyword.startswith(prefix)
+            for keyword in bug["keywords"]
+            for prefix in self.keyword_prefixes
+        ):
+            return True
+
+        bug_whiteboard = bug["whiteboard"].lower()
+        if any(prefix in bug_whiteboard for prefix in self.whiteboard_prefixes):
+            return True
+
+        return False
+
+
+class IsMemoryBug(SingleBugFeature):
+    """Determine if the bug is related to memory based on given bug data."""
+
+    name = "Is Memory Bug"
+    type_name = "memory"
+    keyword_prefixes = ("memory-",)
+    whiteboard_prefixes = ("[overhead", "[memshrink")
+
+    def __call__(
+        self,
+        bug: bugzilla.BugDict,
+        bug_map: dict[int, bugzilla.BugDict] | None = None,
+    ) -> bool:
+        if bug_map is not None:
+            for bug_id in bug["blocks"]:
+                if bug_id not in bug_map:
+                    continue
+
+                alias = bug_map[bug_id]["alias"]
+                if alias and alias.startswith("memshrink"):
+                    return True
+
+        if any(
+            keyword.startswith(prefix)
+            for keyword in bug["keywords"]
+            for prefix in self.keyword_prefixes
+        ):
+            return True
+
+        bug_whiteboard = bug["whiteboard"].lower()
+        if any(prefix in bug_whiteboard for prefix in self.whiteboard_prefixes):
+            return True
+
+        return False
+
+
+class IsPowerBug(SingleBugFeature):
+    """Determine if the bug is related to power based on given bug data."""
+
+    name = "Is Power Bug"
+    type_name = "power"
+    keyword_prefixes = ("power",)
+    whiteboard_prefixes = ("[power",)
+
+    def __call__(
+        self,
+        bug: bugzilla.BugDict,
+        bug_map: dict[int, bugzilla.BugDict] | None = None,
+    ) -> bool:
+        if any(
+            keyword.startswith(prefix)
+            for keyword in bug["keywords"]
+            for prefix in self.keyword_prefixes
+        ):
+            return True
+
+        bug_whiteboard = bug["whiteboard"].lower()
+        if any(prefix in bug_whiteboard for prefix in self.whiteboard_prefixes):
+            return True
+
+        return False
+
+
+class IsSecurityBug(SingleBugFeature):
+    """Determine if the bug is related to security based on given bug data."""
+
+    name = "Is Security Bug"
+    type_name = "security"
+    keyword_prefixes = ("sec-", "csectype-")
+    whiteboard_prefixes = ("[client-bounty-form", "[sec-survey")
+
+    def __call__(
+        self,
+        bug: bugzilla.BugDict,
+        bug_map: dict[int, bugzilla.BugDict] | None = None,
+    ) -> bool:
+        if any(
+            keyword.startswith(prefix)
+            for keyword in bug["keywords"]
+            for prefix in self.keyword_prefixes
+        ):
+            return True
+
+        bug_whiteboard = bug["whiteboard"].lower()
+        if any(prefix in bug_whiteboard for prefix in self.whiteboard_prefixes):
+            return True
+
+        return False
+
+
+class IsCrashBug(SingleBugFeature):
+    """Determine if the bug is related to crash based on given bug data."""
+
+    name = "Is Crash Bug"
+    type_name = "crash"
+    keyword_prefixes = ("crash", "crashreportid")
+
+    def __call__(
+        self,
+        bug: bugzilla.BugDict,
+        bug_map: dict[int, bugzilla.BugDict] | None = None,
+    ) -> bool:
+        # Checking for `[@` will exclude some bugs that do not have valid
+        # signatures: https://mzl.la/46XAqRF
+        if bug.get("cf_crash_signature") and "[@" in bug["cf_crash_signature"]:
+            return True
+
+        if any(
+            keyword.startswith(prefix)
+            for keyword in bug["keywords"]
+            for prefix in self.keyword_prefixes
+        ):
+            return True
+
+        return False
+
+
+class BugTypes(SingleBugFeature):
+    """Determine bug type."""
+
+    name = "Infer Bug Type"
+    bug_type_extractors: list = [
+        IsCrashBug(),
+        IsMemoryBug(),
+        IsPerformanceBug(),
+        IsPowerBug(),
+        IsSecurityBug(),
+    ]
+
+    def __call__(
+        self,
+        bug: bugzilla.BugDict,
+        bug_map: dict[int, bugzilla.BugDict] | None = None,
+    ) -> list[str]:
+        """Infer bug types based on various bug characteristics.
+
+        Args:
+        - bug (bugzilla.BugDict): A dictionary containing bug data.
+        - bug_map (Optional[dict[int, bugzilla.BugDict]]): A mapping
+            of bug IDs to bug dictionaries. Default is None.
+
+        Returns:
+        - list[str]: A list of inferred bug types (e.g., "memory", "power",
+            "performance", "security", "crash").
+        """
+        return [
+            is_type.type_name
+            for is_type in self.bug_type_extractors
+            if is_type(bug, bug_map)
+        ]

