Parse context propagation from the metadata

Summary: Parse the context propagation details or specs from the metadata

Reviewed By: fahndrich

Differential Revision: D64304874

fbshipit-source-id: 91cfb125aa72965202145f391e0818428240819c

sapp/analysis_output.py

@@ -11,7 +11,7 @@
 from dataclasses import dataclass
 from glob import glob
 from pathlib import Path
-from typing import Any, Dict, IO, Iterable, List, Optional, Set
+from typing import Any, Dict, IO, Iterable, List, Literal, Optional, Set
 
 from .sharded_files import ShardedFile
 
@@ -63,6 +63,28 @@ class PartialFlowToMark:
     feature: str
 
 
+@dataclass
+class ContextPropagation:
+    """
+    This is a specification of a context propagation that the user wishes us to use.
+    Context propagations are used to propagate context breadcrumbs from one frame
+    in an issue to another.
+
+    `code` represents the issue code that the context/breadcrumb is derived from.
+    `pattern` is the pattern that the breadcrumb should match.
+    `ignore_callee_port_and_location` is a boolean that indicates whether the callee
+    port and location should be ignored when finding frames to propagate the
+    context/breadcrumb to.
+    `frame_type` is a string that indicates the type of the frame the context/breadcrumb
+    will be derived from.
+    """
+
+    code: int
+    pattern: str
+    frame_type: Literal["precondition", "postcondition"]
+    ignore_callee_port_and_location: bool
+
+
 @dataclass
 class Metadata:
     # Used to relativize paths in the results
@@ -81,6 +103,9 @@ class Metadata:
     partial_flows_to_mark: List[PartialFlowToMark] = dataclasses.field(
         default_factory=list
     )
+    context_propagation_specs: List[ContextPropagation] = dataclasses.field(
+        default_factory=list
+    )
 
     def merge(self, o: "Metadata") -> "Metadata":
         return Metadata(
@@ -96,6 +121,8 @@ def merge(self, o: "Metadata") -> "Metadata":
             + o.class_type_intervals_filenames,
             category_coverage=self.category_coverage,  # should all be the same
             partial_flows_to_mark=self.partial_flows_to_mark + o.partial_flows_to_mark,
+            context_propagation_specs=self.context_propagation_specs
+            + o.context_propagation_specs,
         )
 
 
@@ -200,6 +227,9 @@ def from_directories(cls, directories: List[str]) -> "AnalysisOutput":
             partial_flows_to_mark = _parse_partial_flows_to_mark(
                 metadata, "partial_flows"
             )
+            context_propagation_specs = _parse_context_propagation_specs(
+                metadata, "context_propagation_specs"
+            )
             this_metadata = Metadata(
                 analysis_tool_version=metadata["version"],
                 commit_hash=metadata.get("commit"),
@@ -212,6 +242,7 @@ def from_directories(cls, directories: List[str]) -> "AnalysisOutput":
                 class_type_intervals_filenames=class_type_intervals_filenames,
                 category_coverage=metadata.get("category_coverage", []),
                 partial_flows_to_mark=partial_flows_to_mark,
+                context_propagation_specs=context_propagation_specs,
             )
             if not main_metadata:
                 main_metadata = this_metadata
@@ -256,6 +287,9 @@ def from_directory(cls, directory: str) -> "AnalysisOutput":
             metadata, "class_type_intervals_filename", directory
         )
         partial_flows_to_mark = _parse_partial_flows_to_mark(metadata, "partial_flows")
+        context_propagation_specs = _parse_context_propagation_specs(
+            metadata, "context_propagation_specs"
+        )
         return cls(
             directory=directory,
             filename_specs=filename_specs,
@@ -272,6 +306,7 @@ def from_directory(cls, directory: str) -> "AnalysisOutput":
                 class_type_intervals_filenames=class_type_intervals_filenames,
                 category_coverage=metadata.get("category_coverage", []),
                 partial_flows_to_mark=partial_flows_to_mark,
+                context_propagation_specs=context_propagation_specs,
             ),
         )
 
