Groovy Essentials for Nextflow Developers¶
Nextflow is built on Groovy, a powerful dynamic language that runs on the Java Virtual Machine. You can write a lot of Nextflow without ever feeling like you've learned Groovy - many workflows use only basic syntax for variables, maps, and lists. Most Nextflow tutorials focus on workflow orchestration (channels, processes, and data flow), and you can go surprisingly far with just that.
However, when you need to manipulate data, parse complex filenames, implement conditional logic, or build robust production workflows, you're writing Groovy code - and knowing a few key Groovy concepts can dramatically improve your ability to solve real-world problems efficiently. Understanding where Nextflow ends and Groovy begins helps you write clearer, more maintainable workflows.
This side quest takes you on a hands-on journey from basic concepts to production-ready patterns. We'll transform a simple CSV-reading workflow into a sophisticated bioinformatics pipeline, evolving it step-by-step through realistic challenges:
- Understanding boundaries: Distinguish between Nextflow operators and Groovy methods, and master when to use each
- Data manipulation: Extract, transform, and subset maps and collections using Groovy's powerful operators
- String processing: Parse complex file naming schemes with regex patterns and master variable interpolation
- Reusable functions: Extract complex logic into named functions for cleaner, more maintainable workflows
- Dynamic logic: Build processes that adapt to different input types and use closures for dynamic resource allocation
- Conditional routing: Intelligently route samples through different processes based on their metadata characteristics
- Safe operations: Handle missing data gracefully with null-safe operators and validate inputs with clear error messages
- Configuration-based handlers: Use workflow event handlers for logging, notifications, and lifecycle management
0. Warmup¶
0.1. Prerequisites¶
Before taking on this side quest you should:
- Complete the Hello Nextflow tutorial or have equivalent experience
- Understand basic Nextflow concepts (processes, channels, workflows)
- Have basic familiarity with common programming constructs used in Groovy syntax (variables, maps, lists)
This tutorial will explain Groovy concepts as we encounter them, so you don't need extensive prior Groovy knowledge. We'll start with fundamental concepts and build up to advanced patterns.
0.2. Starting Point¶
Navigate to the project directory:
The data directory contains sample files and a main workflow file we'll evolve throughout.
> tree
.
├── collect.nf
├── data
│ ├── samples.csv
│ └── sequences
│ ├── SAMPLE_001_S1_L001_R1_001.fastq
│ ├── SAMPLE_002_S2_L001_R1_001.fastq
│ └── SAMPLE_003_S3_L001_R1_001.fastq
├── main.nf
├── modules
│ ├── fastp.nf
│ ├── generate_report.nf
│ └── trimgalore.nf
└── nextflow.config
4 directories, 10 files
Our sample CSV contains information about biological samples that need different processing based on their characteristics:
sample_id,organism,tissue_type,sequencing_depth,file_path,quality_score
SAMPLE_001,human,liver,30000000,data/sequences/SAMPLE_001_S1_L001_R1_001.fastq,38.5
SAMPLE_002,mouse,brain,25000000,data/sequences/SAMPLE_002_S2_L001_R1_001.fastq,35.2
SAMPLE_003,human,kidney,45000000,data/sequences/SAMPLE_003_S3_L001_R1_001.fastq,42.1
We'll use this realistic dataset to explore practical Groovy techniques that you'll encounter in real bioinformatics workflows.
1. Nextflow vs Groovy: Understanding the Boundaries¶
1.1. Identifying What's What¶
Nextflow developers often confuse Nextflow constructs with Groovy language features. Let's build a workflow demonstrating how they work together.
Step 1: Basic Nextflow Workflow¶
Start with a simple workflow that just reads the CSV file (we've already done this for you in main.nf):
| main.nf | |
|---|---|
The workflow block defines our pipeline structure, while Channel.fromPath() creates a channel from a file path. The .splitCsv() operator processes the CSV file and converts each row into a map data structure.
