Runners

Overview

annopi-ts does not hardcode ata or annotask binary names. Instead, it resolves global runner paths from merged pipeline deps and generates runcmd strings in tasks.yml.

Why no local subcommand?

Both ata and annotask accept the same local interface:

runner -i script.sh -l <lines> -t <threads>

• ata always runs locally with -i
• annotask defaults to local mode when no subcommand is given

So in local mode, annopi-ts generates commands without a local subcommand. This keeps ata and annotask interchangeable for local execution.

For cluster submission, only annotask supports the qsubsge subcommand:

annotask qsubsge -i script.sh -l N -t M --cpu C --queue Q --project P

Configuring paths in pipeline.yml

Set global software paths under deps:

name: my_pipeline
version: 1.0.0

deps:
  ata: /Volumes/data/GOPATH/bin/ata
  annotask: /opt/annotask

tasks:
  local_task:
    command: echo hello
    resources:
      executor: local

  cluster_task:
    command: echo cluster
    resources:
      executor: qsubsge
      cpu: 4
      queue: all.q

Resolution rules

localRunner    = deps.ata ?? deps.annotask ?? defaultAta
annotaskRunner = deps.annotask ?? defaultAnnotask

• localRunner is used for all executor: local tasks
• annotaskRunner is used for all executor: qsubsge tasks

If deps.ata is not set but deps.annotask is, annotask serves as the local runner too.

Default discovery

When deps are not set, annopi-ts looks for:

1. /Volumes/data/GOPATH/bin/ata and /Volumes/data/GOPATH/bin/annotask
2. Binaries named ata / annotask on PATH
3. Falls back to bare command names ata / annotask

Example tasks.yml output

tasks:
  1-0-cellranger:
    runcmd: annotask qsubsge -i shell/1-0-cellranger.sh -l 1 -t 10 --cpu 8 --queue sci.q --project DEMO
    depends: []
  2-0-qc:
    runcmd: /Volumes/data/GOPATH/bin/ata -i shell/2-0-qc.sh -l 4 -t 10
    depends: [1-0-cellranger]

Note: local runcmd has no local keyword and no --project flag.

Related tools

Install ata for local development. annotask requires DRMAA headers to build the qsubsge module.

Other deps

deps in pipeline.yml also holds tool paths used in task commands (not runners):

deps:
  cellranger: /opt/cellranger-7.0.0
  reference: /data/refdata-gex-GRCh38-2020-A

These are expanded into ${deps.cellranger} etc. during conf. Runner keys (ata, annotask) are reserved for execution path resolution.