bugbug/models/bugtype.py

@@ -18,79 +18,6 @@ from bugbug.model import BugModel
 
 logger = logging.getLogger(__name__)
 
-KEYWORD_DICT = {
-    "sec-": "security",
-    "csectype-": "security",
-    "memory-": "memory",
-    "crash": "crash",
-    "crashreportid": "crash",
-    "perf": "performance",
-    "topperf": "performance",
-    "main-thread-io": "performance",
-    "power": "power",
-}
-TYPE_LIST = sorted(set(KEYWORD_DICT.values()))
-
-
-def bug_to_types(
-    bug: bugzilla.BugDict, bug_map: dict[int, bugzilla.BugDict] | None = None
-) -> list[str]:
-    types = set()
-
-    bug_whiteboard = bug["whiteboard"].lower()
-
-    if any(
-        f"{whiteboard_text}" in bug_whiteboard
-        for whiteboard_text in ("overhead", "memshrink")
-    ):
-        types.add("memory")
-
-    if "[power" in bug_whiteboard:
-        types.add("power")
-
-    if any(
-        f"[{whiteboard_text}" in bug_whiteboard
-        for whiteboard_text in (
-            "fxperf",
-            "fxperfsize",
-            "snappy",
-            "pdfjs-c-performance",
-            "pdfjs-performance",
-            "sp3",
-        )
-    ):
-        types.add("performance")
-
-    if any(
-        f"[{whiteboard_text}" in bug_whiteboard
-        for whiteboard_text in ("client-bounty-form", "sec-survey")
-    ):
-        types.add("security")
-
-    if "cf_performance_impact" in bug and bug["cf_performance_impact"] not in (
-        "---",
-        "?",
-    ):
-        types.add("performance")
-
-    if "cf_crash_signature" in bug and bug["cf_crash_signature"] not in ("", "---"):
-        types.add("crash")
-
-    if bug_map is not None:
-        for bug_id in bug["blocks"]:
-            if bug_id not in bug_map:
-                continue
-
-            alias = bug_map[bug_id]["alias"]
-            if alias and alias.startswith("memshrink"):
-                types.add("memory")
-
-    for keyword_start, type in KEYWORD_DICT.items():
-        if any(keyword.startswith(keyword_start) for keyword in bug["keywords"]):
-            types.add(type)
-
-    return list(types)
-
 
 class BugTypeModel(BugModel):
     def __init__(self, lemmatization=False, historical=False):
@@ -98,12 +25,20 @@ class BugTypeModel(BugModel):
 
         self.calculate_importance = False
 
+        self.bug_type_extractors = bug_features.BugTypes.bug_type_extractors
+
+        label_keyword_prefixes = {
+            keyword
+            for extractor in self.bug_type_extractors
+            for keyword in extractor.keyword_prefixes
+        }
+
         feature_extractors = [
             bug_features.HasSTR(),
             bug_features.Severity(),
             # Ignore keywords that would make the ML completely skewed
             # (we are going to use them as 100% rules in the evaluation phase).
-            bug_features.Keywords(set(KEYWORD_DICT.keys())),
+            bug_features.Keywords(label_keyword_prefixes),
             bug_features.IsCoverityIssue(),
             bug_features.HasCrashSignature(),
             bug_features.HasURL(),
@@ -170,20 +105,23 @@ class BugTypeModel(BugModel):
         bug_map = {bug["id"]: bug for bug in bugzilla.get_bugs()}
 
