Struct Cache 

pub struct Cache {
    pub root: PathBuf,
    /* private fields */
}

The cache for one run; its workspace snapshot is taken on first use.

Fields

root: PathBuf

Implementations

impl Cache

pub fn prefetch_warm( &self, workspace: &Workspace, graph: &TaskGraph, prepared: &BTreeMap<String, PreparedTask>, restores: impl Fn(&str) -> bool, )

Starts fetching, for each task’s warm entry that asks for it, the remote save its restore would fetch, so it is here by the time the task runs. An entry whose paths are on disk, or that a save here would restore, fetches nothing, and a group shared by several tasks is fetched once. A few are fetched at once, those of the tasks with the fewest dependencies to wait for first. Only the tasks restores accepts, which restore warm state as they start, are fetched for; a task whose result is cached never uses what was fetched for it.

pub fn abandon_prefetches(&self)

Stops fetching ahead without waiting for the fetches under way, as a cancelled run does, which no task will take.

impl Cache

pub fn new(root: PathBuf) -> Self

pub fn for_observation(workspace: &Workspace) -> Self

Records verified executions without configuring or using result stores.

pub fn for_workspace( workspace: &Workspace, environment: &BTreeMap<OsString, OsString>, ) -> Self

The workspace’s cache, with the remote store its nx.json s3 or qk:storeServer key configures, if any. Credentials and modes are read from environment, which includes the workspace’s dotenv files. A remote store that cannot be used is reported and left out; the local cache still works.

pub fn cold(self) -> Self

This cache with every task run from scratch, as --cold asks: see Cache::cold’s field.

pub fn finish(&self, workspace: &Workspace)

Saves what later runs reuse and waits for background saves, uploads and retired-output cleanup, reporting failures.

pub fn run( &self, workspace: &Workspace, graph: &TaskGraph, task: &Task, prepared: &PreparedTask, cancelled: &AtomicBool, ) -> Result<TaskResult>