Add caching to bottlenecks in the message store (#5605)

Some functions can't be cached without impacting the correctness
with the current design.

Co-authored-by: Daniël van Noord <[email protected]>

Author: Pierre-Sassoulas

pylint/lint/pylinter.py

@@ -1389,7 +1389,11 @@ def _get_message_state_scope(
         return None
 
     def _is_one_message_enabled(self, msgid: str, line: Optional[int]) -> bool:
-        """Checks state of a single message"""
+        """Checks state of a single message for the current file
+
+        This function can't be cached as it depends on self.file_state which can
+        change.
+        """
         if line is None:
             return self._msgs_state.get(msgid, True)
         try:
@@ -1426,10 +1430,15 @@ def is_message_enabled(
         line: Optional[int] = None,
         confidence: Optional[interfaces.Confidence] = None,
     ) -> bool:
-        """return whether the message associated to the given message id is
-        enabled
+        """Return whether this message is enabled for the current file, line and confidence level.
+
+        This function can't be cached right now as the line is the line of
+        the currently analysed file (self.file_state), if it changes, then the
+        result for the same msg_descr/line might need to change.
 
-        msgid may be either a numeric or symbolic message id.
+        :param msg_descr: Either the msgid or the symbol for a MessageDefinition
+        :param line: The line of the currently analysed file
+        :param confidence: The confidence of the message
         """
         if self.config.confidence and confidence:
             if confidence.name not in self.config.confidence:

pylint/message/message_definition_store.py

@@ -2,6 +2,7 @@
 # For details: https://github.com/PyCQA/pylint/blob/main/LICENSE
 
 import collections
+import functools
 from typing import TYPE_CHECKING, Dict, List, Tuple, ValuesView
 
 from pylint.exceptions import UnknownMessageError
@@ -46,8 +47,14 @@ def register_message(self, message: MessageDefinition) -> None:
         self._messages_definitions[message.msgid] = message
         self._msgs_by_category[message.msgid[0]].append(message.msgid)
 
+    @functools.lru_cache()
     def get_message_definitions(self, msgid_or_symbol: str) -> List[MessageDefinition]:
-        """Returns the Message definition for either a numeric or symbolic id."""
+        """Returns the Message definition for either a numeric or symbolic id.
+
+        The cache has no limit as its size will likely stay minimal. For each message we store
+        about 1000 characters, so even if we would have 1000 messages the cache would only
+        take up ~= 1 Mb.
+        """
         return [
             self._messages_definitions[m]
             for m in self.message_id_store.get_active_msgids(msgid_or_symbol)

pylint/message/message_id_store.py

@@ -1,5 +1,6 @@
 # Licensed under the GPL: https://www.gnu.org/licenses/old-licenses/gpl-2.0.html
 # For details: https://github.com/PyCQA/pylint/blob/main/LICENSE
+import functools
 from typing import Dict, List, NoReturn, Optional, Tuple
 
 from pylint.exceptions import InvalidMessageError, UnknownMessageError
@@ -101,18 +102,23 @@ def _raise_duplicate_msgid(symbol: str, msgid: str, other_msgid: str) -> NoRetur
         )
         raise InvalidMessageError(error_message)
 
+    @functools.lru_cache()
     def get_active_msgids(self, msgid_or_symbol: str) -> List[str]:
-        """Return msgids but the input can be a symbol."""
-        # Only msgid can have a digit as second letter
-        is_msgid: bool = msgid_or_symbol[1:].isdigit()
-        msgid = None
-        if is_msgid:
+        """Return msgids but the input can be a symbol.
+
+        The cache has no limit as its size will likely stay minimal. For each message we store
+        about 1000 characters, so even if we would have 1000 messages the cache would only
+        take up ~= 1 Mb.
+        """
+        msgid: Optional[str]
+        if msgid_or_symbol[1:].isdigit():
+            # Only msgid can have a digit as second letter
             msgid = msgid_or_symbol.upper()
             symbol = self.__msgid_to_symbol.get(msgid)
         else:
             msgid = self.__symbol_to_msgid.get(msgid_or_symbol)
             symbol = msgid_or_symbol
-        if msgid is None or symbol is None or not msgid or not symbol:
+        if not msgid or not symbol:
             error_msg = f"No such message id or symbol '{msgid_or_symbol}'."
             raise UnknownMessageError(error_msg)
         return self.__old_names.get(msgid, [msgid])