Change treatment of files that fail to load structured data

This new approach should allow us to cache the result of the failed
attempt to load structured data on a Lintable. The exception is stored
inside the object, so we can later produce detailed errors.

src/ansiblelint/_internal/rules.py

@@ -17,7 +17,7 @@
 LOAD_FAILURE_MD = """\
 # load-failure
 
-Linter failed to process a YAML file, probably because it is either:
+"Linter failed to process a file, possible invalid file. Possible reasons:
 
 * contains unsupported encoding (only UTF-8 is supported)
 * not an Ansible file
@@ -175,7 +175,7 @@ class LoadingFailureRule(BaseRule):
     """Failed to load or parse file."""
 
     id = "load-failure"
-    description = "Linter failed to process a YAML file, possible not an Ansible file."
+    description = "Linter failed to process a file, possible invalid file."
     severity = "VERY_HIGH"
     tags = ["core", "unskippable"]
     version_added = "v4.3.0"

src/ansiblelint/constants.py

@@ -1,5 +1,6 @@
 """Constants used by AnsibleLint."""
 import os.path
+from enum import Enum
 from typing import Literal
 
 DEFAULT_RULESDIR = os.path.join(os.path.dirname(__file__), "rules")
@@ -152,3 +153,15 @@ def main():
 #
 # https://github.com/ansible/ansible-lint-action/issues/138
 GIT_CMD = ["git", "-c", f"safe.directory={os.getcwd()}"]
+
+
+class States(Enum):
+    """States used are used as sentinel values in various places."""
+
+    NOT_LOADED = "File not loaded"
+    LOAD_FAILED = "File failed to load"
+    UNKNOWN_DATA = "Unknown data"
+
+    def __bool__(self) -> bool:
+        """Ensure all states evaluate as False as booleans."""
+        return False

src/ansiblelint/file_utils.py

@@ -14,12 +14,12 @@
 from tempfile import NamedTemporaryFile
 from typing import TYPE_CHECKING, Any, Iterator, cast
 
-# import wcmatch
 import wcmatch.pathlib
 from wcmatch.wcmatch import RECURSIVE, WcMatch
+from yaml.error import YAMLError
 
 from ansiblelint.config import BASE_KINDS, options
-from ansiblelint.constants import GIT_CMD, FileType
+from ansiblelint.constants import GIT_CMD, FileType, States
 
 if TYPE_CHECKING:
     # https://github.com/PyCQA/pylint/issues/3979
@@ -189,8 +189,9 @@ def __init__(
         self.dir: str = ""
         self.kind: FileType | None = None
         self.stop_processing = False  # Set to stop other rules from running
-        self._data: Any = None
+        self._data: Any = States.NOT_LOADED
         self.line_skips: dict[int, set[str]] = defaultdict(set)
+        self.exc: Exception | None = None  # Stores data loading exceptions
 
         if isinstance(name, str):
             name = Path(name)
@@ -341,21 +342,34 @@ def __repr__(self) -> str:
     @property
     def data(self) -> Any:
         """Return loaded data representation for current file, if possible."""
-        if not self._data:
-            if str(self.base_kind) == "text/yaml":
-                from ansiblelint.utils import (  # pylint: disable=import-outside-toplevel
-                    parse_yaml_linenumbers,
-                )
-
-                self._data = parse_yaml_linenumbers(self)
-                # Lazy import to avoid delays and cyclic-imports
-                if "append_skipped_rules" not in globals():
-                    # pylint: disable=import-outside-toplevel
-                    from ansiblelint.skip_utils import append_skipped_rules
-
-                self._data = append_skipped_rules(self._data, self)
-
-            # will remain None if we do not know how to load that file-type
+        if self._data == States.NOT_LOADED:
+            if self.path.is_dir():
+                self._data = None
+                return self._data
+            try:
+                if str(self.base_kind) == "text/yaml":
+                    from ansiblelint.utils import (  # pylint: disable=import-outside-toplevel
+                        parse_yaml_linenumbers,
+                    )
+
+                    self._data = parse_yaml_linenumbers(self)
+                    # Lazy import to avoid delays and cyclic-imports
+                    if "append_skipped_rules" not in globals():
+                        # pylint: disable=import-outside-toplevel
+                        from ansiblelint.skip_utils import append_skipped_rules
+
+                    self._data = append_skipped_rules(self._data, self)
+                else:
+                    logging.debug(
+                        "data set to None for %s due to being of %s kind.",
+                        self.path,
+                        self.base_kind,
+                    )
+                    self._data = States.UNKNOWN_DATA
+
+            except (RuntimeError, FileNotFoundError, YAMLError) as exc:
+                self._data = None
+                self.exc = exc
         return self._data
 
 

src/ansiblelint/runner.py

@@ -12,6 +12,7 @@
 import ansiblelint.skip_utils
 import ansiblelint.utils
 from ansiblelint._internal.rules import LoadingFailureRule
+from ansiblelint.constants import States
 from ansiblelint.errors import MatchError
 from ansiblelint.file_utils import Lintable, expand_dirs_in_lintables
 from ansiblelint.rules.syntax_check import AnsibleSyntaxCheckRule
@@ -122,14 +123,12 @@ def run(self) -> list[MatchError]:  # noqa: C901
                 _logger.debug("Excluded %s", lintable)
                 self.lintables.remove(lintable)
                 continue
-            try:
-                lintable.data
-            except (RuntimeError, FileNotFoundError) as exc:
+            if isinstance(lintable.data, States) and lintable.exc:
                 matches.append(
                     MatchError(
                         filename=lintable,
-                        message=str(exc),
-                        details=str(exc.__cause__),
+                        message=str(lintable.exc),
+                        details=str(lintable.exc.__cause__),
                         rule=LoadingFailureRule(),
                     )
                 )