|
|
@@ -13,7 +13,9 @@ from sentry.grouping.strategies.base import (
|
|
|
from sentry.interfaces.message import Message
|
|
|
from sentry.utils import metrics
|
|
|
|
|
|
-_irrelevant_re = re.compile(
|
|
|
+_parameterization_regex = re.compile(
|
|
|
+ # The `(?x)` tells the regex compiler to ingore comments and unescaped whitespace,
|
|
|
+ # so we can use newlines and indentation for better legibility.
|
|
|
r"""(?x)
|
|
|
(?P<email>
|
|
|
[a-zA-Z0-9.!#$%&'*+/=?^_`{|}~-]+@[a-zA-Z0-9-]+(?:\.[a-zA-Z0-9-]+)*
|
|
|
@@ -96,32 +98,44 @@ _irrelevant_re = re.compile(
|
|
|
\b\d+\b
|
|
|
) |
|
|
|
(?P<quoted_str>
|
|
|
+ # The `=` here guarantees we'll only match the value half of key-value pairs,
|
|
|
+ # rather than all quoted strings
|
|
|
='([\w\s]+)'
|
|
|
)
|
|
|
"""
|
|
|
)
|
|
|
|
|
|
|
|
|
-def trim_message_for_grouping(string: str) -> str:
|
|
|
- """Replace values from a group's message to hide P.I.I. and improve grouping when no
|
|
|
- stacktrace available.
|
|
|
+def normalize_message_for_grouping(message: str) -> str:
|
|
|
+ """Replace values from a group's message with placeholders (to hide P.I.I. and
|
|
|
+ improve grouping when no stacktrace is available) and trim to at most 2 lines.
|
|
|
"""
|
|
|
- s = "\n".join(islice((x for x in string.splitlines() if x.strip()), 2)).strip()
|
|
|
- if s != string:
|
|
|
- s += "..."
|
|
|
+ trimmed = "\n".join(
|
|
|
+ # If there are multiple lines, grab the first two non-empty ones.
|
|
|
+ islice(
|
|
|
+ (x for x in message.splitlines() if x.strip()),
|
|
|
+ 2,
|
|
|
+ )
|
|
|
+ )
|
|
|
+ if trimmed != message:
|
|
|
+ trimmed += "..."
|
|
|
|
|
|
def _handle_match(match: Match[str]) -> str:
|
|
|
- # e.g. hex, 0x40000015
|
|
|
+ # Find the first (should be only) non-None match entry, and sub in the placeholder. For
|
|
|
+ # example, given the groupdict item `('hex', '0x40000015')`, this returns '<hex>' as a
|
|
|
+ # replacement for the original value in the string.
|
|
|
for key, value in match.groupdict().items():
|
|
|
if value is not None:
|
|
|
- # key can be one of the keys from _irrelevant_re, thus, not a large cardinality
|
|
|
- # tracking the key helps distinguish what kinds of replacements are happening
|
|
|
+ # `key` can only be one of the keys from `_parameterization_regex`, thus, not a large
|
|
|
+ # cardinality. Tracking the key helps distinguish what kinds of replacements are happening.
|
|
|
metrics.incr("grouping.value_trimmed_from_message", tags={"key": key})
|
|
|
- # For quoted_str we want to preserver the = symbol
|
|
|
+ # For `quoted_str` we want to preserve the `=` symbol, which we include in
|
|
|
+ # the match in order not to replace random quoted strings in contexts other
|
|
|
+ # than key-value pairs
|
|
|
return f"=<{key}>" if key == "quoted_str" else f"<{key}>"
|
|
|
return ""
|
|
|
|
|
|
- return _irrelevant_re.sub(_handle_match, s)
|
|
|
+ return _parameterization_regex.sub(_handle_match, trimmed)
|
|
|
|
|
|
|
|
|
@strategy(ids=["message:v1"], interface=Message, score=0)
|
|
|
@@ -129,14 +143,14 @@ def trim_message_for_grouping(string: str) -> str:
|
|
|
def message_v1(
|
|
|
interface: Message, event: Event, context: GroupingContext, **meta: Any
|
|
|
) -> ReturnedVariants:
|
|
|
- if context["trim_message"]:
|
|
|
- message_in = interface.message or interface.formatted or ""
|
|
|
- message_trimmed = trim_message_for_grouping(message_in)
|
|
|
- hint = "stripped common values" if message_in != message_trimmed else None
|
|
|
+ if context["normalize_message"]:
|
|
|
+ raw = interface.message or interface.formatted or ""
|
|
|
+ normalized = normalize_message_for_grouping(raw)
|
|
|
+ hint = "stripped event-specific values" if raw != normalized else None
|
|
|
return {
|
|
|
context["variant"]: GroupingComponent(
|
|
|
id="message",
|
|
|
- values=[message_trimmed],
|
|
|
+ values=[normalized],
|
|
|
hint=hint,
|
|
|
)
|
|
|
}
|
Katie Byers