willdo

Crates.iowilldo
lib.rswilldo
version0.0.1
sourcesrc
created_at2024-11-26 18:01:49.011447
updated_at2024-11-26 18:01:49.011447
descriptionTask manager with DAG
homepage
repositoryhttps://gitlab.com/BrightOpen/willdo
max_upload_size
id1462101
size197,873
Robert Cutajar (jocutajar)

documentation

https://docs.rs/willdo

README

WillDo

Will do some work when the conditions are ripe!

It is a generic task manager for running jobs with a directed acyclic graph (DAG) dependency resolution. Configuration is loaded from simple yaml files:

job: baking
script:
    - echo oven 250
    - echo insert yummy
    - sleep 3600
    - echo oven off
    - echo take yummy
---
job: oops
when:
    - done: baking
      code: 
        min: 1 # that is non-zero, baking failed
script: echo alarm!
---
job: cleanup
when:
    - any:
        - done: baking # assuming 0 - success
        - done: oops
script: echo tidy up

Jobs are effectively scheduled when their parent dependencies (when) are satisfied. Dependencies can be conditional with a simple code arithmetic conditions and combined into complex rules (all/any). By default, all dependencies must be satisfied.

Triggers (then) are opposite - child - relations.

The crate includes a subprocess commander behind a feature of same name, so any shell or repl can be utilized immediately to run your scripts. The default commander is () which does nothing.

You can configure an interpretter:

interpretter: sh
provider: subprocess
executable: dash
---
job: making
runs: sh
script:
    - test -f .skip && echo skipped || echo make

In order to discern the outcome of a script command, the commander must provide feedback to send back the numeric code of the last operation. This is not easily done in a subprocess and is achieved by printing a uniquely stamped string that is parsed from child process outputs.

Usage

  1. Use it as a crate in your project - see examples and docs to get started. The Graph implements Display with a digraph output which can be rendered with graphviz to visualize the state. You can provide your own Commander to interpret job scripts.

  2. Build the examples that can run jobs as a CLI, you could install the examples with cargo, but I would not recommend it at this stage as the interfaces are not stable.

TODO

  • Pass resources between jobs, such as exit codes, outputs, variables.
  • Preserve and load job state so we can pick up where we left off.
  • An official and stable CLI.
  • Later: other configuration sources.
  • Later: fancy interactive TUI.
  • Maybe: more ways to work with the graph.
Commit count: 0

cargo fmt