Processes¶

Scripts à la carte¶

The process script is interpreted by Nextflow as a Bash script by default, but you are not limited to it.

You can use your favourite scripting language (e.g. Perl, Python, Ruby, R, etc), or even mix them in the same pipeline.

A pipeline may be composed by processes that execute very different tasks. Using Nextflow you can choose the scripting language that better fits the task carried out by a specified process. For example for some processes R could be more useful than Perl, in other you may need to use Python because it provides better access to a library or an API, etc.

To use a scripting other than Bash, simply start your process script with the corresponding shebang declaration. For example:

process perlStuff {

    """
    #!/usr/bin/perl

    print 'Hi there!' . '\n';
    """

}

process pyStuff {

    """
    #!/usr/bin/python

    x = 'Hello'
    y = 'world!'
    print "%s - %s" % (x,y)
    """

}

Conditional scripts¶

Complex process scripts may need to evaluate conditions on the input parameters or use traditional flow control statements (i.e. if, switch, etc) in order to execute specific script commands, depending on the current inputs configuration.

Process scripts can contain conditional statements by simply prefixing the script block with the keyword script:. By doing that the interpreter will evaluate all the following statements as a code block that must return the script string to be executed. It's much easier to use than to explain, for example:

seq_to_align = ...
mode = 'tcoffee'

process align {
    input:
    file seq_to_aln from sequences

    script:
    if( mode == 'tcoffee' )
        """
        t_coffee -in $seq_to_aln > out_file
        """

    else if( mode == 'mafft' )
        """
        mafft --anysymbol --parttree --quiet $seq_to_aln > out_file
        """

    else if( mode == 'clustalo' )
        """
        clustalo -i $seq_to_aln -o out_file
        """

    else
        error "Invalid alignment mode: ${mode}"

}

In the above example the process will execute the script fragment depending on the value of the mode parameter. By default it will execute the tcoffee command, changing the mode variable to mafft or clustalo value, the other branches will be executed.

Template¶

Process script can be externalised by using template files which can be reused across different processes and tested independently by the overall pipeline execution.

A template is simply a shell script file that Nextflow is able to execute by using the template function as shown below:

process template_example {

    input:
    val STR from 'this', 'that'

    script:
    template 'my_script.sh'

}

Nextflow looks for the my_script.sh template file in the directory templates that must exist in the same folder where the Nextflow script file is located (any other location can be provided by using an absolute template path).

The template script can contain any piece of code that can be executed by the underlying system. For example:

#!/bin/bash
echo "process started at `date`"
echo $STR
:
echo "process completed"

Shell¶

The shell block is a string statement that defines the shell command executed by the process to carry out its task. It is an alternative to the Script definition with an important difference, it uses the exclamation mark ! character as the variable placeholder for Nextflow variables in place of the usual dollar character.

In this way it is possible to use both Nextflow and Bash variables in the same piece of code without having to escape the latter and making process scripts more readable and easy to maintain. For example:

process myTask {

    input:
    val str from 'Hello', 'Hola', 'Bonjour'

    shell:
    '''
    echo User $USER says !{str}
    '''

}

In the above trivial example the $USER variable is managed by the Bash interpreter, while !{str} is handled as a process input variable managed by Nextflow.

Native execution¶

Nextflow processes can execute native code other than system scripts as shown in the previous paragraphs.

This means that instead of specifying the process command to be executed as a string script, you can define it by providing one or more language statements, as you would do in the rest of the pipeline script. Simply starting the script definition block with the exec: keyword, for example:

x = Channel.from( 'a', 'b', 'c')

process simpleSum {
    input:
    val x

    exec:
    println "Hello Mr. $x"
}

Will display:

Hello Mr. b
Hello Mr. a
Hello Mr. c

Input of generic values¶

The val qualifier allows you to receive data of any type as input. It can be accessed in the process script by using the specified input name, as shown in the following example:

num = Channel.from( 1, 2, 3 )

process basicExample {
  input:
  val x from num

  "echo process job $x"

