fix: release the id increment lock after generation (#87)

- the success path of DataAdapter._id leaked the RLock acquisition, deadlocking inserts from background threads (scheduler jobs) on non Mongo adapters
- regression test asserts the lock is acquirable from another thread

Author: joamag

CHANGELOG.md

@@ -17,7 +17,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
-*
+* Identifier generation (`DataAdapter._id`) no longer leaks the increment lock, which deadlocked inserts from background threads (e.g. scheduler jobs over the tiny adapter) - [#86](https://github.com/hivesolutions/appier/issues/86)
 
 ## [1.46.0] - 2026-05-31
 

src/appier/data.py

@@ -105,7 +105,7 @@ def _id(self):
         try:
             token += struct.pack(">i", self._inc)[1:4]
             self._inc = (self._inc + 1) % 0xFFFFFF
-        except Exception:
+        finally:
             self._inc_lock.release()
         token_s = binascii.hexlify(token)
         token_s = legacy.str(token_s)

src/appier/test/data.py

@@ -30,6 +30,7 @@
 
 import os
 import tempfile
+import threading
 import unittest
 
 import appier
@@ -43,6 +44,24 @@ def test_id(self):
         self.assertEqual(type(identifier), str)
         self.assertEqual(len(identifier), 24)
 
+    def test_id_lock_release(self):
+        adapter = appier.DataAdapter()
+        adapter._id()
+
+        result = []
+
+        def acquire():
+            acquired = adapter._inc_lock.acquire(False)
+            if acquired:
+                adapter._inc_lock.release()
+            result.append(acquired)
+
+        thread = threading.Thread(target=acquire)
+        thread.start()
+        thread.join()
+
+        self.assertEqual(result, [True])
+
     def test_drop_db_missing(self):
         fd, file_path = tempfile.mkstemp()
         os.close(fd)