Merge pull request #8 from pinecone-io/experimental/index-vector-transform-remove-type-check

pinecone.Index vector transform remove type check

Author: benjaminran

CHANGELOG.md

@@ -1,5 +1,11 @@
 # Changelog
 
+## Unreleased Changes
+
+### Changed
+
+- Some type validations were moved to the backend for performance reasons. In these cases a 4xx ApiException will be returned instead of an ApiTypeError.
+
 ## [2.0.2] - 2021-10-21
 
 ### Changed

README.md

@@ -5,6 +5,19 @@ For more information, see the docs at https://www.pinecone.io/docs/
 
 ## Installation
 
+Install a released version from pip:
+```shell
+pip install pinecone-client
+```
+
+Or the latest development version:
+```shell
+pip install git+https://git@github.com/pinecone-io/pinecone-python-client.git
+```
+
+Or a specific development version:
 ```shell
 pip install git+https://git@github.com/pinecone-io/pinecone-python-client.git
+pip install git+https://git@github.com/pinecone-io/pinecone-python-client.git@example-branch-name
+pip install git+https://git@github.com/pinecone-io/pinecone-python-client.git@259deff
 ```

pinecone/__version__

@@ -1 +1 @@
-2.0.2
+2.0.2.a

pinecone/index.py

@@ -48,17 +48,20 @@ def __init__(self, index_name: str, pool_threads=1):
     @sentry
     @validate_and_convert_errors
     def upsert(self, vectors, **kwargs):
+        _check_type = kwargs.pop('_check_type', False)
+
         def _vector_transform(item):
             if isinstance(item, Vector):
                 return item
             if isinstance(item, tuple):
                 id, values, metadata = fix_tuple_length(item, 3)
-                return Vector(id=id, values=values, metadata=metadata or {})
+                return Vector(id=id, values=values, metadata=metadata or {}, _check_type=_check_type)
             raise ValueError(f"Invalid vector value passed: cannot interpret type {type(item)}")
 
         return self._vector_api.upsert(
             UpsertRequest(
                 vectors=list(map(_vector_transform, vectors)),
+                _check_type=_check_type,
                 **{k: v for k, v in kwargs.items() if k not in _OPENAPI_ENDPOINT_PARAMS}
             ),
             **{k: v for k, v in kwargs.items() if k in _OPENAPI_ENDPOINT_PARAMS}
@@ -77,19 +80,22 @@ def fetch(self, *args, **kwargs):
     @sentry
     @validate_and_convert_errors
     def query(self, queries, **kwargs):
+        _check_type = kwargs.pop('_check_type', False)
+
         def _query_transform(item):
             if isinstance(item, QueryVector):
                 return item
             if isinstance(item, tuple):
                 values, filter = fix_tuple_length(item, 2)
-                return QueryVector(values=values, filter=filter)
+                return QueryVector(values=values, filter=filter, _check_type=_check_type)
             if isinstance(item, Iterable):
-                return QueryVector(values=item)
+                return QueryVector(values=item, _check_type=_check_type)
             raise ValueError(f"Invalid query vector value passed: cannot interpret type {type(item)}")
 
         return self._vector_api.query(
             QueryRequest(
                 queries=list(map(_query_transform, queries)),
+                _check_type=_check_type,
                 **{k: v for k, v in kwargs.items() if k not in _OPENAPI_ENDPOINT_PARAMS}
             ),
             **{k: v for k, v in kwargs.items() if k in _OPENAPI_ENDPOINT_PARAMS}

test-requirements.txt

@@ -4,4 +4,5 @@ pytest-cov==2.10.1
 pytest-mock==3.6.1
 tox==3.20.1
 pytest-timeout==1.4.2
-urllib3_mock==0.3.3
+urllib3_mock==0.3.3
+responses==0.15.0

tests/integ/test_index.py

@@ -1,38 +1,113 @@
+import json
+
 import pytest
 from urllib3_mock import Responses
+import responses as req_responses
 
 import pinecone
-from pinecone import ApiTypeError
+from pinecone import ApiTypeError, ApiException
 
+responses_req = Responses()
 responses = Responses('requests.packages.urllib3')
 
+@req_responses.activate
+@responses.activate
+def test_invalid_upsert_request_vector_value_type():
+    environment = 'example-environment'
+    project_name = 'example-project'
+    req_responses.add(
+        'GET', f'https://controller.{environment}.pinecone.io/actions/whoami',
+        status=200, content_type='application/json',
+        body=json.dumps(dict(project_name=project_name, user_label='example-label', user_name='test'))
+    )
+    responses.add(
+        'POST', '/vectors/upsert',
+        status=400, content_type='text/plain',
+        adding_headers={
+            'content-length': '62',
+            'date': 'Thu, 28 Oct 2021 09:14:51 GMT',
+            'server': 'envoy',
+            'connection': 'close'
+        },
+        body='vectors[0].values[1]: invalid value "type" for type TYPE_FLOAT'
+    )
+
+    pinecone.init('example-api-key', environment='example-environment')
+    with pytest.raises(ApiException) as exc_info:
+        index = pinecone.Index('example-index')
+        resp = index.upsert(vectors=[('vec1', [0.1]*8), ('vec2', [0.2]*8)])
+
+    assert len(responses.calls) == 1
+    assert responses.calls[0].request.scheme == 'https'
+    assert responses.calls[0].request.host == 'example-index-example-project.svc.example-environment.pinecone.io'
+    assert responses.calls[0].request.url == '/vectors/upsert'
 
