|
1
|
"""Tests for robust JSON parsing from LLM responses.""" |
|
2
|
|
|
3
|
from video_processor.utils.json_parsing import parse_json_from_response |
|
4
|
|
|
5
|
|
|
6
|
class TestParseJsonFromResponse: |
|
7
|
def test_direct_dict(self): |
|
8
|
assert parse_json_from_response('{"key": "value"}') == {"key": "value"} |
|
9
|
|
|
10
|
def test_direct_array(self): |
|
11
|
assert parse_json_from_response("[1, 2, 3]") == [1, 2, 3] |
|
12
|
|
|
13
|
def test_markdown_fenced_json(self): |
|
14
|
text = '```json\n{"key": "value"}\n```' |
|
15
|
assert parse_json_from_response(text) == {"key": "value"} |
|
16
|
|
|
17
|
def test_markdown_fenced_no_lang(self): |
|
18
|
text = "```\n[1, 2]\n```" |
|
19
|
assert parse_json_from_response(text) == [1, 2] |
|
20
|
|
|
21
|
def test_json_embedded_in_text(self): |
|
22
|
text = 'Here is the result:\n{"name": "test", "value": 42}\nEnd of result.' |
|
23
|
result = parse_json_from_response(text) |
|
24
|
assert result == {"name": "test", "value": 42} |
|
25
|
|
|
26
|
def test_array_embedded_in_text(self): |
|
27
|
text = 'The entities are:\n[{"name": "Alice"}, {"name": "Bob"}]\nThat is all.' |
|
28
|
result = parse_json_from_response(text) |
|
29
|
assert len(result) == 2 |
|
30
|
assert result[0]["name"] == "Alice" |
|
31
|
|
|
32
|
def test_nested_json(self): |
|
33
|
text = '{"outer": {"inner": [1, 2, 3]}}' |
|
34
|
result = parse_json_from_response(text) |
|
35
|
assert result["outer"]["inner"] == [1, 2, 3] |
|
36
|
|
|
37
|
def test_empty_string(self): |
|
38
|
assert parse_json_from_response("") is None |
|
39
|
|
|
40
|
def test_none_input(self): |
|
41
|
assert parse_json_from_response(None) is None |
|
42
|
|
|
43
|
def test_whitespace_only(self): |
|
44
|
assert parse_json_from_response(" \n ") is None |
|
45
|
|
|
46
|
def test_no_json(self): |
|
47
|
assert parse_json_from_response("This is just plain text.") is None |
|
48
|
|
|
49
|
def test_invalid_json(self): |
|
50
|
assert parse_json_from_response("{invalid json}") is None |
|
51
|
|
|
52
|
def test_multiple_json_objects_picks_first(self): |
|
53
|
text = '{"a": 1} and {"b": 2}' |
|
54
|
result = parse_json_from_response(text) |
|
55
|
assert result == {"a": 1} |
|
56
|
|
|
57
|
def test_complex_fenced(self): |
|
58
|
text = """Here is the analysis: |
|
59
|
|
|
60
|
```json |
|
61
|
[ |
|
62
|
{"point": "Architecture uses microservices", "topic": "Architecture"}, |
|
63
|
{"point": "Deployment is automated", "topic": "DevOps"} |
|
64
|
] |
|
65
|
``` |
|
66
|
|
|
67
|
I hope this helps!""" |
|
68
|
result = parse_json_from_response(text) |
|
69
|
assert len(result) == 2 |
|
70
|
assert result[0]["topic"] == "Architecture" |
|
71
|
|