5.1 Composition¶

The AWS deep collector · From docs/INVENTORY_INTERNALS.md on GitHub

AWSDeepInventoryCollector is a stack of mixins on top of AsyncAWSCollector:

class AWSDeepInventoryCollector(
    GovernanceCollectorsMixin,
    NetworkExtCollectorsMixin,
    ApplicationCollectorsMixin,
    DataMLCollectorsMixin,
    IdentityCollectorsMixin,
    ContainerCollectorsMixin,
    ServerlessCollectorsMixin,
    SecurityCollectorsMixin,
    PlatformCollectorsMixin,
    CloudControlCollectorsMixin,
    AsyncAWSCollector,
): ...

Rules that follow from the MRO:

Mixins come before AsyncAWSCollector. A mixin's _collect_s3 overrides the base collector's. AWSServiceMixin declares _region, _account_id and _aio_config as annotations only; it must never define stubs for AsyncAWSCollector methods, or it would shadow them.
Method names are global across all mixins. Two mixins defining the same helper silently shadow each other (the one earlier in the MRO wins). Prefix package-private helpers (the data/ML package uses _dm_role_ref for that reason) and check with grep -rn "def <name>" cloudg/inventory/aws_services before adding one.
Package mixins are themselves compositions (ApplicationCollectorsMixin combines twelve per-area mixins), so each package keeps one place in the MRO.

Constructor: AWSDeepInventoryCollector(session, region="us-east-1", account_id=None, tagging_sweep=True, **options). tagging_sweep stays the fourth positional parameter for 0.5.0 compatibility; every other option is keyword-only and validated by the DeepInventoryOptions dataclass, so an unknown option name raises TypeError.

Option	Default	Effect
`tagging_sweep`	`True`	Run the Tagging API sweep.
`is_primary_region`	`True`	Collect account-wide tasks here.
`services`	all	Families / task names to include.
`exclude_services`	none	Families / task names to skip (wins over `services`).
`kubernetes`	`True`	Map in-cluster objects of EKS clusters. Also requires that `kubernetes` is not excluded and that `all`, `kubernetes`, `containers` or `eks` is selected.
`kubernetes_timeout`	`10`	Kubernetes API timeout, seconds.
`iam_resource_edges`	`True`	Link principals to resources their identity policies name.
`max_images_per_repository`	`20`	ECR images sampled per repository (for tags, scan findings).
`stack_resources`	`True`	Link CloudFormation stacks to their resources.
`cloud_control`	`True`	Run the Cloud Control sweep.
`cloud_control_types`	all	Only these CloudFormation types or prefixes.
`cloud_control_exclude`	none	Skip these types or prefixes.
`cloud_control_concurrency`	`6`	Types listed in parallel per region.