ref(utils): SDK name tag normalizer (getsentry#59504)

getsentry#59501 Normalizes SDK tags to reduce their cardinality. 
Related to getsentry#59075 and getsentry#59379.

- non-Sentry SDK tags are ignored (collapsed into `"other"`)
- official Sentry SDK tags are normalized and shortened: 
    -  `sentry.javascript.*` are mostly kept as-is
    -  `sentry.native.*` are collapsed to 3 levels
    -  all other `sentry.*` are collapsed to 2 levels

---------

Co-authored-by: Katie Byers <katie.byers@sentry.io>
Co-authored-by: getsantry[bot] <66042841+getsantry[bot]@users.noreply.github.com>

src/sentry/utils/tag_normalization.py

@@ -0,0 +1,94 @@
+import functools
+import re
+
+_KNOWN_TAGS = {
+    "sentry.cocoa",
+    "sentry.dart",
+    "sentry.dotnet",
+    "sentry.elixir",
+    "sentry.go",
+    "sentry.java",
+    "sentry.javascript.angular",
+    "sentry.javascript.browser",
+    "sentry.javascript.capacitor",
+    "sentry.javascript.cordova",
+    "sentry.javascript.deno",
+    "sentry.javascript.electron",
+    "sentry.javascript.ember",
+    "sentry.javascript.gatsby",
+    "sentry.javascript.nextjs",
+    "sentry.javascript.node",
+    "sentry.javascript.react",
+    "sentry.javascript.react.native",
+    "sentry.javascript.remix",
+    "sentry.javascript.serverless",
+    "sentry.javascript.svelte",
+    "sentry.javascript.sveltekit",
+    "sentry.javascript.vue",
+    "sentry.kubernetes",
+    "sentry.native.android",
+    "sentry.native.dotnet",
+    "sentry.native.unity",
+    "sentry.native.unreal",
+    "sentry.objc",
+    "sentry.perl",
+    "sentry.php",
+    "sentry.python",
+    "sentry.ruby",
+    "sentry.rust",
+    "sentry.swift",
+}
+
+
+_SYNONYMOUS_TAGS = {
+    "sentry.cordova": "sentery.javascript.cordova",
+    "sentry.electron": "sentry.javascript.electron",
+    "sentry.javascript.angular.ivy": "sentry.javascript.angular",
+    "sentry.javascript.node.experimental": "sentry.javascript.node",
+    "sentry.javascript.react.expo": "sentry.javascript.react",
+    "sentry.javascript.react.native.expo": "sentry.javascript.react.native",
+    "sentry.laravel": "sentry.php.laravel",
+    "sentry.react": "sentry.javascript.react",
+    "sentry.symfony": "sentry.php.symfony",
+    "sentry.unity": "sentry.native.unity",
+}
+
+# TODO: Should we be grouping by origin SDK instead? (For example, should we be
+#       combining all flutter events rather than all native events?)
+#       See https://github.com/getsentry/sentry/pull/59504#discussion_r1385483963
+
+
+@functools.lru_cache(maxsize=300)
+def normalize_sdk_tag(tag: str) -> str:
+    """
+     Normalize tags coming from SDKs to more manageable canonical form, by:
+
+     - combining synonymous tags (`sentry.react` -> `sentry.javascript.react`),
+     - ignoring framework differences (`sentry.python.flask` and `sentry.python.django` -> `sentry.python`)
+     - collapsing all community/third-party SDKs into a single `other` category
+
+    Note: Some platforms may keep their framework-specific values, as needed for analytics.
+    """
+
+    # replace non-word characters with dots (normalize sentry-foo to sentry.foo)
+    tag = re.sub(r"[\W_]+", ".", tag)
+
+    # collapse known synonymous tags
+    tag = _SYNONYMOUS_TAGS.get(tag, tag)
+
+    # ignore non-sentry SDK tags
+    if not tag.startswith("sentry."):
+        return "other"
+
+    # collapse tags other than JavaScript / Native to their top-level SDK
+
+    if not tag.split(".")[1] in {"javascript", "native"}:
+        tag = ".".join(tag.split(".", 2)[0:2])
+
+    if tag.split(".")[1] == "native":
+        tag = ".".join(tag.split(".", 3)[0:3])
+
+    if tag not in _KNOWN_TAGS:
+        tag = "other"
+
+    return tag

tests/sentry/utils/test_tag_normalization.py

@@ -0,0 +1,72 @@
+import pytest
+
+from sentry.utils.tag_normalization import normalize_sdk_tag
+
+
+@pytest.mark.parametrize(
+    ("tag", "expected"),
+    (
+        ("sentry-javascript-angular", "sentry.javascript.angular"),
+        ("sentry_python", "sentry.python"),
+    ),
+)
+def test_normalizes_to_dots(tag, expected):
+    assert normalize_sdk_tag(tag) == expected
+
+
+@pytest.mark.parametrize(
+    ("tag", "expected"),
+    (
+        ("sentry.javascript.angular", "sentry.javascript.angular"),
+        (
+            "sentry.javascript.react.native",
+            "sentry.javascript.react.native",
+        ),
+        ("sentry.python.django", "sentry.python"),
+        (
+            "sentry.native.android.flutter",
+            "sentry.native.android",
+        ),
+    ),
+)
+def test_shortens_non_js(tag, expected):
+    assert normalize_sdk_tag(tag) == expected
+
+
+@pytest.mark.parametrize(
+    ("tag", "expected"),
+    (
+        ("sentry.javascript.angular", "sentry.javascript.angular"),
+        ("sentry.javascript.angular.ivy", "sentry.javascript.angular"),
+        ("sentry.symfony", "sentry.php"),
+        ("sentry.unity", "sentry.native.unity"),
+        ("sentry.javascript.react.native.expo", "sentry.javascript.react.native"),
+    ),
+)
+def test_uses_synonyms(tag, expected):
+    assert normalize_sdk_tag(tag) == expected
+
+
+@pytest.mark.parametrize(
+    ("tag", "expected"),
+    (("foo.baz.bar", "other"), ("sentryfoo", "other"), ("raven", "other")),
+)
+def test_non_sentry_to_other(tag, expected):
+    assert normalize_sdk_tag(tag) == expected
+
+
+@pytest.mark.parametrize(
+    ("tag", "expected"),
+    (("sentry.sparql", "other"), ("sentry.terraform.hcl", "other"), ("sentry-native", "other")),
+)
+def test_unknown_sentry_to_other(tag, expected):
+    assert normalize_sdk_tag(tag) == expected
+
+
+def test_responses_cached():
+    normalize_sdk_tag.cache_clear()
+    assert normalize_sdk_tag("sentry.javascript.react") == "sentry.javascript.react"
+    assert normalize_sdk_tag("sentry.javascript.react") == "sentry.javascript.react"
+
+    assert normalize_sdk_tag.cache_info().hits == 1
+    assert normalize_sdk_tag.cache_info().misses == 1