Streamlining Design and Deployment of Complex Workflows

This framework allows you to design and implement complex pipelines, and deploy them on your institution's computing cluster. This has been built keeping in mind the needs of bioinformatics workflows. However, it is easily extendable to any field where a series of steps (shell commands) are to be executed in a (work)flow.

Build Status cran downloads

Latest documentation:

Flowr framework allows you to design and implement complex pipelines, and deploy them on your institution's computing cluster. This has been built keeping in mind the needs of bioinformatics workflows. However, it is easily extendable to any field where a series of steps (shell commands) are to be executed in a (work)flow.


  • No new syntax or language. Put all shell commands as a tsv file called flow mat.
  • Define the flow of steps using a simple tsv file (serial, scatter, gather, burst...) called flow def.
  • Works on your laptop/server or cluster (/cloud).
  • Supports multiple cluster computing platforms (torque, lsf, sge, slurm ...), cloud (star cluster) OR a local machine.
  • One line installation (install.packages("flowr"))
  • Reproducible and transparent, with cleanly structured execution logs
  • Track and re-run flows
  • Lean and Portable, with easy installation
  • Fine grain control over resources (CPU, memory, walltime, etc.) of each step.



A few lines, to get started

## Official stable release from CRAN (updated every other month)
## visit for more details
install.packages("flowr",  repos = "")
# or a latest version from DRAT, provide cran for dependencies
install.packages("flowr", repos = c(CRAN="", DRAT=""))
library(flowr) ## load the library
setup() ## copy flowr bash script; and create a folder flowr under home.
## Run an example pipeline
flowr run x=sleep_pipe platform=local execute=TRUE

Example pipelines inst/pipelines


  • For a quick overview, you may browse through, these introductory slides.
  • The overview provides additional details regarding the ideas and concepts used in flowr
  • We have a tutorial which can walk you through creating a new pipeline
  • Additionally, a subset of important functions are described in the package reference page
  • You may follow detailed instructions on installing and configuring
  • You can use flow creator, a shiny app to aid in designing a shiny new flow. This provides a good example of the concepts


This package is under active-development, you may watch for changes using the watch link above.


Please feel free to raise a github issue with questions and comments.


  • Jianhua Zhang
  • Samir Amin
  • Roger Moye
  • Kadir Akdemir
  • Ethan Mao
  • Henry Song
  • An excellent resource for writing your own R packages:


title: "flowr" subtitle: "Streamlining Workflows" author: Sahil Seth date: "r Sys.Date()" output: html_document packagedocs: toc: true navpills: |

  • Overview
  • Install
  • Tutorial
  • Help
  • News
  • Github
  • brand: |- flowr icon copyright: Licence MIT source: "" ---

    flowr 0.9.10 (dates)


    tl;dr (summary of changes)

    • Flowr Rscript gets further enhancements, taking advantage of improved funr
    • run function now accepts paths.
    # 'cherries' version
    cd <path to pipelines>
    flowr x=mypipline
    # 'dates' version
    flowr x=<path to pipelines>/mypipline
    • Previously, flowr expected a specific structure, now using ~/.flowr.conf, one may specify their own structure - enabling flexibility.
    # 'cherries' version
    ├── conf
    │   ├── flowr.conf
    ├── pipelines
    │   ├── sleep_pipe.R
    ├── runs
    # 'dates' version (backward compatible)
    # this file controls the location of these folders:
    flow_base_path	~/flowr
    flow_conf_path	{{flow_base_path}}/conf  ## custom configuration files
    flow_run_path	~/flowr/runs  ## home of all executed flows, you may change this
    flow_pipe_paths	~/flowr/pipelines,<add new paths...>
    • a few bug fixes in to_flow
    • several other cosmetic changes to messages

    flowr (cherries)


    tl;dr (summary of changes)

    • Better handling of multiple flows in terms of running and re-running.
    • Nicer and cleaner messages.
    • Add two additional lines in flowr.conf (modules_cmds and local_cores), after upgrading the package. Everything else is compatible with previous versions.

    additions/changes to flowr.conf file

    • New: option local_cores, which determines (max) number of cores to use when running local jobs.

