|
| 1 | +from collections import defaultdict |
| 2 | +from dataclasses import asdict, dataclass |
| 3 | +from dataclasses import field as dataclass_field |
| 4 | +from typing import Any |
| 5 | + |
| 6 | +from rest_framework.exceptions import ParseError |
| 7 | +from rest_framework.request import Request |
| 8 | +from rest_framework.response import Response |
| 9 | +from sentry_protos.snuba.v1.endpoint_trace_item_attributes_pb2 import TraceItemAttributeNamesRequest |
| 10 | +from sentry_protos.snuba.v1.request_common_pb2 import RequestMeta |
| 11 | +from sentry_protos.snuba.v1.trace_item_attribute_pb2 import AttributeKey |
| 12 | + |
| 13 | +from sentry.api.api_publish_status import ApiPublishStatus |
| 14 | +from sentry.api.bases import NoProjects, OrganizationEventsEndpointBase |
| 15 | +from sentry.api.utils import handle_query_errors |
| 16 | +from sentry.exceptions import InvalidSearchQuery |
| 17 | +from sentry.models.organization import Organization |
| 18 | +from sentry.search.eap import constants |
| 19 | +from sentry.search.eap.columns import ResolvedAttribute |
| 20 | +from sentry.search.eap.resolver import SearchResolver |
| 21 | +from sentry.search.eap.types import SearchResolverConfig |
| 22 | +from sentry.search.events import fields |
| 23 | +from sentry.snuba.referrer import Referrer |
| 24 | +from sentry.snuba.utils import RPC_DATASETS |
| 25 | +from sentry.utils import snuba_rpc |
| 26 | +from sentry.utils.concurrent import ContextPropagatingThreadPoolExecutor |
| 27 | + |
| 28 | + |
| 29 | +@dataclass(kw_only=True) |
| 30 | +class Validation: |
| 31 | + valid: bool |
| 32 | + error: str | None |
| 33 | + |
| 34 | + |
| 35 | +@dataclass(kw_only=True) |
| 36 | +class AttributeValidation(Validation): |
| 37 | + name: str |
| 38 | + # None when its an error |
| 39 | + attrType: str | None |
| 40 | + |
| 41 | + |
| 42 | +@dataclass(kw_only=True) |
| 43 | +class ValidationResponse: |
| 44 | + valid: bool |
| 45 | + projects: list[Validation] = dataclass_field(default_factory=list) |
| 46 | + dataset: list[Validation] = dataclass_field(default_factory=list) |
| 47 | + field: list[AttributeValidation] = dataclass_field(default_factory=list) |
| 48 | + orderby: list[AttributeValidation] = dataclass_field(default_factory=list) |
| 49 | + |
| 50 | + |
| 51 | +def serialize_type(search_type: constants.SearchType) -> str: |
| 52 | + proto_type = constants.TYPE_MAP.get(search_type) |
| 53 | + if proto_type == constants.STRING: |
| 54 | + return "string" |
| 55 | + if proto_type == constants.BOOLEAN: |
| 56 | + return "boolean" |
| 57 | + # DOUBLE, INT, or anything else numeric |
| 58 | + return "number" |
| 59 | + |
| 60 | + |
| 61 | +MAX_ATTRIBUTE_VALIDATION_THREADS = 3 |
| 62 | + |
| 63 | + |
| 64 | +def _check_attributes_by_type( |
| 65 | + meta: RequestMeta, |
| 66 | + attr_type: AttributeKey.Type.ValueType, |
| 67 | + attributes: list[ResolvedAttribute], |
| 68 | +) -> set[tuple[AttributeKey.Type.ValueType, str]]: |
| 69 | + """Check which typed attribute names exist in storage for the active window.""" |
| 70 | + if not attributes: |
| 71 | + return set() |
| 72 | + |
| 73 | + requested_names = set(attribute.internal_name for attribute in attributes) |
| 74 | + # TODO(wmak): Need to update snuba here so we can pass the list of attributes, snuba currently does a hasAll if we |
| 75 | + # pass names in a OrFilter which means only rows with _all_ attributes will return |
| 76 | + attrs_request = TraceItemAttributeNamesRequest( |
| 77 | + meta=meta, |
| 78 | + limit=10_000, |
