Skip to content

Latest commit

 

History

History
293 lines (186 loc) · 10.3 KB

README.md

File metadata and controls

293 lines (186 loc) · 10.3 KB

pbsmrtpipe

(Beta) Workflow Engine for working with PacBio data.

Official Latest Documentation

Circle CI

Features

  • Integrates natively with Pacific Biosciences sequence data
  • Runs Scatter/Gather model to scale bioinformatic analysis
  • Configurable HPC scheduler (SGE, PBS, Slurm)
  • pbcommand Python Tool Contract API to define tasks
  • Simple Bindings model to define pipelines and pipeline templates
  • HTML Reports
  • interfaces with smrtflow

Defining Tasks using Tool Contracts

The pbcommand library can be used to call python functions natively, or generate an CLI interface that can be emit Tool Contract JSON files, or run Resolved Tool Contract JSON files.

An example is shown below using the 'quick' model that allows you to call python functions and define metadata about the task, such as the FileType, or number of processors to use. The below example is included to provide a complete overview of how the system works. For specific details on the ToolContract data model and commandline interface, please see pbcommand for details.

from pbcommand.models import FileTypes
from pbcommand.cli import registry_builder, registry_runner

log = logging.getLogger(__name__)

registry = registry_builder("pbcommand", "python -m pbcommand.cli.examples.dev_quick_hello_world ")


def _example_main(input_files, output_files, **kwargs):
    log.info("Running example main with {i} {o} kw:{k}".format(i=input_files,
                                                               o=output_files, k=kwargs))
    # write mock output files, otherwise the End-to-End test will fail
    xs = output_files if isinstance(output_files, (list, tuple)) else [output_files]
    for x in xs:
        with open(x, 'w') as writer:
            writer.write("Mock data\n")
    return 0


@registry("dev_qhello_world", "0.2.1", FileTypes.FASTA, FileTypes.FASTA, nproc=1, options=dict(alpha=1234))
def run_rtc(rtc):
    return _example_main(rtc.task.input_files[0], rtc.task.output_files[0], nproc=rtc.task.nproc)


@registry("dev_fastq2fasta", "0.1.0", FileTypes.FASTQ, FileTypes.FASTA)
def run_rtc(rtc):
    return _example_main(rtc.task.input_files[0], rtc.task.output_files[0])


if __name__ == '__main__':
    sys.exit(registry_runner(registry, sys.argv[1:]))

The tool contracts can now be emitted to a directory and tool ids can be used in Pipeline Bindings to define the workflow graph.

$> python -m pbcommand.cli.examples.dev_quick_hello_world -o /path/to/my-tool-contracts

Defining a Pipeline

A pipeline template can be defined using a simple micro language to bind outputs to inputs.

See pbsmrtpipe.pb_pipelines_dev for examples.

The bindings are a list of tuples that are used to define the mappings of the output of a task to the input of task.

This will map the 0-th output of pbsmrtpipe.tasks.dev_txt_to_fasta to the first input of task pbsmrtpipe.tasks.dev_filter_fasta.

b1 = [('pbsmrtpipe.tasks.dev_txt_to_fasta:0', 'pbsmrtpipe.tasks.dev_filter_fasta:0')]

Registering a pipeline can be written in python. An entry-point must be defined as the external input to the pipeline.

import logging

from pbsmrtpipe.core import register_pipeline
from pbsmrtpipe.constants import to_pipeline_ns

log = logging.getLogger(__name__)


@register_pipeline("pipelines.pbsmrtpipe.dev_local", "Dev Local Hello Pipeline")
def get_dev_local_pipeline():
    """Simple example pipeline"""
    
    b1 = [('$entry:e_01', 'pbsmrtpipe.tasks.dev_hello_world:0')]

    b2 = [('pbsmrtpipe.tasks.dev_hello_world:0', 'pbsmrtpipe.tasks.dev_hello_worlder:0')]

    b3 = [('pbsmrtpipe.tasks.dev_hello_world:0', 'pbsmrtpipe.tasks.dev_txt_to_fasta:0')]

    b4 = [('pbsmrtpipe.tasks.dev_txt_to_fasta:0', 'pbsmrtpipe.tasks.dev_filter_fasta:0')]

    return b1 + b2 + b3 + b4

Once the pipeline is registered it can be referenced in pipeline-template and will accessible via the CLI using pbsmrtpipe show-templates

The pipeline can by run by referencing id pipelines.pbsmrtpipe.dev_local using --entry-point e_01:/path/to/file.txt as a CLI arg.

CLI of running pbsmrtpipe

> pbsmrtpipe pipeline my_workflow.xml --output=/path/to/output_dir entry_01:/path/to/file.txt entry_id_02:/path/to/file2.txt

> pbsmrtpipe pipeline-id pbsmrtpipe.pipelines.dev_pipeline --output=/path/to/output_dir entry_01:/path/to/file.txt entry_id_02:/path/to/file2.txt

> pbsmrtpipe show-templates # Displays all registered pipelines templates

> pbsmrtpipe show-template-details [template id] # Displays pipeline template details

Help

(mk_pbsmrtpipe_test)pbsmrtpipe $> pbsmrtpipe --help
usage: pbsmrtpipe [-h] [-v]
                  {pipeline,pipeline-id,task,show-templates,show-template-details,show-tasks,show-task-details,show-workflow-options}
                  ...