    • New: flow_pipe_paths now supports multiple paths, seperated by comma. The fetch_pipes() would split the vector at commas.

    • IMP: New version needs additional components in the flowr.conf file

    • New: Now you can add a module_cmds variable to the config file, and this will be prefixed in all script of the pipeline. An example could be:

    # version >=
    # max number of cores to use when running on a local server
    local_cores 4
    # default module of a pipeline
    # version >=
    module_cmds ''
    # examples: one may define all modules used in a pipeline here, 
    # further one may specify any other command which should be run before 
    # script executes.
    #module_cmds 'module load samtools;export PATH=$PATH:/apps/bin'

    addition/changes to status()

    • New: status gets a new argument to turn off progress bar if needed.
    • New enhanced get_wds/status, so that if current wd contains a flow_details file, status is shown for this folder and not sub-folder(s).
    ## now this works well !
    flowr status x=.

    addition/changes to run() and rerun() functions

    • New: run function now accepts a custom configuration [conf], parameter. See help(flowr::run) for more details. This enables one, to specify custom parameters used for that pipeline.

    • New: Now rerun() supports multiple folders. Basically, one may specify a parent folder which has multiple flowr runs and ask it to re-run ALL of them again, from a specific intermediate step.

    • New: Flowr creates a new folder if there are multiple samples in the flowmat; basically containerizes the run, keeping the logs clean and debugging life easier.

    • New: run() now supports, re-running as well. i.e. One would generate a new set of commands etc. but execute in the previous folder; possibly from a inter-mediate step (experimental feature).

    other changes

    • New: to_flowdef() can now guess submission and dependency types (experimental feature).
    • IMP: to_flowdef now adds a parameter nodes, to enable specifying number of nodes required per-job.
    • IMP: opts_flow$get replaces get_opts, for reliability etc. Also this closely follows how knitr options are set.
    • fixed bugs in documentation (changed the formatting of output messages).

    flowr 0.9.8 (blueberries)


    • Modified the output of status function, to add a status column. Specifically, this uses information from other columns and summarizes whether a specific step is pending, processing, completed or errored.
    |                | total| started| completed| exit_status|status     |
    |001.alnCmd1     |   109|     109|       109|           0|completed  |
    |007.markCmd     |     3|       3|         0|           0|processing |
    • []

      • Switched default value of [email protected] to "created".
      • When using status on several folders, it used to be a little cluttered. Have added spaces, so that this looks prettier now.
    • Introducing verbose levels:

      One can set the level of verboseness using opts_flow$set(verbose=2). Where the level may be 0, 1, 2.... Level 1 is good for most purposes, where as, level 0 is almost silent, producing messages only when neccessary. While level 2 is good when developing a new pipeline, additional details useful for debugging are provided by level 3.

    • Detailed checking of flowdef

    checking if required columns are present...
    checking if resources columns are present...
    checking if dependency column has valid names...
    checking if submission column has valid names...
    checking for missing rows in def...
    checking for extra rows in def...
    checking submission and dependency types...
        jobname	prev.sub_type --> dep_type --> sub_type: relationship
        1: aln1_a	none --> none --> scatter 
        2: aln2_a	scatter --> none --> scatter 
        3: sampe_a	scatter --> serial --> scatter rel: complex one:one
        4: fixrg_a	scatter --> serial --> scatter rel: complex one:one
        5: merge_a	scatter --> gather --> serial rel: many:one
        6: markdup_a	serial --> serial --> serial rel: simple one:one
        7: target_a	serial --> serial --> serial rel: simple one:one
        8: realign_a	serial --> burst --> scatter rel: one:many
        9: baserecalib_a	scatter --> serial --> scatter rel: complex one:one
        10: printreads_a	scatter --> serial --> scatter rel: complex one:one
    • rerun []:

      • Previously one could specify a starting point from where a re-run flow would initiate execution. Now one may also specify an arbitary number of of steps to re-run using select and ignore; which may need to run again.
    • job killing and submission now sport a progress bar:

      • |============================================================ | 70%
      • This is especially useful flows with thousands of jobs
    • Fixed 2 important bugs in, template file, where it was missing the -n argument.