}

In the above example the process is executed three times, each time a value is received from the channel num and used to process the script. Thus, it results in an output similar to the one shown below:

process job 3
process job 1
process job 2

When the val has the same name as the channel from where the data is received, the from part can be omitted. Thus the above example can be written as shown below:

num = Channel.from( 1, 2, 3 )

process basicExample {
  input:
  val num

  "echo process job $num"

}

Input of files¶

The file qualifier allows the handling of file values in the process execution context. This means that Nextflow will stage it in the process execution directory, and it can be access in the script by using the name specified in the input declaration. For example:

proteins = Channel.fromPath( '/some/path/*.fa' )

process blastThemAll {
  input:
  file query_file from proteins

  "blastp -query ${query_file} -db nr"

}

In the above example all the files ending with the suffix .fa are sent over the channel proteins. Then, these files are received by the process which will execute a BLAST query on each of them.

When the file input name is the same as the channel name, the from part of the input declaration can be omitted. Thus, the above example could be written as shown below:

proteins = Channel.fromPath( '/some/path/*.fa' )

process blastThemAll {
  input:
  file proteins

  "blastp -query $proteins -db nr"

}

It's worth noting that in the above examples, the name of the file in the file-system is not touched, you can access the file even without knowing its name because you can reference it in the process script by using the variable whose name is specified in the input file parameter declaration.

There may be cases where your task needs to use a file whose name is fixed, it does not have to change along with the actual provided file. In this case you can specify its name by specifying the name attribute in the input file parameter declaration, as shown in the following example:

input:
    file query_file name 'query.fa' from proteins

Or alternatively using a shorter syntax:

input:
    file 'query.fa' from proteins

Using this, the previous example can be re-written as shown below:

proteins = Channel.fromPath( '/some/path/*.fa' )

process blastThemAll {
  input:
  file 'query.fa' from proteins

  "blastp -query query.fa -db nr"

}

What happens in this example is that each file, that the process receives, is staged with the name query.fa in a different execution context (i.e. the folder where the job is executed) and an independent process execution is launched.

Multiple input files¶

A process can declare as input file a channel that emits a collection of values, instead of a simple value.

In this case, the script variable defined by the input file parameter will hold a list of files. You can use it as shown before, referring to all the files in the list, or by accessing a specific entry using the usual square brackets notation.

When a target file name is defined in the input parameter and a collection of files is received by the process, the file name will be appended by a numerical suffix representing its ordinal position in the list. For example:

fasta = Channel.fromPath( "/some/path/*.fa" ).buffer(size:3)

process blastThemAll {
    input:
    file 'seq' from fasta

    "echo seq*"

}

Will output:

seq1 seq2 seq3
seq1 seq2 seq3
...

The target input file name can contain the * and ? wildcards, that can be used to control the name of staged files. The following table shows how the wildcards are replaced depending on the cardinality of the received input collection.

The following fragment shows how a wildcard can be used in the input file declaration:

fasta = Channel.fromPath( "/some/path/*.fa" ).buffer(size:3)

process blastThemAll {
    input:
    file 'seq?.fa' from fasta

    "cat seq1.fa seq2.fa seq3.fa"

}

Dynamic input file names¶

When the input file name is specified by using the name file clause or the short string notation, you are allowed to use other input values as variables in the file name string. For example:

process simpleCount {
  input:
  val x from species
  file "${x}.fa" from genomes

  """
  cat ${x}.fa | grep '>'
  """
}

In the above example, the input file name is set by using the current value of the x input value.

This allows the input files to be staged in the script working directory with a name that is coherent with the current execution context.

Input of type 'stdin'¶

The stdin input qualifier allows you the forwarding of the value received from a channel to the standard input of the command executed by the process. For example:

str = Channel.from('hello', 'hola', 'bonjour', 'ciao').map { it+'\n' }

process printAll {
   input:
   stdin str

   """
   cat -
   """

}

It will output:

hola
bonjour
ciao
hello

Input of type 'env'¶

The env qualifier allows you to define an environment variable in the process execution context based on the value received from the channel. For example:

str = Channel.from('hello', 'hola', 'bonjour', 'ciao')

process printEnv {

    input:
    env HELLO from str

    '''
    echo $HELLO world!
    '''

}

hello world!
ciao world!
bonjour world!
hola world!