Pbsmrtpipe workflow engine

positional arguments:
  {pipeline,pipeline-id,task,show-templates,show-template-details,show-tasks,show-task-details,show-workflow-options}
                        commands
    pipeline            Run a pipeline using a pipeline template or with
                        explict Bindings and EntryPoints.
    pipeline-id         Run a registered pipline by specifiying the pipline
                        id.
    task                Run Task by id.
    show-templates      List all pipeline templates. A pipeline 'id' can be
                        referenced in your my_pipeline.xml file using
                        '<import-template
                        id="pbsmrtpipe.pipelines.my_pipeline_id" />. This can
                        replace the explicit listing of EntryPoints and
                        Bindings.
    show-template-details
                        Show details about a specific Pipeline template.
    show-tasks          Show completed list of Tasks by id
    show-task-details   Show Details of a particular task by id (e.g.,
                        'pbsmrtpipe.tasks.filter_report'). Use 'show-tasks' to
                        get a completed list of registered tasks.
    show-workflow-options
                        Display all workflow level options that can be set in
                        <options /> for preset.xml

optional arguments:
  -h, --help            show this help message and exit
  -v, --version         show program's version number and exit

Development and Testing Pipelines

"testkit" is the test framework for testing pipelines that is included with pbsmrtpipe.

Examples of leveraging the testkit are included in the pbsmrtpipe/testkit-data dir.

The tests are configurable via an INI/CFG file (e.g., testkit.cfg). This file defines the input entry points, pipeline id and tests to run after the pipeline has completed.

Example from testkit-data/dev_01/testkit.cfg

[pbsmrtpipe:pipeline]

id = dev_01
description = Dev example
author = mkocher

pipeline_xml = workflow_id.xml
preset_xml = preset.xml

[entry_points]
e_01 = input.txt


[tests]
# Tests can be loaded from any python module
# specifically, Any TestBase subclass in pbsmrtpipe.teskit.core.test_zero will be loaded
pbsmrtpipe.testkit.core = test_zero, test_resources, test_datastore

Executing the test

$> pbtestkit-runner --debug testkit.cfg

This will run the specific pipeline via a subprocess call to pbsmrtpipe in job_output dir and run tests on the results.

To re-run only the test assertions, run:

$> pbtestkit-runner  --only-tests testkit.cfg

To re-run from scratch, delete the job_output dir and re-run

$> pbtestkit-runner testkit.cfg

For development and testing, there are make targets to automate common tasks.

Uninstall and Install pbsmrtpipe

make install

Test defined Tasks

make test-tasks

Test defined Pipelines

make test-pipelines

Test defined Chunk Operators

make test-operators

Test All core defined tasks, pipelines and chunk operators

make test-sanity

Running test-data Integration Pipelines with testkit

make test-dev

Run the entire test suite (unittests and integration tests) This should be run before every pull request.

make test-suite

Clean All

make clean-all

Pacbio End to End Pipeline Testing Model

Testing at the commandline level is required. If the tests pass at the commandline level via testkit, there's a high probability the pipeline (with the same data set and options) will run successfully from the SMRTLink UI.

Having tests at the commandline level allow quicker iteration and will avoid debugging the entire software stack to debug pipeline failures.

Summary of the end-to-end model for Pacbio Tool Developers Workflow

  • Update exe and tool within your specific tool repo
  • TEST: Update unittests within your tool repo
  • Update JSON tool contract within the pbsmrtpipe repo (pbsmrtpipe/pbsmrtpipe/registered_tool_contracts_sa3)
  • Update or define new pipeline(s) within pbsmrtpipe.pb_pipelines.*.py
  • TEST: Update or add new testkit pipeline test to /depot/software/smrtanalysis/siv/testkit-jobs/sa3_pipelines in perforce
  • After the next tools and service bundle build, the updated tasks and pipeline should will be accessible via smrtlink-bihourly

Every production pipeline MUST have at least one testkit job defined.

Todo: Add comments on running pipeline service tests from testkit.cfg using pbtestkit-service-runner

DISCLAIMER

THIS WEBSITE AND CONTENT AND ALL SITE-RELATED SERVICES, INCLUDING ANY DATA, ARE PROVIDED "AS IS," WITH ALL FAULTS, WITH NO REPRESENTATIONS OR WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, ANY WARRANTIES OF MERCHANTABILITY, SATISFACTORY QUALITY, NON-INFRINGEMENT OR FITNESS FOR A PARTICULAR PURPOSE. YOU ASSUME TOTAL RESPONSIBILITY AND RISK FOR YOUR USE OF THIS SITE, ALL SITE-RELATED SERVICES, AND ANY THIRD PARTY WEBSITES OR APPLICATIONS. NO ORAL OR WRITTEN INFORMATION OR ADVICE SHALL CREATE A WARRANTY OF ANY KIND. ANY REFERENCES TO SPECIFIC PRODUCTS OR SERVICES ON THE WEBSITES DO NOT CONSTITUTE OR IMPLY A RECOMMENDATION OR ENDORSEMENT BY PACIFIC BIOSCIENCES.