| 79 | + type=attr_type, |
| 80 | + ) |
| 81 | + attrs_response = snuba_rpc.attribute_names_rpc(attrs_request) |
| 82 | + return { |
| 83 | + (attr_type, attribute.name) |
| 84 | + for attribute in attrs_response.attributes |
| 85 | + if attribute.name in requested_names |
| 86 | + } |
| 87 | + |
| 88 | + |
| 89 | +def check_attributes_exist( |
| 90 | + resolver: SearchResolver, |
| 91 | + dataset: Any, |
| 92 | + attrs_by_type: dict[AttributeKey.Type.ValueType, list[ResolvedAttribute]], |
| 93 | +) -> set[tuple[AttributeKey.Type.ValueType, str]]: |
| 94 | + """Check which typed attribute internal names exist in storage.""" |
| 95 | + if not attrs_by_type: |
| 96 | + return set() |
| 97 | + |
| 98 | + meta = resolver.resolve_meta(referrer=Referrer.API_TRACE_ITEM_ATTRIBUTE_VALIDATE.value) |
| 99 | + |
| 100 | + found: set[tuple[AttributeKey.Type.ValueType, str]] = set() |
| 101 | + with ContextPropagatingThreadPoolExecutor( |
| 102 | + thread_name_prefix="attr_validate", |
| 103 | + max_workers=MAX_ATTRIBUTE_VALIDATION_THREADS, |
| 104 | + ) as pool: |
| 105 | + futures = [ |
| 106 | + pool.submit(_check_attributes_by_type, meta, attr_type, names) |
| 107 | + for attr_type, names in attrs_by_type.items() |
| 108 | + ] |
| 109 | + for future in futures: |
| 110 | + found.update(future.result()) |
| 111 | + |
| 112 | + return found |
| 113 | + |
| 114 | + |
| 115 | +class OrganizationEventsValidateEndpoint(OrganizationEventsEndpointBase): |
| 116 | + publish_status = { |
| 117 | + "GET": ApiPublishStatus.EXPERIMENTAL, |
| 118 | + } |
| 119 | + |
| 120 | + def serialize_response( |
| 121 | + self, |
| 122 | + validity: ValidationResponse, |
| 123 | + ) -> Response: |
| 124 | + return Response( |
| 125 | + status=200 if validity.valid else 400, |
| 126 | + data=asdict(validity), |
| 127 | + ) |
| 128 | + |
| 129 | + def get(self, request: Request, organization: Organization) -> Response: |
| 130 | + if not self.has_feature(organization, request): |
| 131 | + return Response(status=400) |
| 132 | + |
| 133 | + response = ValidationResponse(valid=True) |
| 134 | + |
| 135 | + try: |
| 136 | + snuba_params = self.get_snuba_params( |
| 137 | + request, |
| 138 | + organization, |
| 139 | + ) |
| 140 | + except NoProjects: |
| 141 | + response.valid = False |
| 142 | + response.projects.append( |
| 143 | + Validation(valid=False, error="At least one valid project is required to query") |
| 144 | + ) |
| 145 | + return self.serialize_response(response) |
| 146 | + |
| 147 | + try: |
| 148 | + dataset = self.get_dataset(request, organization) |
| 149 | + except ParseError as error: |
| 150 | + response.valid = False |
| 151 | + response.dataset.append(Validation(valid=False, error=str(error))) |
| 152 | + return self.serialize_response(response) |
| 153 | + |
| 154 | + if dataset not in RPC_DATASETS: |
| 155 | + response.dataset.append( |
| 156 | + Validation( |
| 157 | + valid=True, |
| 158 | + error="This dataset is not compatible with the validate endpoint, your request may still be valid", |
| 159 | + ) |
| 160 | + ) |
| 161 | + # Can't continue if this isn't a RPC dataset |
| 162 | + return self.serialize_response(response) |
| 163 | + |
| 164 | + resolver = dataset.get_resolver(snuba_params, SearchResolverConfig()) |
| 165 | + definitions = resolver.definitions |
| 166 | + |
| 167 | + # Validate selected_columns |
| 168 | + selected_columns = self.get_field_list(organization, request) |
| 169 | + attributes_to_lookup = defaultdict(list) |
| 170 | + column_validity: list[AttributeValidation] = [] |
