:hedgehog: PostHog is the leading platform for building self-driving products. Our developer tools – AI observability, analytics, session replay, flags, experiments, error tracking, logs, and more – capture all the context agents need to diagnose problems, uncover opportunities, and ship fixes. Steer it all from Slack, web, desktop, or the MCP.
Python ab-testingai-analyticsanalyticscdpdata-warehouseexperimentsfeature-flagsjavascriptproduct-analyticspythonreactsession-replaysurveystypescriptweb-analytics
import re
import json
from pathlib import Path
from django.test import SimpleTestCase
from parameterized import parameterized
from posthog.schema import PropertyOperator, SpanPropertyFilter, SpanPropertyFilterType
from products.tracing.backend.logic import translate_span_filter
PROMPTS_DIR = Path(__file__).parents[2] / "mcp" / "prompts"
def _span_filter(key: str, value: object) -> SpanPropertyFilter:
return SpanPropertyFilter(
key=key,
operator=PropertyOperator.EXACT,
type=SpanPropertyFilterType.SPAN,
value=value,
)
class TestTranslateSpanFilter(SimpleTestCase):
@parameterized.expand(
[
# status_code normalises to string-digit codes. Integers and digit strings
# must survive — an integer must NOT collapse to [] (which filters nothing).
("status_code_int", "status_code", 2, ["2"]),
("status_code_int_zero", "status_code", 0, ["0"]),
("status_code_digit_string", "status_code", "2", ["2"]),
("status_code_error_label", "status_code", "Error", ["2"]),
("status_code_ok_label", "status_code", "OK", ["0", "1"]),
("status_code_list_ints", "status_code", [0, 2], ["0", "2"]),
# kind normalises to string-digit codes (same form as status_code). Digit strings
# and ints must survive — they must NOT collapse to [].
("kind_int", "kind", 2, ["2"]),
("kind_digit_string", "kind", "3", ["3"]),
("kind_server_label", "kind", "Server", ["2"]),
("kind_list_digit_strings", "kind", ["2", "3"], ["2", "3"]),
]
)
def test_normalises_status_code_and_kind_values(self, _name, key, value, expected):
span_filter = _span_filter(key, value)
translate_span_filter(span_filter)
self.assertEqual(span_filter.value, expected)
@parameterized.expand(
[
# Duration is given in milliseconds and converted to nanoseconds. The result must be an
# integer string — a `.0`-suffixed value can't be cast to the UInt64 `duration_nano`
# column and makes ClickHouse throw. Floats (e.g. the slider max) are the regression.
("duration_int", 2, "2000000"),
("duration_float", 2000000000.0, "2000000000000000"),
("duration_float_string", "2000000000.0", "2000000000000000"),
("duration_int_string", "2", "2000000"),
]
)
def test_duration_converts_to_integer_nanoseconds(self, _name, value, expected):
span_filter = _span_filter("duration", value)
translate_span_filter(span_filter)
self.assertEqual(span_filter.key, "duration_nano")
self.assertEqual(span_filter.value, expected)
@parameterized.expand(
[
# The MCP prompts tell agents what unit to send. These examples must mean the
# threshold their prose states once the backend translates them.
("query-apm-spans.md", 1_000_000_000),
("apm-spans-count.md", 30_000_000_000),
]
)
def test_documented_mcp_duration_example_matches_stated_threshold(self, prompt, expected_nano):
match = re.search(r'\{ "key": "duration", [^}]*\}', (PROMPTS_DIR / prompt).read_text())
assert match is not None
span_filter = SpanPropertyFilter(**json.loads(match.group(0)))
translate_span_filter(span_filter)
self.assertEqual(span_filter.value, str(expected_nano))
def test_duration_list_converts_to_integer_nanoseconds(self):
span_filter = _span_filter("duration", [1.0, 2000000000.0])
translate_span_filter(span_filter)
self.assertEqual(span_filter.key, "duration_nano")
self.assertEqual(span_filter.value, ["1000000", "2000000000000000"])
@parameterized.expand(
[
("status_code", "status_code", 2, ["2"]),
("kind", "kind", "3", ["3"]),
]
)
def test_translation_is_idempotent(self, _name, key, value, expected):
span_filter = _span_filter(key, value)
translate_span_filter(span_filter)
translate_span_filter(span_filter)
self.assertEqual(span_filter.value, expected)