    • Now the status function has a new argument use_cache. If enabled, it skips fetching statuses of previously completed jobs; this really speeding things up.

      • Also we have added a progress bar to show the status of this summarization.
    • Several detailed changes to the documentation.

    flowr 0.9.7 (apples)

    2015-08-22 (

    • This release adds and changes functionality of several functions.
    • A new function run(), creates and submits a pipeline. Specifically it follows the following steps:
      • One supplies the name of the pipeline, which is used to fetch the pipeline using: fetch_pipe()
      • create flowmat by running a function called mypipeline(), mypipeline is the name of the pipeline.
      • load configuration file, with paths to tools etc using load_conf()
      • fetch the flow definition using fetch_pipe, and load it using as.flowdef()
      • Further, create a flow object using to_flow()
      • Finally, submit to the cluster, submit_flow()
    • kill(): now a S3 functions, and operates on both a flow object and flow_wd folder
    • check(): Now works on flowdef and flowmat
    • as.flowmat(), as.flowdef(): easy ways to fetch and check these tables
    • fetch() along with fetch_pipes() and fetch_conf() simplify finding files
    • Reduce function overload, moving several functions a seperate params pkg
      • moved read_sheet(), write_sheet()
      • moved get_opts(), set_opts()
      • moved .load_conf() load_conf()
      • Here is a link to params package
      • kable function is now a part of params, that removes the dependency to knitr package
    • plot_flow: supports flowdef


    • [] 2015-07-23
      • Using PBS/LSF script instead of one line commands
      • Format now a script, parsed using whisker
      • to_flowmat() and to_flowdef() introduced; creating of these tables simplified
      • Refactored code for plot_flow, handles multiple dependencies beautifully
    • [] 2015-07-08
      • add to_flow()
      • kill_flow()
      • supports moab
      • burst as dependency type
      • node can be character
      • final job names changed to: sprintf("%s_%s-%s", basename([email protected]_path), [email protected], i)
      • Using PBS/LSF script instead of one line commands
      • Format now a script, parsed using whisker



    • add read_sample_sheet
    • clean package dependencies
    • []
      • switch cat to messages
      • added to_flow
    • []
      • Added more examples
      • Update Vignette with more examples
      • squash issues in DESCRIPTION
    • []
      • Added some vignettes for a simple and comples example
      • satiate CRAN checks



    • Added several new functions kill_flow(), which kills the flow.
    • Added experimental rerun_flow which still needs some work
    • updated documentation to satiate R CMD check
    • now job() checks for a dependency_type is previous job is provided
    • Added some slides inst
    • Integrated documentation with Sphnix
    • submit_flow, now generates flow_details.txt, flow.pdf, flow_status.txt
    • status() updates: flow_details.txt and flow_status.txt
    • save flow.rda files with all details on the flow
    • update flow details with job exit status, to be used in rerun


    • add more examples
    • update and cleanup documentation
    • integrate with sqlite with status
    • add example pipelines
    • assign UUID to each job/flow. and use that to track them.
    • Work on test_queue()
    • Add recipies

    flowr 0.85


    • get_flow_details():
    • Attach a job_number
    • Attach a flow level unique ID
    • update flow details with exited jobs
    • Create trigger upon start
    • add started/pending column in details and summary
    • plot_flow(): make plot type 1 as default
    • get_flow_status(): -exit_status: reports number of exited jobs (not sum of exit codes)


    Loosely, following fruits A-Z.

    Reference manual

    It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.


    0.9.11 by Sahil Seth, a year ago

    Report a bug at

    Browse source code at

    Authors: Sahil Seth [aut, cre]

    Documentation:   PDF Manual  

    Task views: High-Performance and Parallel Computing with R, Reproducible Research

    MIT + file LICENSE license

    Imports diagram, whisker, tools, readr

    Depends on methods, params, utils

    Suggests reshape2, knitr, testthat, funr, rmarkdown

    Enhances ggplot2, openxlsx, pacman, cowplot

    See at CRAN