Input of type 'set'¶

The set qualifier allows you to group multiple parameters in a single parameter definition. It can be useful when a process receives, in input, tuples of values that need to be handled separately. Each element in the tuple is associated to a corresponding element with the set definition. For example:

tuple = Channel.from( [1, 'alpha'], [2, 'beta'], [3, 'delta'] )

process setExample {
    input:
    set val(x), file('latin.txt')  from tuple

    """
    echo Processing $x
    cat - latin.txt > copy
    """

}

In the above example the set parameter is used to define the value x and the file latin.txt, which will receive a value from the same channel.

In the set declaration items can be defined by using the following qualifiers: val, env, file and stdin.

A shorter notation can be used by applying the following substitution rules:

Thus the previous example could be rewritten as follows:

tuple = Channel.from( [1, 'alpha'], [2, 'beta'], [3, 'delta'] )

process setExample {
    input:
    set x, 'latin.txt' from tuple

    """
    echo Processing $x
    cat - latin.txt > copy
    """

}

File names can be defined in dynamic manner as explained in the Dynamic input file names section.

Input repeaters¶

The each qualifier allows you to repeat the execution of a process for each item in a collection, every time a new data is received. For example:

sequences = Channel.fromPath('*.fa')
methods = ['regular', 'expresso', 'psicoffee']

process alignSequences {
  input:
  file seq from sequences
  each mode from methods

  """
  t_coffee -in $seq -mode $mode > result
  """
}

In the above example every time a file of sequences is received as input by the process, it executes three tasks running a T-coffee alignment with a different value for the mode parameter. This is useful when you need to repeat the same task for a given set of parameters.

Since version 0.25+ input repeaters can be applied to files as well. For example:

sequences = Channel.fromPath('*.fa')
methods = ['regular', 'expresso']
libraries = [ file('PQ001.lib'), file('PQ002.lib'), file('PQ003.lib') ]

process alignSequences {
  input:
  file seq from sequences
  each mode from methods
  each file(lib) from libraries

  """
  t_coffee -in $seq -mode $mode -lib $lib > result
  """
}

In the latter example for any sequence input file emitted by the sequences channel are executed 6 alignments, 3 using the regular method against each library files, and other 3 by using the expresso method always against the same library files.

Understand how multiple input channels work¶

A key feature of processes is the ability to handle inputs from multiple channels.

When two or more channels are declared as process inputs, the process stops until there's a complete input configuration ie. it receives an input value from all the channels declared as input.

When this condition is verified, it consumes the input values coming from the respective channels, and spawns a task execution, then repeat the same logic until one or more channels have no more content.

This means channel values are consumed serially one after another and the first empty channel cause the process execution to stop even if there are other values in other channels.

For example:

process foo {
  echo true
  input:
  val x from Channel.from(1,2)
  val y from Channel.from('a','b','c')
  script:
   """
   echo $x and $y
   """
}

The process foo is executed two times because the first input channel only provides two values and therefore the c element is discarded. It prints:

1 and a
2 and b

This kind of channel is created by the Channel.value factory method or implicitly when a process input specifies a simple value in the from clause.

By definition, a Value channel is bound to a single value and it can be read unlimited times without consuming its content.

These properties make that when mixing a value channel with one or more (queue) channels, it does not affect the process termination which only depends by the other channels and its content is applied repeatedly.

To better understand this behavior compare the previous example with the following one:

process bar {
  echo true
  input:
  val x from Channel.value(1)
  val y from Channel.from('a','b','c')
  script:
   """
   echo $x and $y
   """
}

The above snippet executes the bar process three times because the first input is a value channel, therefore its content can be read as many times as needed. The process termination is determined by the content of the second channel. It prints:

1 and a
1 and b
1 and c

See also: Channel types.