         for bug_data in bug_map.values():
-            target = np.zeros(len(TYPE_LIST))
-            for type_ in bug_to_types(bug_data, bug_map):
-                target[TYPE_LIST.index(type_)] = 1
+            target = np.zeros(len(self.bug_type_extractors))
+            for i, is_type in enumerate(self.bug_type_extractors):
+                if is_type(bug_data, bug_map):
+                    target[i] = 1
 
             classes[int(bug_data["id"])] = target
 
-        for type_ in TYPE_LIST:
+        bug_types = [extractor.type_name for extractor in self.bug_type_extractors]
+
+        for i, bug_type in enumerate(bug_types):
             logger.info(
                 "%d %s bugs",
-                sum(target[TYPE_LIST.index(type_)] == 1 for target in classes.values()),
-                type_,
+                sum(target[i] for target in classes.values()),
+                bug_type,
             )
 
-        return classes, TYPE_LIST
+        return classes, bug_types
 
     def get_feature_names(self):
         return self.clf.named_steps["union"].get_feature_names_out()
@@ -194,11 +132,14 @@ class BugTypeModel(BugModel):
         classes: dict[int, np.ndarray],
         probabilities: bool,
     ):
+        bug_map = {bug["id"]: bug for bug in bugs}
+
         for i, bug in enumerate(bugs):
-            for type_ in bug_to_types(bug):
-                if probabilities:
-                    classes[i][TYPE_LIST.index(type_)] = 1.0
-                else:
-                    classes[i][TYPE_LIST.index(type_)] = 1
+            for j, is_type_applicable in enumerate(self.bug_type_extractors):
+                if is_type_applicable(bug, bug_map):
+                    if probabilities:
+                        classes[i][j] = 1.0
+                    else:
+                        classes[i][j] = 1
 
         return classes

scripts/generate_landings_risk_report.py

@@ -30,7 +30,6 @@ from dateutil.relativedelta import relativedelta
 from tqdm import tqdm
 
 from bugbug import bug_features, bugzilla, db, phabricator, repository, test_scheduling
-from bugbug.models.bugtype import bug_to_types
 from bugbug.models.regressor import BUG_FIXING_COMMITS_DB, RegressorModel
 from bugbug.utils import (
     download_check_etag,
@@ -555,6 +554,8 @@ class LandingsRiskReportGenerator(object):
 
         component_team_mapping = get_component_team_mapping()
 
+        bug_to_types = bug_features.BugTypes()
+
         bug_summaries = []
         for bug_id in bugs:
             if bug_id not in bug_map:

tests/fixtures/bug_features/bug_types.json

@@ -0,0 +1,5 @@
+{"keywords": ["meta", "perf"], "whiteboard": "", "cf_crash_signature": ""}
+{"keywords": ["memory-leak", "regression"], "whiteboard": "[MemShrink:P1]", "cf_crash_signature": ""}
+{"whiteboard": "", "keywords": ["power"]}
+{"keywords": ["sec-want"], "whiteboard": "[sg:want][psm-padlock]"}
+{"keywords": ["crash", "regression"], "whiteboard": "", "cf_crash_signature": "[@ audiounit_property_listener_callback]"}

tests/test_bug_features.py

@@ -12,6 +12,7 @@ from bugbug.bug_features import (
     BlockedBugsNumber,
     BugExtractor,
     BugReporter,
+    BugTypes,
     CommentCount,
     CommentLength,
     Component,
@@ -178,3 +179,11 @@ def test_BugExtractor():
         BugExtractor([HasSTR(), HasSTR()], [fileref(), url()])
     with pytest.raises(AssertionError):
         BugExtractor([HasSTR(), HasURL()], [fileref(), fileref()])
+
+
+def test_BugTypes(read) -> None:
+    read(
+        "bug_types.json",
+        BugTypes,
+        [["performance"], ["memory"], ["power"], ["security"], ["crash"]],
+    )