Enable the on-disk topology cache by default (match the .NET client)

The .NET client enables the topology cache by default (DisableTopologyCache=false,
TopologyCacheLocation=AppContext.BaseDirectory). Align the Python port so it has the
same purpose/behaviour, not just the same mechanism:

- Add the disable_topology_cache convention (default False), mirroring .NET DisableTopologyCache.
- Default topology_cache_location to os.getcwd() (the closest zero-dependency analog of
  .NET's AppContext.BaseDirectory).
- RequestExecutor gates both the write and the seed-from-cache on `not disable_topology_cache`.

Every client now persists topology and, when its initial urls are unreachable on startup,
seeds the last-known topology from disk. TestBase.tearDown already removes *topology files
from the working directory, so the default-on behaviour leaves no test artifacts (verified
across a broad sample; 0 files left behind).

Author: poissoncorp

ravendb/documents/conventions.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import inspect
+import os
 import threading
 from abc import abstractmethod, ABC
 from datetime import timedelta, datetime
@@ -51,8 +52,10 @@ def __init__(self):
 
         # Flags
         self.disable_topology_updates = False
-        # Opt-in on-disk topology cache: directory to persist/read topology files, or None to disable.
-        self.topology_cache_location: Optional[str] = None
+        # On-disk topology cache (mirrors the .NET client): enabled by default. Topology is persisted to, and
+        # - when the initial urls are unreachable on startup - seeded from topology_cache_location.
+        self.disable_topology_cache = False
+        self.topology_cache_location: Optional[str] = os.getcwd()
         self._optimistic_concurrency_mode = None
         # Track which setter the user touched so we can reject mixing them.
         self._use_optimistic_concurrency_was_set = False

ravendb/http/request_executor.py

@@ -417,7 +417,7 @@ def __supply_async():
 
                 self._topology_etag = self._node_selector.topology.etag
 
-                if self.conventions.topology_cache_location:
+                if not self.conventions.disable_topology_cache and self.conventions.topology_cache_location:
                     topology_local_cache.try_save(
                         self.conventions.topology_cache_location,
                         topology_local_cache.server_hash(parameters.node.url, self._database_name),
@@ -497,7 +497,7 @@ def __run(errors: list):
 
     def _try_load_topology_from_cache(self, url):
         return topology_local_cache.try_load(
-            self.conventions.topology_cache_location,
+            None if self.conventions.disable_topology_cache else self.conventions.topology_cache_location,
             topology_local_cache.server_hash(url, self._database_name),
             topology_local_cache.DATABASE_TOPOLOGY_EXTENSION,
         )
@@ -1471,7 +1471,7 @@ def __supply_async():
                 new_topology = Topology(results.etag, nodes)
                 self._topology_etag = results.etag
 
-                if self.conventions.topology_cache_location:
+                if not self.conventions.disable_topology_cache and self.conventions.topology_cache_location:
                     topology_local_cache.try_save(
                         self.conventions.topology_cache_location,
                         topology_local_cache.server_hash(parameters.node.url),
@@ -1505,7 +1505,7 @@ def __supply_async():
 
     def _try_load_topology_from_cache(self, url):
         return topology_local_cache.try_load(
-            self.conventions.topology_cache_location,
+            None if self.conventions.disable_topology_cache else self.conventions.topology_cache_location,
             topology_local_cache.server_hash(url),
             topology_local_cache.CLUSTER_TOPOLOGY_EXTENSION,
         )