Files
qdrant/openapi/tests/openapi_integration/test_basic_retrieve_api.py
Luis Cossío e90a03e00b Group by key (#1768)
* test: test must_not is_null

* vcs: ignore vscode files

* feat: group-by initial implementation

* cargo fmt

* refactor: same request behavior on reco and search

* refactor: get rid of RefCell

* refactor-fix: correct hashmap keys, and early stops

* chore: small improvements

* feat: groups aggregator

* fix: pull changes from other files

* cargo fix

* cargo fmt

* docs: edit docstrings

* allow dead code (while the complete feature is beint built)

* chore: restructure

* feat: introduce GroupKey, minor other improvements

* cargo fmt

* chore: specify aggregator visibility

* fix: oops, leaking "private" type

* refactor-fix: restructure and refactor group_by

* cargo fix

* fix: don't panic when there is no group-by field

* remove print statements

* amend: `>=`  -> `==`

* perf: remove double clone

* chore: sync aggregator from other branch

* chore: cleanup print statemets

* test: ignore big tests

* cargo fmt

* refactor: add early stop when the groups have been filled, improve code

* chore: sync aggregator, remove print from test

* refactor: consider shard_selection, improve collection_by_name handling

* feat: add bucketing to table of content

* refactor: better errors, improve tests

* test: add integration tests

* feat: add endpoints

* refactor: introduce ScoredPoint wrapper, restructure types

* sync aggregator

* edit internal grouping visibility

* feat: group_by internals

* cargo fmt

* cargo fmt

* refactor: turn inner fn into closure

* test: fix test to support new vector output representation

* feat: wire up grouping with actix

* expose grouped_by field

* fix: change output group format

* feat: wire up openapi

* fix: finish wiring up grouping in actix

* tests: fix test_group.py

* cargo fmt

* refactor: extract constants

* remove Hash from ScoredPoint

* `Option<collection_by_name>` -> `collection_by_name`

* fix: handle better cases on `match_on`

* fix: consider that subsequent calls can bring better results

* cargo fmt

* fix clippy warnings

* cargo fmt

* refactor: move `Group` to `types`, localize `hydrate_from`, remove `Deref` impls

* refactor `add_points`

* refactor: turn `GroupKey` into enum

* refactor: make `HashablePoint` inner struct private

* feat: add grpc layer, make new `PointGroup` type to use as output

* fix: update openapi models

* docs: update grpc docs

* fix merge errors

* refactor: add BaseGroupRequest to make code DRYer, improve doc comments

* cargo fmt

* perf: increase precision; choose best groups by score

* misc: add more integration tests, fix review comments

* cargo fmt

* fix: reimplement interface to flatten search and recommend requests, excluding offset

* cargo fmt

* refactor: move `r#do` impl to `GroupRequest`

* fix: update grpc docs

* perf: sort in reverse order

* fix: use fist value of a Value::Array

* fix: validate group_by to not support bracket notation, fix int. tests

* fix: update grpc validation

* tests: update collection_tests

* refactor: move validation to the api layers

* Oops: reupdate tests

* refactor: let the derives derive (thanks @ffuugoo)

* refactor: use a new GroupId on the output

also increases performance by copying less

* remove hashable set, take ordering into an account, fix mutliple groups values support

* fmt

* refactor group_id + rename per_group -> group_size, fix clippy

* remove GroupKey wrapper

* @agourlay review fixes

* refactor: `group_min_scores` and `group_max_scores` ->  `group_best_scores`

* refactor: use set difference on `keys_of_unfilled_best_groups`

* refactor: use set intersection on `len_of_filled_best_groups`

* refactor: turn best_group_keys into iterator

* fix: remove [] syntax limitation

* fix: update openapi.json

---------

Co-authored-by: Andrey Vasnetsov <andrey@vasnetsov.com>
2023-05-15 23:05:20 +02:00