@@ -360,3 +395,29 @@ def _parse_partial_flows_to_mark(
             )
         )
     return parsed
+
+
+def _parse_context_propagation_specs(
+    metadata_json: Dict[str, Any], key: str
+) -> List[ContextPropagation]:
+    parsed: List[ContextPropagation] = []
+    context_propagation_specs = metadata_json.get(key, [])
+    for context_propagation_spec in context_propagation_specs:
+        code = context_propagation_spec["code"]
+        context_propagations = context_propagation_spec["context_propagations"]
+        for context_propagation in context_propagations:
+            frame_types = context_propagation["frame_types"]
+            pattern = context_propagation["pattern"]
+            ignore_callee_port_and_location = context_propagation[
+                "ignore_callee_port_and_location"
+            ]
+            for frame_type in frame_types:
+                parsed.append(
+                    ContextPropagation(
+                        code=code,
+                        pattern=pattern,
+                        frame_type=frame_type,
+                        ignore_callee_port_and_location=ignore_callee_port_and_location,
+                    )
+                )
+    return parsed

sapp/pipeline/propagate_context_to_leaf_frames.py

@@ -10,6 +10,7 @@
 from collections import defaultdict, deque
 from typing import Dict, List, Set, Tuple
 
+from ..analysis_output import ContextPropagation
 from ..models import IssueInstance, SharedTextKind, TraceFrame, TraceKind
 from ..trace_graph import TraceGraph
 from . import PipelineStep, Summary
@@ -33,26 +34,29 @@ class PropagateContextToLeafFrames(PipelineStep[TraceGraph, TraceGraph]):
 
     def __init__(
         self,
-        issue_code: int,
-        feature_pattern: str,
-        frame_kind: TraceKind,
-        ignore_callee_port: bool = False,
+        context_propagation: ContextPropagation,
     ) -> None:
         super().__init__()
         # pyre-fixme[13]: Attribute `summary` is never initialized.
         self.summary: Summary
         # pyre-fixme[13]: Attribute `graph` is never initialized.
         self.graph: TraceGraph
-        self.feature_pattern = feature_pattern
-        self.issue_code = issue_code
-        self.frame_kind = frame_kind
-        # pyre-fixme[8]: Expected `Dict[FrameID, TaintKindToState]` for 1st param but got `defaultdict`.
+        self.feature_pattern: str = context_propagation.pattern
+        self.issue_code: int = context_propagation.code
+        self.frame_kind: TraceKind = (
+            TraceKind.PRECONDITION
+            if context_propagation.frame_type == "precondition"
+            else TraceKind.POSTCONDITION
+        )
+        # pyre-fixme[8]: Expected `Dict[FrameID, TaintKindToState]` for param 1
         self.visited: Dict[FrameID, TaintKindToState] = defaultdict(
             lambda: defaultdict(lambda: PerTaintKindState())
         )
         self.leaf_features_added = 0
         self.leaf_frames = 0
-        self.ignore_callee_port = ignore_callee_port
+        self.ignore_callee_port_and_location: bool = (
+            context_propagation.ignore_callee_port_and_location
+        )
 
     def _subtract_kinds(
         self,
@@ -204,7 +208,7 @@ def _add_contextual_features_to_neighbor_frames(
                     candidate.callee_port != trace_frame.callee_port
                     or candidate.callee_location != trace_frame.callee_location
                 )
-                and (not self.ignore_callee_port)
+                and (not self.ignore_callee_port_and_location)
             ):
                 continue
 
@@ -217,7 +221,10 @@ def run(self, input: TraceGraph, summary: Summary) -> Tuple[TraceGraph, Summary]
         self.graph = graph
 
         log.info(
-            f"Propagating feature {self.feature_pattern} in issues {self.issue_code} to {self.frame_kind} leaves"
+            f"Propagating feature {self.feature_pattern} in issues"
+            + f" {self.issue_code} to {self.frame_kind} leaves."
+            + " Callee port and location will"
+            + f"{'' if self.ignore_callee_port_and_location else ' not'} be ignored"
         )
 
         for instance in graph.get_issue_instances():
@@ -238,7 +245,8 @@ def run(self, input: TraceGraph, summary: Summary) -> Tuple[TraceGraph, Summary]
                 self._add_contextual_features_to_frame(trace_frame, features)
                 self._add_contextual_features_to_neighbor_frames(trace_frame, features)
         log.info(
-            f"Added {self.leaf_features_added} features to {self.leaf_frames} trace frames"
+            f"Added {self.leaf_features_added} features to {self.leaf_frames}"
+            + " trace frames. "
         )
 
         return graph, summary