Output values¶

The val qualifier allows to output a value defined in the script context. In a common usage scenario, this is a value which has been defined in the input declaration block, as shown in the following example:

methods = ['prot','dna', 'rna']

process foo {
  input:
  val x from methods

  output:
  val x into receiver

  """
  echo $x > file
  """

}

receiver.println { "Received: $it" }

Valid output values are value literals, input values identifiers, variables accessible in the process scope and value expressions. For example:

process foo {
  input:
  file fasta from 'dummy'

  output:
  val x into var_channel
  val 'BB11' into str_channel
  val "${fasta.baseName}.out" into exp_channel

  script:
  x = fasta.name
  """
  cat $x > file
  """
}

Output files¶

The file qualifier allows to output one or more files, produced by the process, over the specified channel. For example:

process randomNum {

   output:
   file 'result.txt' into numbers

   '''
   echo $RANDOM > result.txt
   '''

}

numbers.subscribe { println "Received: " + it.text }

In the above example the process, when executed, creates a file named result.txt containing a random number. Since a file parameter using the same name is declared between the outputs, when the task is completed that file is sent over the numbers channel. A downstream process declaring the same channel as input will be able to receive it.

Multiple output files¶

When an output file name contains a * or ? wildcard character it is interpreted as a glob path matcher. This allows to capture multiple files into a list object and output them as a sole emission. For example:

process splitLetters {

    output:
    file 'chunk_*' into letters

    '''
    printf 'Hola' | split -b 1 - chunk_
    '''
}

letters
    .flatMap()
    .subscribe { println "File: ${it.name} => ${it.text}" }

It prints:

File: chunk_aa => H
File: chunk_ab => o
File: chunk_ac => l
File: chunk_ad => a

Some caveats on glob pattern behavior:

• Input files are not included in the list of possible matches.
• Glob pattern matches against both files and directories path.
• When a two stars pattern ** is used to recourse across directories, only file paths are matched i.e. directories are not included in the result list.

By using the mode attribute the previous example can be re-written as show below:

process splitLetters {

    output:
    file 'chunk_*' into letters mode flatten

    '''
    printf 'Hola' | split -b 1 - chunk_
    '''
}

letters .subscribe { println "File: ${it.name} => ${it.text}" }

Read more about glob syntax at the following link What is a glob?

Dynamic output file names¶

When an output file name needs to be expressed dynamically, it is possible to define it using a dynamic evaluated string which references values defined in the input declaration block or in the script global context. For example:

process align {
  input:
  val x from species
  file seq from sequences

  output:
  file "${x}.aln" into genomes

  """
  t_coffee -in $seq > ${x}.aln
  """
}

In the above example, each time the process is executed an alignment file is produced whose name depends on the actual value of the x input.

Output 'stdout' special file¶

The stdout qualifier allows to capture the stdout output of the executed process and send it over the channel specified in the output parameter declaration. For example:

process echoSomething {
    output:
    stdout channel

    "echo Hello world!"

}

channel.subscribe { print "I say..  $it" }

Output 'set' of values¶

The set qualifier allows to send multiple values into a single channel. This feature is useful when you need to group together the results of multiple executions of the same process, as shown in the following example:

query = Channel.fromPath '*.fa'
species = Channel.from 'human', 'cow', 'horse'

process blast {

input:
    val species
    file query

output:
    set val(species), file('result') into blastOuts


"blast -db nr -query $query" > result

}

In the above example a BLAST task is executed for each pair of species and query that are received. When the task completes a new tuple containing the value for species and the file result is sent to the blastOuts channel.

A set declaration can contain any combination of the following qualifiers, previously described: val, file and stdout.

output:
    set species, 'result' into blastOuts

File names can be defined in a dynamic manner as explained in the Dynamic output file names section.

Optional Output¶

In most cases a process is expected to generate output that is added to the output channel. However, there are situations where it is valid for a process to not generate output. In these cases optional true may be added to the output declaration, which tells Nextflow not to fail the process if the declared output is not created.

output:
    file("output.txt") optional true into outChannel

In this example, the process is normally expected to generate an output.txt file, but in the cases where the file is legitimately missing, the process does not fail. outChannel is only populated by those processes that do generate output.txt.

afterScript¶

The afterScript directive allows you to execute a custom (Bash) snippet immediately after the main process has run. This may be useful to clean up your staging area.

beforeScript¶

The beforeScript directive allows you to execute a custom (Bash) snippet before the main process script is run. This may be useful to initialise the underlying cluster environment or for other custom initialisation.