353 lines
9.5 KiB
Python

import pytest
from .helpers.collection_setup import basic_collection_setup, drop_collection
from .helpers.helpers import request_with_validation
collection_name = 'test_collection'
@pytest.fixture(autouse=True)
def setup():
basic_collection_setup(collection_name=collection_name)
yield
drop_collection(collection_name=collection_name)
def test_points_retrieve():
points_retrieve()
def points_retrieve():
response = request_with_validation(
api='/collections/{collection_name}/points/{id}',
method="GET",
path_params={'collection_name': collection_name, 'id': 2},
)
assert response.ok
response = request_with_validation(
api='/collections/{collection_name}/points',
method="POST",
path_params={'collection_name': collection_name},
body={
"ids": [1, 2]
}
)
assert response.ok
assert len(response.json()['result']) == 2
response = request_with_validation(
api='/collections/{collection_name}',
method="GET",
path_params={'collection_name': collection_name},
)
assert response.ok
assert response.json()['result']['vectors_count'] == 8
response = request_with_validation(
api='/collections/{collection_name}/points/search',
method="POST",
path_params={'collection_name': collection_name},
body={
"vector": [0.2, 0.1, 0.9, 0.7],
"limit": 3
}
)
assert response.ok
assert len(response.json()['result']) == 3
response = request_with_validation(
api='/collections/{collection_name}/points/search',
method="POST",
path_params={'collection_name': collection_name},
body={
"filter": {
"should": [
{
"key": "city",
"match": {
"value": "London"
}
}
]
},
"vector": [0.2, 0.1, 0.9, 0.7],
"limit": 3
}
)
assert response.ok
# only 2 London records in collection
assert len(response.json()['result']) == 2
response = request_with_validation(
api='/collections/{collection_name}/points/scroll',
method="POST",
path_params={'collection_name': collection_name},
body={"offset": 2, "limit": 2, "with_vector": True}
)
assert response.ok
assert len(response.json()['result']['points']) == 2
def test_exclude_payload():
exclude_payload()
def exclude_payload():
response = request_with_validation(
api='/collections/{collection_name}/points/search',
method="POST",
path_params={'collection_name': collection_name},
body={
"vector": [0.2, 0.1, 0.9, 0.7],
"limit": 5,
"filter": {
"should": [
{
"key": "city",
"match": {
"value": "London"
}
}
]
},
"with_payload": {
"exclude": ["city"]
}
}
)
assert response.ok
assert len(response.json()['result']) > 0
for result in response.json()['result']:
assert 'city' not in result['payload']
def test_is_empty_condition():
is_empty_condition()
def is_empty_condition():
response = request_with_validation(
api='/collections/{collection_name}/points/search',
method="POST",
path_params={'collection_name': collection_name},
body={
"vector": [0.2, 0.1, 0.9, 0.7],
"limit": 5,
"filter": {
"should": [
{
"is_empty": {
"key": "city"
}
}
]
},
"with_payload": True
}
)
assert response.ok
json = response.json()
assert len(json['result']) == 4
ids = [x['id'] for x in json['result']]
assert 5 in ids
assert 6 in ids
assert 7 in ids
assert 8 in ids
def test_is_null_condition():
is_null_condition()
def is_null_condition():
response = request_with_validation(
api='/collections/{collection_name}/points/search',
method="POST",
path_params={'collection_name': collection_name},
body={
"vector": [0.2, 0.1, 0.9, 0.7],
"limit": 5,
"filter": {
"should": [
{
"is_null": {
"key": "city"
}
}
]
},
"with_payload": True
}
)
assert response.ok
json = response.json()
assert len(json['result']) == 1
ids = [x['id'] for x in json['result']]
assert 7 in ids
# With must_not (as recommended in docs)
def must_not_is_null(field: str):
response = request_with_validation(
api='/collections/{collection_name}/points/search',
method="POST",
path_params={'collection_name': collection_name},
body={
"vector": [0.2, 0.1, 0.9, 0.7],
"limit": 5,
"filter": {
"must_not": [
{
"is_null": {
"key": field
}
}
]
},
"with_payload": True
}
)
assert response.ok
json = response.json()
assert len(json['result']) == 5
ids = [x['id'] for x in json['result']]
assert 5 not in ids
assert 6 not in ids
assert 7 not in ids
assert 1 in ids
assert 2 in ids
must_not_is_null("city")
must_not_is_null("city[]")
def test_recommendation():
recommendation()
def recommendation():
response = request_with_validation(
api='/collections/{collection_name}/points/recommend',
method="POST",
path_params={'collection_name': collection_name},
body={
"limit": 3,
"negative": [],
"positive": [1],
"with_vector": False,
"with_payload": True
}
)
assert len(response.json()['result']) == 3
assert response.json()['result'][0]['payload'] is not None
assert response.ok
def test_query_nested():
query_nested()
def query_nested():
response = request_with_validation(
api='/collections/{collection_name}/points',
method="PUT",
path_params={'collection_name': collection_name},
query_params={'wait': 'true'},
body={
"points": [
{
"id": 8,
"vector": [0.15, 0.31, 0.76, 0.74],
"payload": {
"database_id": {
"type": "keyword",
"value": "8594ff5d-265f-4785-a9f5-b3b4b9665506"
}
}
}
]
}
)
assert response.ok
response = request_with_validation(
api='/collections/{collection_name}/points/scroll',
method="POST",
path_params={'collection_name': collection_name},
body={
"offset": None,
"limit": 10,
"with_vector": True,
"filter": {
"must": [
{
"key": "database_id.value",
"match": {
"value": "8594ff5d-265f-4785-a9f5-b3b4b9665506"
}
}
]
}
}
)
assert response.ok
assert len(response.json()['result']['points']) == 1
def test_with_vectors_alias_of_with_vector():
database_id = "8594ff5d-265f-adfh-a9f5-b3b4b9665506"
vector = [0.15, 0.31, 0.76, 0.74]
response = request_with_validation(
api='/collections/{collection_name}/points',
method="PUT",
path_params={'collection_name': collection_name},
query_params={'wait': 'true'},
body={
"points": [
{
"id": 8,
"vector": vector,
"payload": {
"database_id": database_id,
}
}
]
}
)
assert response.ok
def scroll_with_vector(keyword):
response = request_with_validation(
api='/collections/{collection_name}/points/scroll',
method='POST',
path_params={'collection_name': collection_name},
body={
keyword: True, # <--- should make no difference
"filter": {
"must": [
{
"key": "database_id",
"match": {
"value": database_id
}
}
]
},
"limit": 1,
}
)
assert response.ok
body = response.json()
assert body["result"]["points"][0]["vector"] == vector
scroll_with_vector("with_vector")
scroll_with_vector("with_vectors")