Glossary ​

This page defines the terms used throughout the Sprocket documentation.

Attempt ​

One try to execute a task. Sprocket stores each attempt in a numbered attempts/<n>/ directory with its command, standard output, standard error, and working directory. See Retries.

Baseline ​

A sprocket-baseline.toml file that records existing diagnostics so sprocket check and sprocket lint can suppress them while still reporting new diagnostics. Stale entries cause the check to fail. See Baselines.

Call ​

A use of a task inside a workflow. During a workflow run, Sprocket gives each task call its own directory under calls/, with execution attempts below it. See Workflow runs.

Call cache ​

The persistent cache that lets Sprocket reuse the result of a previously successful task execution with matching inputs and execution context. Call caching is off by default, and only a task that succeeds on its first attempt is eligible. See Call cache.

Container ​

The isolated environment in which Sprocket runs a task. A task selects a container image through its WDL container requirement or runtime key, and the execution backend starts the image. See Containers.

Diagnostic ​

A message produced while Sprocket parses, analyzes, lints, or formats WDL. A diagnostic has a severity such as error, warning, or note; check and lint options control which severities are shown or cause failure. See sprocket check and sprocket lint.

Digest ​

A value Sprocket uses to detect file changes for the call cache. A weak digest uses file metadata, a strongish digest uses the metadata and first 10 MiB of content, and a strong digest hashes the full contents. See Content digests.

Document ​

A WDL source file or URL that declares a WDL version and can contain imports, structs, tasks, and a workflow. Sprocket can discover, analyze, and load a document together with its imported dependencies. See Inputs and targets.

Execution engine ​

The complete system that evaluates and runs WDL: Sprocket's orchestration engine together with one or more execution runtimes. See the Execution model.

Execution runtime / backend ​

The component that carries out dispatched task work in a particular environment. The current command and configuration reference uses execution backend for Docker, TES, LSF with Apptainer, and Slurm with Apptainer implementations. See Execution backends.

Index ​

An optional, user-organized tree of relative symlinks under an output directory's index/ directory. sprocket run --index-on <PATH> links a run's outputs.json and eligible file and directory outputs at a stable path without copying them. See Output indexing.

Input ​

A named value supplied to a task or workflow. A fully qualified input name combines the target and input names, such as main.name; when you pass --target main, Sprocket can prefix unqualified run inputs with main. See Inputs and targets.

_latest ​

A symlink under runs/<target>/ that points to the most recent run directory for that target. If Windows cannot create the symlink, the run continues without it. See The _latest symlink.

Lint rule ​

An opinionated check from wdl-lint that promotes idiomatic, maintainable WDL. Lint rules have tags and are enabled by sprocket lint or sprocket check --lint. See Lint and validation rules.

Lock file ​

The experimental sprocket.lock file written by sprocket dev lock. It records container image manifest checksums and the time they were resolved. Sprocket 0.31.0 does not yet consume this file during a run. See sprocket dev lock.

LSP / analyzer ​

The Language Server Protocol service started by sprocket analyzer. Editor clients use it for diagnostics, navigation, references, renaming, formatting, semantic highlighting, and snippets. See sprocket analyzer.

Module ​

An experimental reusable WDL project described by module.json. Its resolved dependencies are recorded separately in module-lock.json. See sprocket dev module.

Orchestration engine ​

The part of Sprocket that evaluates workflows, stages data, connects inputs and outputs, schedules and monitors calls, and dispatches task work to an execution runtime. See the Execution model.

Output ​

A named value produced by a task or workflow. Sprocket serializes run outputs to outputs.json; the output index can also link file and directory outputs. See Run contents.

Output directory ​

The self-contained directory where Sprocket stores execution data. It defaults to ./out and contains the provenance database, immutable runs/ hierarchy, and optional index/ hierarchy. See Output directory.

Provenance database ​

The sprocket.db SQLite database in the output directory. It records sessions, runs, and task executions, using paths relative to the database so the output directory remains portable. Its schema is internal and may change in any release, so query it through Sprocket's commands or REST API rather than directly. See Provenance database.

Retry ​

Another attempt after a task fails. Configuration and a task's maxRetries value control how many retries Sprocket permits; --disable-retries disables them for one local run. See Retries.

Run ​

One execution of a selected task or workflow, with its inputs, outputs, and status. Sprocket stores each run in a timestamped directory under runs/<target>/. See Runs and the index.

Session ​

A group of related run submissions in the provenance database. Each sprocket run invocation creates a session; one sprocket dev server start process creates a session shared by the runs submitted to that server. See Sessions.

SIF ​

A Singularity Image Format (.sif) container image used by Apptainer-based execution backends. Sprocket can keep images per run under apptainer-images/ or use a configured shared image cache. See Containers.

.sprocketignore ​

An ignore file that controls WDL document discovery. Its patterns work like .gitignore patterns and apply to the file's directory and its descendants. See Ignoring WDL files and directories.

sprocket.toml ​

Sprocket's configuration file. Sprocket merges configuration from system, executable-adjacent, working-directory, environment, and command-line sources according to its configuration load order. The configuration reference lists every key.

Target ​

The task or workflow selected for an operation such as running, validating, or generating an input template. Sprocket can often infer a run target from fully qualified inputs, a document's workflow, or its sole task; use --target when selection is ambiguous. See Targets.

Task ​

A WDL unit of computation with inputs, a command, requirements or runtime settings, and optional outputs. A workflow invokes a task through a call, and Sprocket sends the resulting work to an execution backend. See the Execution model.

Validation rule ​

A WDL form or analysis check from wdl-analysis. Validation rules run under both sprocket check and sprocket lint and do not expose lint tags. See Lint and validation rules.

WDL ​

The Workflow Description Language, the open workflow language that Sprocket parses, analyzes, and executes. See Open, tailored standard.

Workflow ​

A WDL graph that declares inputs and outputs and connects task calls and expressions. Sprocket evaluates the workflow locally and dispatches ready task work to its configured execution backend. See the Execution model.