feat: better errors for remote SDK (#1722)

* Adds nicer errors to remote SDK, that expose useful properties like
`request_id` and `status_code`.
* Makes sure the Python tracebacks print nicely by mapping the `source`
field from a Rust error to the `__cause__` field.

python/python/tests/test_remote_db.py

@@ -1,12 +1,14 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
+import contextlib
 import http.server
 import threading
 from unittest.mock import MagicMock
 import uuid
 
 import lancedb
+from lancedb.remote.errors import HttpError, RetryError
 import pyarrow as pa
 from lancedb.remote.client import VectorQuery, VectorQueryResult
 import pytest
@@ -98,6 +100,33 @@ def make_mock_http_handler(handler):
     return MockLanceDBHandler
 
 
+@contextlib.asynccontextmanager
+async def mock_lancedb_connection(handler):
+    with http.server.HTTPServer(
+        ("localhost", 8080), make_mock_http_handler(handler)
+    ) as server:
+        handle = threading.Thread(target=server.serve_forever)
+        handle.start()
+
+        db = await lancedb.connect_async(
+            "db://dev",
+            api_key="fake",
+            host_override="http://localhost:8080",
+            client_config={
+                "retry_config": {"retries": 2},
+                "timeout_config": {
+                    "connect_timeout": 1,
+                },
+            },
+        )
+
+        try:
+            yield db
+        finally:
+            server.shutdown()
+            handle.join()
+
+
 @pytest.mark.asyncio
 async def test_async_remote_db():
     def handler(request):
@@ -114,28 +143,50 @@ async def test_async_remote_db():
         request.end_headers()
         request.wfile.write(b'{"tables": []}')
 
-    def run_server():
-        with http.server.HTTPServer(
-            ("localhost", 8080), make_mock_http_handler(handler)
-        ) as server:
-            # we will only make one request
-            server.handle_request()
+    async with mock_lancedb_connection(handler) as db:
+        table_names = await db.table_names()
+        assert table_names == []
 
-    handle = threading.Thread(target=run_server)
-    handle.start()
 
-    db = await lancedb.connect_async(
-        "db://dev",
-        api_key="fake",
-        host_override="http://localhost:8080",
-        client_config={
-            "retry_config": {"retries": 2},
-            "timeout_config": {
-                "connect_timeout": 1,
-            },
-        },
-    )
-    table_names = await db.table_names()
-    assert table_names == []
+@pytest.mark.asyncio
+async def test_http_error():
+    request_id_holder = {"request_id": None}
 
-    handle.join()
+    def handler(request):
+        request_id_holder["request_id"] = request.headers["x-request-id"]
+
+        request.send_response(507)
+        request.end_headers()
+        request.wfile.write(b"Internal Server Error")
+
+    async with mock_lancedb_connection(handler) as db:
+        with pytest.raises(HttpError, match="Internal Server Error") as exc_info:
+            await db.table_names()
+
+        assert exc_info.value.request_id == request_id_holder["request_id"]
+        assert exc_info.value.status_code == 507
+
+
+@pytest.mark.asyncio
+async def test_retry_error():
+    request_id_holder = {"request_id": None}
+
+    def handler(request):
+        request_id_holder["request_id"] = request.headers["x-request-id"]
+
+        request.send_response(429)
+        request.end_headers()
+        request.wfile.write(b"Try again later")
+
+    async with mock_lancedb_connection(handler) as db:
+        with pytest.raises(RetryError, match="Hit retry limit") as exc_info:
+            await db.table_names()
+
+        assert exc_info.value.request_id == request_id_holder["request_id"]
+        assert exc_info.value.status_code == 429
+
+        cause = exc_info.value.__cause__
+        assert isinstance(cause, HttpError)
+        assert "Try again later" in str(cause)
+        assert cause.request_id == request_id_holder["request_id"]
+        assert cause.status_code == 429