add test for opinion in diff sentence (Azure#13524)

Author: iscai-msft

sdk/textanalytics/azure-ai-textanalytics/tests/test_json_pointer.py

@@ -0,0 +1,111 @@
+# coding=utf-8
+# --------------------------------------------------------------------------
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License. See License.txt in the project root for
+# license information.
+# --------------------------------------------------------------------------
+import pytest
+from azure.ai.textanalytics._models import (
+    AnalyzeSentimentResult,
+    AspectSentiment,
+    OpinionSentiment,
+    SentenceSentiment,
+    _get_indices,
+)
+
+from azure.ai.textanalytics._response_handlers import sentiment_result
+
+from azure.ai.textanalytics._generated.v3_1_preview_1 import models as _generated_models
+
+
+@pytest.fixture
+def generated_aspect_opinion_confidence_scores():
+    return _generated_models.AspectConfidenceScoreLabel(
+        positive=1.0,
+        neutral=0.0,
+        negative=0.0,
+    )
+
+@pytest.fixture
+def generated_sentiment_confidence_score():
+    return _generated_models.SentimentConfidenceScorePerLabel(
+        positive=1.0,
+        neutral=0.0,
+        negative=0.0,
+    )
+
+@pytest.fixture
+def generated_aspect_relation():
+    return _generated_models.AspectRelation(
+        relation_type="opinion",
+        ref="#/documents/0/sentences/1/opinions/0"
+    )
+
+@pytest.fixture
+def generated_aspect(generated_aspect_opinion_confidence_scores, generated_aspect_relation):
+    return _generated_models.SentenceAspect(
+        text="aspect",
+        sentiment="positive",
+        confidence_scores=generated_aspect_opinion_confidence_scores,
+        offset=0,
+        length=6,
+        relations=[generated_aspect_relation],
+    )
+
+@pytest.fixture
+def generated_opinion(generated_aspect_opinion_confidence_scores):
+    return _generated_models.SentenceOpinion(
+        text="good",
+        sentiment="positive",
+        confidence_scores=generated_aspect_opinion_confidence_scores,
+        offset=0,
+        length=4,
+        is_negated=False,
+    )
+
+def generated_sentence_sentiment(generated_sentiment_confidence_score, index, aspects=[], opinions=[]):
+    return _generated_models.SentenceSentiment(
+        text="not relevant",
+        sentiment="positive",
+        confidence_scores=generated_sentiment_confidence_score,
+        offset=0,
+        length=12,
+        aspects=aspects,
+        opinions=opinions,
+    )
+
+@pytest.fixture
+def generated_document_sentiment(generated_aspect, generated_opinion, generated_sentiment_confidence_score):
+    aspect_sentence = generated_sentence_sentiment(generated_sentiment_confidence_score, index=0, aspects=[generated_aspect])
+    opinion_sentence = generated_sentence_sentiment(generated_sentiment_confidence_score, index=1, opinions=[generated_opinion])
+
+    return _generated_models.DocumentSentiment(
+        id=1,
+        sentiment="positive",
+        confidence_scores=generated_sentiment_confidence_score,
+        sentences=[aspect_sentence, opinion_sentence],
+        warnings=[],
+    )
+
+@pytest.fixture
+def generated_sentiment_response(generated_document_sentiment):
+    return _generated_models.SentimentResponse(
+        documents=[generated_document_sentiment],
+        errors=[],
+        model_version="0000-00-00",
+    )
+
+
+class TestJsonPointer():
+
+    def test_json_pointer_parsing(self):
+        assert [1, 0, 15] == _get_indices("#/documents/1/sentences/0/opinions/15")
+
+    def test_opinion_different_sentence_aspect(self, generated_sentiment_response):
+        # the first sentence has the aspect, and the second sentence has the opinion
+        # the desired behavior is the first wrapped sentence object has an aspect, and it's opinion
+        # is in the second sentence.
+        # the second sentence will have no mined opinions, since we define that as an aspect and opinion duo
+        wrapped_sentiment = sentiment_result(response="not relevant", obj=generated_sentiment_response, response_headers={})[0]
+        assert wrapped_sentiment.sentences[0].mined_opinions[0].opinions[0].text == "good"
+        assert not wrapped_sentiment.sentences[1].mined_opinions