Run this workflow to see the raw CSV data:
You should see output like:
Launching `main.nf` [marvelous_tuckerman] DSL2 - revision: 6113e05c17
[sample_id:SAMPLE_001, organism:human, tissue_type:liver, sequencing_depth:30000000, file_path:data/sequences/SAMPLE_001_S1_L001_R1_001.fastq, quality_score:38.5]
[sample_id:SAMPLE_002, organism:mouse, tissue_type:brain, sequencing_depth:25000000, file_path:data/sequences/SAMPLE_002_S2_L001_R1_001.fastq, quality_score:35.2]
[sample_id:SAMPLE_003, organism:human, tissue_type:kidney, sequencing_depth:45000000, file_path:data/sequences/SAMPLE_003_S3_L001_R1_001.fastq, quality_score:42.1]
Step 2: Adding the Map Operator¶
Now we're going to use some Groovy code to transform the data, using the .map() operator you will probably already be familiar with. This operator takes a 'closure' where we can write Groovy code to transform each item.
Note
A closure is a block of code that can be passed around and executed later. Think of it as a function that you define inline. In Groovy, closures are written with curly braces { } and can take parameters. They're fundamental to how Nextflow operators work and if you've been writing Nextflow for a while, you may already have been using them without realizing it!
Here's what that map operation looks like:
This is our first Groovy closure—an anonymous function you can pass as an argument. Closures are a core Groovy concept (similar to lambdas in Python or arrow functions in JavaScript) and are essential for working with Nextflow operators.
The closure { row -> return row } takes a parameter row (could be any name: item, sample, etc.). You can also use the implicit variable it instead: .map { return it }, though naming parameters improves clarity.
When Nextflow processes each channel item, it passes that item to your closure. Here, row holds one CSV row at a time.
Apply this change and run the workflow:
You'll see the same output as before, because we're simply returning the input unchanged. This confirms that the map operator is working correctly. Now let's start transforming the data.
Step 3: Creating a Map Data Structure¶
Now we're going to write pure Groovy code inside our closure. Everything from this point forward in this section is Groovy syntax and methods, not Nextflow operators.
This is pure Groovy code. The sample_meta map is a key-value data structure (like dictionaries in Python, objects in JavaScript, or hashes in Ruby) storing related information: sample ID, organism, tissue type, sequencing depth, and quality score.
We use Groovy's string manipulation methods like .toLowerCase() and .replaceAll() to clean up our data, and type conversion methods like .toInteger() and .toDouble() to convert string data from the CSV into the appropriate numeric types.
Apply this change and run the workflow:
You should see the refined map output like:
[id:sample_001, organism:human, tissue:liver, depth:30000000, quality:38.5]
[id:sample_002, organism:mouse, tissue:brain, depth:25000000, quality:35.2]
[id:sample_003, organism:human, tissue:kidney, depth:45000000, quality:42.1]
Step 4: Adding Conditional Logic¶
Now let's add more Groovy logic - this time using a ternary operator to make decisions based on data values.
Make the following change:
The ternary operator is a shorthand for an if/else statement that follows the pattern condition ? value_if_true : value_if_false. This line means: "If the quality is greater than 40, use 'high', otherwise use 'normal'". Its cousin, the Elvis operator (?:), provides default values when something is null or empty - we'll explore that pattern later in this tutorial.
The map addition operator + creates a new map rather than modifying the existing one. This line creates a new map that contains all the key-value pairs from sample_meta plus the new priority key.
Note
Never modify maps passed into closures - always create new ones using + (for example). In Nextflow, the same data often flows through multiple operations simultaneously. Modifying a map in-place can cause unpredictable side effects when other operations reference that same object. Creating new maps ensures each operation has its own clean copy.
Run the modified workflow:
You should see output like:
[id:sample_001, organism:human, tissue:liver, depth:30000000, quality:38.5, priority:normal]
[id:sample_002, organism:mouse, tissue:brain, depth:25000000, quality:35.2, priority:normal]
[id:sample_003, organism:human, tissue:kidney, depth:45000000, quality:42.1, priority:high]
We've successfully added conditional logic to enrich our metadata with a priority level based on quality scores.
