#rlsbl.resolved_target

#rlsbl.resolved_target

Resolution of a member's release surface into ResolvedTarget records, linking separately-configured targets and pipelines for the release flow.

Targets and pipelines are configured separately-but-linked (Phase 6.1): each pipeline entry declares an explicit target link (a target name, or null for a target-less publisher such as a docs/site deploy). This module turns that linked configuration into the flat, per-pipeline view the release flow wants.

A :class:ResolvedTarget is one publishable (target, pipeline) pair:

• A target served by N pipelines produces N ResolvedTargets -- the release flow

publishes once per pipeline, so per-pair records let it iterate directly without re-deriving the target for each pipeline.

• A pipeline-less target (e.g. plain / spec, which version-bump but do

not publish through a pipeline) produces a single ResolvedTarget with pipeline=None.

Target-less pipelines (target: null) are NOT targets and must never be faked into a ResolvedTarget. They are surfaced separately as deploys by :func:partition_pipelines (and MemberContext.deploy_pipelines).

Kept dependency-light: imports only :mod:rlsbl.errors. The caller supplies already-detected targets and already-loaded pipelines, so this module never touches disk or the pipeline registry.

#ResolvedTarget

One publishable (target, pipeline) pair.

Fields: target: the detected target entry (a TargetEntry). path: the target's resolved directory path (same value as MemberContext.target_paths[name] -- carried here so consumers need not re-resolve it). pipeline: the linked pipeline object, or None for a pipeline-less target (version-bump-only targets like plain / spec). publish_mode: the member's effective publish_mode string, carried verbatim from config ("ci" / "none") -- never collapsed to a private boolean. artifact_kind: the pipeline config's artifact value where present (e.g. Go's library / binary, wired in a later phase), or None. Carried as the raw config value. primary: whether this record is the release's primary/registry target. At most one record in a resolved list is primary; the release flow derives the registry name, registry-target instance, and primary path from it. Defaults to False; set by :func:resolve_targets when a primary_name is supplied.

#name

def name(self) -> str

The target's name (convenience for self.target.name).

#_pipeline_target_link

def _pipeline_target_link(pipeline) -> Optional[str]

The pipeline's explicit target link (Phase 6.1's .target attribute).

load_pipelines sets instance.target from the entry's target key (a target name string, or None for a target-less publisher). This is the single point of coupling to that wiring; getattr keeps the resolver from crashing un-actionably on a pipeline built outside load_pipelines.

#partition_pipelines

def partition_pipelines(pipelines: dict) -> tuple[dict, list]

Split loaded pipelines by their target link.

Returns (by_target, deploys) where:

• by_target maps a target name to the list of pipelines linking it

(insertion order preserved -- a target served by multiple pipelines keeps its pipelines in config order).

• deploys is the list of target-less pipelines (target: null), in

config order.

Args:

• pipelines: the dict returned by load_pipelines (name -> pipeline).

#resolve_targets

def resolve_targets(targets: list, pipelines: dict, publish_mode: str, primary_name: Optional[str]=None) -> list

Resolve targets + linked pipelines into a flat list of ResolvedTargets.

One :class:ResolvedTarget is produced per (target, linked pipeline) pair. A pipeline-less target yields a single record with pipeline=None. Target-less pipelines are excluded here (see partition_pipelines / deploys).

When primary_name is given (the release file's include[0]), exactly one record is marked primary=True: the FIRST record (in config / detection order) whose name equals primary_name. When a target is served by multiple pipelines, the records for that target are contiguous and the first is chosen -- a deterministic, config-order choice. If no record matches primary_name (or it is None), nothing is marked; non-release contexts (version sync, workspace checks) leave it unset.

Args:

• targets: detected target entries (list of TargetEntry).
• pipelines: loaded pipelines (dict name -> pipeline, from

load_pipelines).

• publish_mode: the member's effective publish_mode string, carried

verbatim onto every produced record.

• primary_name: the primary/registry target name (include[0] from the

release file), or None when no primary is known.

Raises:

• ConfigError: if a pipeline links a target name absent from targets

(a dangling reference). validate_pipeline_target_links catches this earlier at config-validation time; this is a loud backstop so the resolver never produces a silently-incomplete view from unvalidated input.