Merge pull request #335 from nateprewitt/cache_race_condition

Fix concurrency issues with cache

Author: jamesls

.github/workflows/run-tests.yml

@@ -9,7 +9,7 @@ jobs:
     strategy:
       matrix:
         os: [ubuntu-latest]
-        python-version: ["3.7", "3.8", "3.9", "3.10", "3.11-dev"]
+        python-version: ["3.9", "3.10", "3.11"]
 
     steps:
     - uses: actions/checkout@v2
@@ -25,4 +25,4 @@ jobs:
         pip install dist/*.whl
     - name: Test with pytest
       run: |
-        cd tests/ && py.test --cov jmespath --cov-report term-missing
+        cd tests/ && python -m pytest --cov jmespath --cov-report term-missing

CHANGELOG.rst

@@ -1,3 +1,10 @@
+Next Release (TBD)
+==================
+
+* Fix concurrency issue with cache
+  (`pr #335 <https://github.com/jmespath/jmespath.py/pull/335>`__)
+
+
 1.0.1
 =====
 

jmespath/parser.py

@@ -25,8 +25,6 @@
   consuming from the token iterator one token at a time.
 
 """
-import random
-
 from jmespath import lexer
 from jmespath.compat import with_repr_method
 from jmespath import ast
@@ -73,7 +71,7 @@ class Parser(object):
     # The _MAX_SIZE most recent expressions are cached in
     # _CACHE dict.
     _CACHE = {}
-    _MAX_SIZE = 128
+    _MAX_SIZE = 512
 
     def __init__(self, lookahead=2):
         self.tokenizer = None
@@ -82,13 +80,26 @@ def __init__(self, lookahead=2):
         self._index = 0
 
     def parse(self, expression):
-        cached = self._CACHE.get(expression)
-        if cached is not None:
-            return cached
+        try:
+            return self._CACHE[expression]
+        except KeyError:
+            pass
         parsed_result = self._do_parse(expression)
+        if len(self._CACHE) >= self._MAX_SIZE:
+            try:
+                del self._CACHE[next(iter(self._CACHE))]
+            except (KeyError, StopIteration, RuntimeError):
+                # KeyError - Another thread else already deleted the key.
+                # RuntimeError - Another modified the cache.
+                # StopIteration - (Unlikely) Cache is empty.
+                #
+                # If we encounter an error we should NOT be adding to the
+                # cache.  To ensure we do not exceed self._MAX_SIZE, we
+                # can only add to the cache if we successfully removed
+                # an element from the cache, otherwise this can grow
+                # unbounded.
+                return parsed_result
         self._CACHE[expression] = parsed_result
-        if len(self._CACHE) > self._MAX_SIZE:
-            self._free_cache_entries()
         return parsed_result
 
     def _do_parse(self, expression):
@@ -488,10 +499,6 @@ def _raise_parse_error_maybe_eof(self, expected_type, token):
         raise exceptions.ParseError(
             lex_position, actual_value, actual_type, message)
 
-    def _free_cache_entries(self):
-        for key in random.sample(list(self._CACHE.keys()), int(self._MAX_SIZE / 2)):
-            self._CACHE.pop(key, None)
-
     @classmethod
     def purge(cls):
         """Clear the expression compilation cache."""