For example:

process foo {

  beforeScript 'source /cluster/bin/setup'

  """
  echo bar
  """

}

cache¶

The cache directive allows you to store the process results to a local cache. When the cache is enabled and the pipeline is launched with the resume option, any following attempt to execute the process, along with the same inputs, will cause the process execution to be skipped, producing the stored data as the actual results.

The caching feature generates a unique key by indexing the process script and inputs. This key is used identify univocally the outputs produced by the process execution.

The cache is enabled by default, you can disable it for a specific process by setting the cache directive to false. For example:

process noCacheThis {
  cache false

  script:
  <your command string here>
}

The cache directive possible values are shown in the following table:

conda¶

The conda directive allows for the definition of the process dependencies using the Conda package manager.

Nextflow automatically sets up an environment for the given package names listed by in the conda directive. For example:

process foo {
  conda 'bwa=0.7.15'

  '''
  your_command --here
  '''
}

Multiple packages can be specified separating them with a blank space eg. bwa=0.7.15 fastqc=0.11.5. The name of the channel from where a specific package needs to be downloaded can be specified using the usual Conda notation i.e. prefixing the package with the channel name as shown here bioconda::bwa=0.7.15.

The conda directory also allows the specification of a Conda environment file path or the path of an existing environment directory. See the Conda environments page for further details.

container¶

The container directive allows you to execute the process script in a Docker container.

It requires the Docker daemon to be running in machine where the pipeline is executed, i.e. the local machine when using the local executor or the cluster nodes when the pipeline is deployed through a grid executor.

For example:

process runThisInDocker {

  container 'dockerbox:tag'

  """
  <your holy script here>
  """

}

Simply replace in the above script dockerbox:tag with the Docker image name you want to use.

containerOptions¶

The containerOptions directive allows you to specify any container execution option supported by the underlying container engine (ie. Docker, Singularity, etc). This can be useful to provide container settings only for a specific process e.g. mount a custom path:

process runThisWithDocker {

    container 'busybox:latest'
    containerOptions '--volume /data/db:/db'

    output: file 'output.txt'

    '''
    your_command --data /db > output.txt
    '''
}

cpus¶

The cpus directive allows you to define the number of (logical) CPU required by the process' task. For example:

process big_job {

  cpus 8
  executor 'sge'

  """
  blastp -query input_sequence -num_threads ${task.cpus}
  """
}

This directive is required for tasks that execute multi-process or multi-threaded commands/tools and it is meant to reserve enough CPUs when a pipeline task is executed through a cluster resource manager.

See also: penv, memory, time, queue, maxForks

clusterOptions¶

The clusterOptions directive allows the usage of any native configuration option accepted by your cluster submit command. You can use it to request non-standard resources or use settings that are specific to your cluster and not supported out of the box by Nextflow.

disk¶

The disk directive allows you to define how much local disk storage the process is allowed to use. For example:

process big_job {

    disk '2 GB'
    executor 'cirrus'

    """
    your task script here
    """
}

The following memory unit suffix can be used when specifying the disk value:

See also: cpus, memory time, queue and Dynamic computing resources.

echo¶

By default the stdout produced by the commands executed in all processes is ignored. Setting the echo directive to true you can forward the process stdout to the current top running process stdout file, showing it in the shell terminal.

For example:

process sayHello {
  echo true

  script:
  "echo Hello"
}

Hello

Without specifying echo true you won't see the Hello string printed out when executing the above example.

errorStrategy¶

The errorStrategy directive allows you to define how an error condition is managed by the process. By default when an error status is returned by the executed script, the process stops immediately. This in turn forces the entire pipeline to terminate.

Table of available error strategies:

When setting the errorStrategy directive to ignore the process doesn't stop on an error condition, it just reports a message notifying you of the error event.