Step 4.5: Subsetting Maps with .subMap()¶
While the + operator adds keys to a map, sometimes you need to do the opposite - extract only specific keys. Groovy's .subMap() method is perfect for this.
Let's add a line to create a simplified version of our metadata that only contains identification fields:
Run the modified workflow:
You should see output showing both the full metadata displayed by the view() operation and the extracted subset we printed with println:
N E X T F L O W ~ version 25.04.6
Launching `main.nf` [peaceful_cori] DSL2 - revision: 4cc4a8340f
ID fields only: [id:sample_001, organism:human, tissue:liver]
ID fields only: [id:sample_002, organism:mouse, tissue:brain]
ID fields only: [id:sample_003, organism:human, tissue:kidney]
[id:sample_001, organism:human, tissue:liver, depth:30000000, quality:38.5, priority:normal]
[id:sample_002, organism:mouse, tissue:brain, depth:25000000, quality:35.2, priority:normal]
[id:sample_003, organism:human, tissue:kidney, depth:45000000, quality:42.1, priority:high]
The .subMap() method takes a list of keys and returns a new map containing only those keys. If a key doesn't exist in the original map, it's simply not included in the result.
This is particularly useful when you need to create different metadata versions for different processes - some might need full metadata while others need only minimal identification fields.
Now remove those println statements to restore your workflow to its previous state, as we don't need them going forward.
Map Operations Summary
- Add keys:
map1 + [new_key: value]- Creates new map with additional keys - Extract keys:
map1.subMap(['key1', 'key2'])- Creates new map with only specified keys - Both operations create new maps - Original maps remain unchanged
Step 5: Combining Maps and Returning Results¶
So far, we've only been returning what Nextflow community calls the 'meta map', and we've been ignoring the files those metadata relate to. But if you're writing Nextflow workflows, you probably want to do something with those files.
Let's output a channel structure comprising a tuple of 2 elements: the enriched metadata map and the corresponding file path. This is a common pattern in Nextflow for passing data to processes.
Apply this change and run the workflow:
You should see output like:
[[id:sample_001, organism:human, tissue:liver, depth:30000000, quality:38.5, priority:normal], /workspaces/training/side-quests/groovy_essentials/data/sequences/SAMPLE_001_S1_L001_R1_001.fastq]
[[id:sample_002, organism:mouse, tissue:brain, depth:25000000, quality:35.2, priority:normal], /workspaces/training/side-quests/groovy_essentials/data/sequences/SAMPLE_002_S2_L001_R1_001.fastq]
[[id:sample_003, organism:human, tissue:kidney, depth:45000000, quality:42.1, priority:high], /workspaces/training/side-quests/groovy_essentials/data/sequences/SAMPLE_003_S3_L001_R1_001.fastq]
This [meta, file] tuple structure is a common pattern in Nextflow for passing both metadata and associated files to processes.
Note
Maps and Metadata: Maps are fundamental to working with metadata in Nextflow. For a more detailed explanation of working with metadata maps, see the Working with metadata side quest.
Our workflow demonstrates the core pattern: Nextflow constructs (workflow, Channel.fromPath(), .splitCsv(), .map(), .view()) orchestrate data flow, while basic Groovy constructs (maps [key: value], string methods, type conversions, ternary operators) handle the data processing logic inside the .map() closure.
1.2. Distinguishing Nextflow operators from Groovy functions¶
So far, so good, we can distinguish between Nextflow constructs and basic Groovy constructs. But what about when the syntax overlaps?
A perfect example of this confusion is the collect operation, which exists in both contexts but does completely different things. Groovy's collect transforms each element, while Nextflow's collect gathers all channel elements into a single-item channel.
Let's demonstrate this with some sample data, starting by refreshing ourselves on what the Nextflow collect() operator does. Check out collect.nf:
Steps:
- Define a Groovy list
- Create a channel with
fromList()that emits each sample ID separately - Print each item with
view()as it flows through - Gather all items into a single list with Nextflow's
collect()operator - Print the collected result (single item containing all sample IDs) with a second
view()
We've changed the structure of the channel, but we haven't changed the data itself.
