from photoflow.core.models import ClusterMember
from app.clustering import _epoch, _median


def coverage_members(candidate, assets_by_id, *, outlier_factor: float = 8.0) -> list:
    member_ids = [i for i in candidate.member_ids if i in assets_by_id]
    members = sorted((assets_by_id[i] for i in member_ids), key=lambda a: a.taken_at)
    epochs = [_epoch(a.taken_at) for a in members]
    gaps = [b - a for a, b in zip(epochs, epochs[1:])]
    base = _median(gaps) if gaps else 0.0

    out = []
    for idx, a in enumerate(members):
        is_outlier = False
        if candidate.seed_tag and base > 0:
            left = epochs[idx] - epochs[idx - 1] if idx > 0 else 0
            right = epochs[idx + 1] - epochs[idx] if idx < len(members) - 1 else 0
            nearest = min([g for g in (left, right) if g > 0], default=0)
            if nearest > outlier_factor * base:
                is_outlier = True
        out.append(ClusterMember(
            cluster_id=0, immich_id=a.immich_id,
            member_confidence=candidate.confidence, is_outlier=is_outlier,
            included=True, flagged_coverage=False))

    if candidate.seed_tag and candidate.start_at and candidate.end_at:
        mset = set(member_ids)
        for a in sorted(assets_by_id.values(), key=lambda x: x.taken_at):
            if a.immich_id in mset:
                continue
            if candidate.start_at <= a.taken_at <= candidate.end_at:
                out.append(ClusterMember(
                    cluster_id=0, immich_id=a.immich_id, member_confidence=0.0,
                    is_outlier=False, included=False, flagged_coverage=True))
    return out