For example:

process ignoreAnyError {
   errorStrategy 'ignore'

   script:
   <your command string here>
}

The retry error strategy, allows you to re-submit for execution a process returning an error condition. For example:

process retryIfFail {
   errorStrategy 'retry'

   script:
   <your command string here>
}

The number of times a failing process is re-executed is defined by the maxRetries and maxErrors directives.

See also: maxErrors, maxRetries and Dynamic computing resources.

executor¶

The executor defines the underlying system where processes are executed. By default a process uses the executor defined globally in the nextflow.config file.

The executor directive allows you to configure what executor has to be used by the process, overriding the default configuration. The following values can be used:

The following example shows how to set the process's executor:

process doSomething {

   executor 'sge'

   script:
   <your script here>

}

ext¶

The ext is a special directive used as namespace for user custom process directives. This can be useful for advanced configuration options. For example:

process mapping {
  container "biocontainers/star:${task.ext.version}"

  input:
  file genome from genome_file
  set sampleId, file(reads) from reads_ch

  """
  STAR --genomeDir $genome --readFilesIn $reads
  """
}

In the above example, the process uses a container whose version is controlled by the ext.version property. This can be defined in the nextflow.config file as shown below:

process.ext.version = '2.5.3'

maxErrors¶

The maxErrors directive allows you to specify the maximum number of times a process can fail when using the retry error strategy. By default this directive is disabled, you can set it as shown in the example below:

process retryIfFail {
  errorStrategy 'retry'
  maxErrors 5

  """
  echo 'do this as that .. '
  """
}

See also: errorStrategy and maxRetries.

maxForks¶

The maxForks directive allows you to define the maximum number of process instances that can be executed in parallel. By default this value is equals to the number of CPU cores available minus 1.

If you want to execute a process in a sequential manner, set this directive to one. For example:

process doNotParallelizeIt {

   maxForks 1

   '''
   <your script here>
   '''

}

maxRetries¶

The maxRetries directive allows you to define the maximum number of times a process instance can be re-submitted in case of failure. This value is applied only when using the retry error strategy. By default only one retry is allowed, you can increase this value as shown below:

process retryIfFail {
    errorStrategy 'retry'
    maxRetries 3

    """
    echo 'do this as that .. '
    """
}

See also: errorStrategy and maxErrors.

memory¶

The memory directive allows you to define how much memory the process is allowed to use. For example:

process big_job {

    memory '2 GB'
    executor 'sge'

    """
    your task script here
    """
}

The following memory unit suffix can be used when specifying the memory value:

See also: cpus, time, queue and Dynamic computing resources.

module¶

Environment Modules is a package manager that allows you to dynamically configure your execution environment and easily switch between multiple versions of the same software tool.

If it is available in your system you can use it with Nextflow in order to configure the processes execution environment in your pipeline.

In a process definition you can use the module directive to load a specific module version to be used in the process execution environment. For example:

process basicExample {

  module 'ncbi-blast/2.2.27'

  """
  blastp -query <etc..>
  """
}

You can repeat the module directive for each module you need to load. Alternatively multiple modules can be specified in a single module directive by separating all the module names by using a : (colon) character as shown below:

 process manyModules {

   module 'ncbi-blast/2.2.27:t_coffee/10.0:clustalw/2.1'

   """
   blastp -query <etc..>
   """
}

penv¶

The penv directive allows you to define the parallel environment to be used when submitting a parallel task to the SGE resource manager. For example:

process big_job {

  cpus 4
  penv 'smp'
  executor 'sge'

  """
  blastp -query input_sequence -num_threads ${task.cpus}
  """
}

This configuration depends on the parallel environment provided by your grid engine installation. Refer to your cluster documentation or contact your admin to lean more about this.

See also: cpus, memory, time

pod¶

The pod directive allows the definition of pods specific settings, such as environment variables, secrets and config maps when using the Kubernetes executor.

For example:

process your_task {
  pod env: 'FOO', value: 'bar'

  '''
  echo $FOO
  '''
}

The above snippet defines an environment variable named FOO which value is bar.