+
+@req_responses.activate
 @responses.activate
-def test_unrecognized_response_field():
-    # unrecognized response fields are okay, shouldn't raise an exception
+def test_multiple_indexes():
+    environment = 'example-environment'
+    project_name = 'example-project'
+    index1_name = 'index-1'
+    index2_name = 'index-2'
+    req_responses.add(
+        'GET', f'https://controller.{environment}.pinecone.io/actions/whoami',
+        status=200, content_type='application/json',
+        body=json.dumps(dict(project_name=project_name, user_label='example-label', user_name='test'))
+    )
+    responses.add(
+        'GET', f'/describe_index_stats',
+        status=200, content_type='application/json',
+        adding_headers={
+            'date': 'Thu, 28 Oct 2021 09:14:51 GMT',
+            'server': 'envoy'
+        },
+        body='{"namespaces":{"":{"vectorCount":50000},"example-namespace-2":{"vectorCount":30000}},"dimension":1024}'
+    )
+
     pinecone.init('example-api-key', environment='example-environment')
 
-    # responses.add('GET', '/actions/whoami',  # fixme: requests-based, so mock fails?
-    #               body='{"project_name": "example-project", "user_label": "example-label", "user_name": "test"}',
-    #               status=200, content_type='application/json')
-    responses.add('DELETE', '/vectors/delete',
-                  body='{"deleted_count": 2, "unexpected_key": "xyzzy"}',
-                  status=200, content_type='application/json')
+    index1 = pinecone.Index(index1_name)
+    resp1 = index1.describe_index_stats()
+    assert resp1.dimension == 1024
+    assert responses.calls[0].request.host == f'{index1_name}-{project_name}.svc.{environment}.pinecone.io'
+
+    index2 = pinecone.Index(index2_name)
+    resp2 = index2.describe_index_stats()
+    assert resp2.dimension == 1024
+    assert responses.calls[1].request.host == f'{index2_name}-{project_name}.svc.{environment}.pinecone.io'
 
+
+@req_responses.activate
+@responses.activate
+def test_invalid_delete_response_unrecognized_field():
+    # unrecognized response fields are okay, shouldn't raise an exception
+    environment = 'example-environment'
+    project_name = 'example-project'
+    req_responses.add(
+        'GET', f'https://controller.{environment}.pinecone.io/actions/whoami',
+        status=200, content_type='application/json',
+        body=json.dumps(dict(project_name=project_name, user_label='example-label', user_name='test'))
+    )
+    responses.add(
+        'DELETE', '/vectors/delete',
+        body='{"unexpected_key": "xyzzy"}',
+        status=200, content_type='application/json'
+    )
+
+    pinecone.init('example-api-key', environment=environment)
     index = pinecone.Index('example-index')
     resp = index.delete(ids=['vec1', 'vec2'])
 
-    # assert len(responses.calls) == 1
-    # assert responses.calls[0].request.url == '/vectors/delete?ids=vec1&ids=vec2'
-    # assert responses.calls[0].request.host == 'example-index-unknown.svc.example-environment.pinecone.io'
-    # assert responses.calls[0].request.scheme == 'https'
-
-    assert resp.deleted_count == 2
+    assert len(req_responses.calls) == 1
+    assert responses.calls[0].request.scheme == 'https'
+    assert responses.calls[0].request.host == f'example-index-{project_name}.svc.{environment}.pinecone.io'
+    assert responses.calls[0].request.url == '/vectors/delete?ids=vec1&ids=vec2'
 
 
 @responses.activate
-def test_missing_response_field():
-    # unrecognized response fields are okay, shouldn't raise an exception
+def test_delete_response_missing_field():
+    # missing (optional) response fields are okay, shouldn't raise an exception
     pinecone.init('example-api-key', environment='example-environment')
     responses.add('DELETE', '/vectors/delete',
                   body='{}',
@@ -44,8 +119,9 @@ def test_missing_response_field():
 
 
 @responses.activate
-def test_malformed_response_wrong_type():
-    # unrecognized response fields are okay, shouldn't raise an exception
+def _test_invalid_delete_response_wrong_type():
+    # FIXME: re-enable this test when accepted_count added back to response
+    # wrong-typed response fields should raise an exception
     pinecone.init('example-api-key', environment='example-environment')
 
     responses.add('DELETE', '/vectors/delete',