| 171 | + for column in selected_columns: |
| 172 | + try: |
| 173 | + match = fields.is_function(column) |
| 174 | + if match: |
| 175 | + resolved, _ = resolver.resolve_function(column, match) |
| 176 | + column_validity.append( |
| 177 | + AttributeValidation( |
| 178 | + attrType=serialize_type(resolved.search_type), |
| 179 | + error=None, |
| 180 | + name=column, |
| 181 | + valid=True, |
| 182 | + ) |
| 183 | + ) |
| 184 | + else: |
| 185 | + resolved, _ = resolver.resolve_attribute(column) |
| 186 | + if column in definitions.contexts or column in definitions.columns: |
| 187 | + column_validity.append( |
| 188 | + AttributeValidation( |
| 189 | + attrType=serialize_type(resolved.search_type), |
| 190 | + error="", |
| 191 | + name=column, |
| 192 | + valid=True, |
| 193 | + ) |
| 194 | + ) |
| 195 | + else: |
| 196 | + attributes_to_lookup[resolved.proto_type].append(resolved) |
| 197 | + except InvalidSearchQuery as error: |
| 198 | + response.valid = False |
| 199 | + column_validity.append( |
| 200 | + AttributeValidation( |
| 201 | + attrType=None, |
| 202 | + error=str(error), |
| 203 | + name=column, |
| 204 | + valid=False, |
| 205 | + ) |
| 206 | + ) |
| 207 | + |
| 208 | + if any(len(attributes) > 0 for attributes in attributes_to_lookup.values()): |
| 209 | + # Group by proto type because the storage check is keyed on |
| 210 | + # (proto_type, internal_name) — the same display name can exist |
| 211 | + # as both a string and a number attribute simultaneously. |
| 212 | + with handle_query_errors(): |
| 213 | + existing = check_attributes_exist(resolver, dataset, attributes_to_lookup) |
| 214 | + for attribute_type, attributes in attributes_to_lookup.items(): |
| 215 | + for resolved in attributes: |
| 216 | + if (resolved.proto_type, resolved.internal_name) in existing: |
| 217 | + column_validity.append( |
| 218 | + AttributeValidation( |
| 219 | + attrType=serialize_type(resolved.search_type), |
| 220 | + error="", |
| 221 | + name=resolved.public_alias, |
| 222 | + valid=True, |
| 223 | + ) |
| 224 | + ) |
| 225 | + else: |
| 226 | + response.valid = False |
| 227 | + column_validity.append( |
| 228 | + AttributeValidation( |
| 229 | + attrType=None, |
| 230 | + error="Unknown attribute", |
| 231 | + name=resolved.public_alias, |
| 232 | + valid=False, |
| 233 | + ) |
| 234 | + ) |
| 235 | + response.field = column_validity |
| 236 | + |
| 237 | + # Validate orderby |
| 238 | + orderby_validity = [] |
| 239 | + orderby_columns = self.get_orderby(request) |
| 240 | + if orderby_columns: |
| 241 | + for orderby in orderby_columns: |
| 242 | + stripped_orderby = orderby.lstrip("-") |
| 243 | + found = False |
| 244 | + for field in column_validity: |
| 245 | + if ( |
| 246 | + field.name == stripped_orderby |
| 247 | + or fields.get_function_alias(field.name) == stripped_orderby |
| 248 | + ): |
| 249 | + orderby_validity.append( |
| 250 | + AttributeValidation( |
| 251 | + attrType=field.attrType, error=None, name=orderby, valid=True |
| 252 | + ) |
| 253 | + ) |
| 254 | + found = True |
| 255 | + break |
| 256 | + if not found: |
| 257 | + response.valid = False |
| 258 | + orderby_validity.append( |
| 259 | + AttributeValidation( |
| 260 | + attrType=None, |
| 261 | + error="Orderby must also be a selected field", |
| 262 | + name=orderby, |
| 263 | + valid=False, |
| 264 | + ) |
| 265 | + ) |
| 266 | + response.orderby = orderby_validity |
| 267 | + |
| 268 | + # TODO(wmak): Validate query |
| 269 | + |
| 270 | + return self.serialize_response(response) |
0 commit comments