The pod directive allows the definition of the following options:

When defined in the Nextflow configuration file, a pod setting can be defined using the canonical associative array syntax. For example:

process {
  pod = [env: 'FOO', value: 'bar']
}

When more than one setting needs to be provides they must be enclosed in a list definition as shown below:

process {
  pod = [ [env: 'FOO', value: 'bar'], [secret: 'my-secret/key1', mountPath: '/etc/file.txt'] ]
}

publishDir¶

The publishDir directive allows you to publish the process output files to a specified folder. For example:

process foo {

    publishDir '/data/chunks'

    output:
    file 'chunk_*' into letters

    '''
    printf 'Hola' | split -b 1 - chunk_
    '''
}

The above example splits the string Hola into file chunks of a single byte. When complete the chunk_* output files are published into the /data/chunks folder.

By default files are published to the target folder creating a symbolic link for each process output that links the file produced into the process working directory. This behavior can be modified using the mode parameter.

Table of optional parameters that can be used with the publishDir directive:

Table of publish modes:

process foo {

    publishDir '/data/chunks', mode: 'copy', overwrite: false

    output:
    file 'chunk_*' into letters

    '''
    printf 'Hola' | split -b 1 - chunk_
    '''
}

queue¶

The queue directory allows you to set the queue where jobs are scheduled when using a grid based executor in your pipeline. For example:

process grid_job {

    queue 'long'
    executor 'sge'

    """
    your task script here
    """
}

Multiple queues can be specified by separating their names with a comma for example:

process grid_job {

    queue 'short,long,cn-el6'
    executor 'sge'

    """
    your task script here
    """
}

label¶

The label directive allows the annotation of processes with mnemonic identifier of your choice. For example:

process bigTask {

  label 'big_mem'

  '''
  <task script>
  '''
}

The same label can be applied to more than a process and multiple labels can be applied to the same process using the label directive more than one time.

Labels are useful to organise workflow processes in separate groups which can be referenced in the configuration file to select and configure subset of processes having similar computing requirements.

See the Process selectors documentation for details.

scratch¶

The scratch directive allows you to execute the process in a temporary folder that is local to the execution node.

This is useful when your pipeline is launched by using a grid executor, because it permits to decrease the NFS overhead by running the pipeline processes in a temporary directory in the local disk of the actual execution node. Only the files declared as output in the process definition will be copied in the pipeline working area.

In its basic form simply specify true at the directive value, as shown below:

process simpleTask {

  scratch true

  output:
  file 'data_out'

  '''
  <task script>
  '''
}

By doing this, it tries to execute the script in the directory defined by the variable $TMPDIR in the execution node. If this variable does not exist, it will create a new temporary directory by using the Linux command mktemp.

A custom environment variable, other than $TMPDIR, can be specified by simply using it as the scratch value, for example:

scratch '$MY_GRID_TMP'

Note, it must be wrapped by single quotation characters, otherwise the variable will be evaluated in the pipeline script context.

You can also provide a specific folder path as scratch value, for example:

scratch '/tmp/my/path'

By doing this, a new temporary directory will be created in the specified path each time a process is executed.

Finally, when the ram-disk string is provided as scratch value, the process will be execute in the node RAM virtual disk.

Summary of allowed values:

storeDir¶

The storeDir directive allows you to define a directory that is used as permanent cache for your process results.

In more detail, it affects the process execution in two main ways:

1. The process is executed only if the files declared in the output clause do not exist in the directory specified by the storeDir directive. When the files exist the process execution is skipped and these files are used as the actual process result.
2. Whenever a process is successfully completed the files listed in the output declaration block are moved into the directory specified by the storeDir directive.

