Struct CheckpointPruner 

pub struct CheckpointPruner { /* private fields */ }

Applies a RetentionPolicy to a collection of CheckpointRecords and determines which records should be deleted.

Usage

use ipfrs_tensorlogic::checkpoint_manager::{CheckpointRecord, CheckpointPruner, RetentionPolicy};

let records = vec![
    CheckpointRecord { id: "cp_1".into(), cid: "Qm1".into(), round: 1,
        created_at_ms: 0, size_bytes: 100, crc32: 0, is_pinned: false },
    CheckpointRecord { id: "cp_2".into(), cid: "Qm2".into(), round: 2,
        created_at_ms: 0, size_bytes: 100, crc32: 0, is_pinned: false },
];
let policy = RetentionPolicy { keep_last_n: 1, ..Default::default() };
let mut pruner = CheckpointPruner::new(records, policy);
let to_delete = pruner.prune();
assert_eq!(to_delete.len(), 1);
assert_eq!(to_delete[0].id, "cp_1");

Implementations

impl CheckpointPruner

pub fn new(records: Vec<CheckpointRecord>, policy: RetentionPolicy) -> Self

Create a new pruner. Records are sorted by round ascending.

pub fn with_now_ms(self, now_ms: u64) -> Self

Override the “current time” used for age-based pruning decisions.

If not called, now_ms defaults to 0, which means every checkpoint has an age ≥ 0 and the min_age_ms guard applies correctly.

pub fn prune(&mut self) -> Vec<CheckpointRecord>

Apply the retention policy and return the list of records to delete.

After the call the pruner’s internal record list contains only the surviving records; repeated calls therefore return an empty list.

pub fn surviving_count(&self) -> usize

Number of records currently held by the pruner (i.e. after any pruning).

pub fn total_bytes(&self) -> u64

Sum of size_bytes across all records currently held by the pruner.

pub fn pinned_count(&self) -> usize

Number of pinned records currently held by the pruner.

pub fn records(&self) -> &[CheckpointRecord]

Read-only access to the surviving records.