The following example shows how to use the storeDir directive to create a directory containing a BLAST database for each species specified by an input parameter:

genomes = Channel.fromPath(params.genomes)

process formatBlastDatabases {

  storeDir '/db/genomes'

  input:
  file species from genomes

  output:
  file "${dbName}.*" into blastDb

  script:
  dbName = species.baseName
  """
  makeblastdb -dbtype nucl -in ${species} -out ${dbName}
  """

}

stageInMode¶

The stageInMode directive defines how input files are staged-in to the process work directory. The following values are allowed:

stageOutMode¶

The stageOutMode directive defines how output files are staged-out from the scratch directory to the process work directory. The following values are allowed:

See also: scratch.

tag¶

The tag directive allows you to associate each process executions with a custom label, so that it will be easier to identify them in the log file or in the trace execution report. For example:

process foo {
  tag "$code"

  input:
  val code from 'alpha', 'gamma', 'omega'

  """
  echo $code
  """
}

The above snippet will print a log similar to the following one, where process names contain the tag value:

[6e/28919b] Submitted process > foo (alpha)
[d2/1c6175] Submitted process > foo (gamma)
[1c/3ef220] Submitted process > foo (omega)

See also Trace execution report

time¶

The time directive allows you to define how long a process is allowed to run. For example:

process big_job {

    time '1h'

    """
    your task script here
    """
}

The following time unit suffix can be used when specifying the duration value:

See also: cpus, memory, queue and Dynamic computing resources.

validExitStatus¶

A process is terminated when the executed command returns an error exit status. By default any error status other than 0 is interpreted as an error condition.

The validExitStatus directive allows you to fine control which error status will represent a successful command execution. You can specify a single value or multiple values as shown in the following example:

process returnOk {
    validExitStatus 0,1,2

     script:
     """
     echo Hello
     exit 1
     """
}

In the above example, although the command script ends with a 1 exit status, the process will not return an error condition because the value 1 is declared as a valid status in the validExitStatus directive.

Dynamic directives¶

A directive can be assigned dynamically, during the process execution, so that its actual value can be evaluated depending on the value of one, or more, process' input values.

In order to be defined in a dynamic manner the directive's value needs to be expressed by using a closure statement, as in the following example:

process foo {

  executor 'sge'
  queue { entries > 100 ? 'long' : 'short' }

  input:
  set entries, file(x) from data

  script:
  """
  < your job here >
  """
}

In the above example the queue directive is evaluated dynamically, depending on the input value entries. When it is bigger than 100, jobs will be submitted to the queue long, otherwise the short one will be used.

All directives can be assigned to a dynamic value except the following:

• executor
• maxForks

For example:

process foo {

   queue { entries > 100 ? 'long' : 'short' }

   input:
   set entries, file(x) from data

   script:
   """
   echo Current queue: ${task.queue}
   """
 }

Dynamic computing resources¶

It's a very common scenario that different instances of the same process may have very different needs in terms of computing resources. In such situations requesting, for example, an amount of memory too low will cause some tasks to fail. Instead, using a higher limit that fits all the tasks in your execution could significantly decrease the execution priority of your jobs.

The Dynamic directives evaluation feature can be used to modify the amount of computing resources requested in case of a process failure and try to re-execute it using a higher limit. For example:

process foo {

    memory { 2.GB * task.attempt }
    time { 1.hour * task.attempt }

    errorStrategy { task.exitStatus in 137..140 ? 'retry' : 'terminate' }
    maxRetries 3

    script:
    <your job here>

}

In the above example the memory and execution time limits are defined dynamically. The first time the process is executed the task.attempt is set to 1, thus it will request a two GB of memory and one hour of maximum execution time.

If the task execution fail reporting an exit status in the range between 137 and 140, the task is re-submitted (otherwise terminates immediately). This time the value of task.attempt is 2, thus increasing the amount of the memory to four GB and the time to 2 hours, and so on.

The directive maxRetries set the maximum number of time the same task can be re-executed.

Dynamic Retry with backoff¶

There are cases in which the required execution resources may be temporary unavailable e.g. network congestion. In these cases immediately re-executing the task will likely result in the identical error. A retry with an exponential backoff delay can better recover these error conditions:

process foo {
  errorStrategy { sleep(Math.pow(2, task.attempt) * 200); return 'retry' }
  maxRetries 5
  script:
  '''
  your_command --here
  '''
}