Run the workflow to confirm this:
N E X T F L O W ~ version 25.04.6
Launching `collect.nf` [loving_mendel] DSL2 - revision: e8d054a46e
Individual channel item: sample_001
Individual channel item: sample_002
Individual channel item: sample_003
Nextflow collect() result: [sample_001, sample_002, sample_003] (3 items grouped into 1)
view() returns an output for every channel emission, so we know that this single output contains all 3 original items grouped into one list.
Now let's see Groovy's collect method in action. Modify collect.nf to apply Groovy's collect method to the original list of sample IDs:
In this new snippet we:
- Define a new variable
formatted_idsthat uses Groovy'scollectmethod to transform each sample ID in the original list - Print the result using
println
Run the modified workflow:
N E X T F L O W ~ version 25.04.6
Launching `collect.nf` [cheeky_stonebraker] DSL2 - revision: 2d5039fb47
Groovy collect result: [SPECIMEN_001, SPECIMEN_002, SPECIMEN_003] (3 items transformed into 3)
Individual channel item: sample_001
Individual channel item: sample_002
Individual channel item: sample_003
Nextflow collect() result: [sample_001, sample_002, sample_003] (3 items grouped into 1)
This time, we have NOT changed the structure of the data, we still have 3 items in the list, but we HAVE transformed each item using Groovy's collect method to produce a new list with modified values. This is sort of like using the map operator in Nextflow, but it's pure Groovy code operating on a standard Groovy list.
collect is an extreme case we're using here to make a point. The key lesson is that when you're writing workflows always distinguish between Groovy constructs (data structures) and Nextflow constructs (channels/workflows). Operations can share names but behave completely differently.
1.3. The Spread Operator (*.) - Shorthand for Property Extraction¶
Related to Groovy's collect is the spread operator (*.), which provides a concise way to extract properties from collections. It's essentially syntactic sugar for a common collect pattern.
Let's add a demonstration to our collect.nf file:
Run the updated workflow:
You should see output like:
N E X T F L O W ~ version 25.04.6
Launching `collect.nf` [cranky_galileo] DSL2 - revision: 5f3c8b2a91
Groovy collect result: [SPECIMEN_001, SPECIMEN_002, SPECIMEN_003] (3 items transformed into 3)
Spread operator result: [s1, s2, s3]
Individual channel item: sample_001
Individual channel item: sample_002
Individual channel item: sample_003
Nextflow collect() result: [sample_001, sample_002, sample_003] (3 items grouped into 1)
The spread operator *. is a shorthand for a common collect pattern:
// These are equivalent:
def ids = samples*.id
def ids = samples.collect { it.id }
// Also works with method calls:
def names = files*.getName()
def names = files.collect { it.getName() }
The spread operator is particularly useful when you need to extract a single property from a list of objects - it's more readable than writing out the full collect closure.
When to Use Groovy's Spread vs Collect
- Use spread (
*.) for simple property access:samples*.id,files*.name - Use collect for transformations or complex logic:
samples.collect { it.id.toUpperCase() },samples.collect { [it.id, it.quality > 40] }
Takeaway¶
In this section, you've learned:
- It takes both Nextflow and Groovy: Nextflow provides the workflow structure and data flow, while Groovy provides the data manipulation and logic
- Distinguishing Nextflow from Groovy: How to identify which language construct you're using given the context
- Context matters: The same operation name can have completely different behaviors
Understanding these boundaries is essential for debugging, documentation, and writing maintainable workflows.
Next we'll dive deeper into Groovy's powerful string processing capabilities, which are essential for handling real-world data.
2. String Processing and Dynamic Script Generation¶
Mastering Groovy's string processing separates brittle workflows from robust pipelines. This section covers parsing complex file names, dynamic script generation, and variable interpolation.
2.1. Pattern Matching and Regular Expressions¶
Bioinformatics files often have complex naming conventions encoding metadata. Let's extract this automatically with Groovy's pattern matching.
We're going to return to our main.nf workflow and add some pattern matching logic to extract additional sample information from file names. The FASTQ files in our dataset follow Illumina-style naming conventions with names like SAMPLE_001_S1_L001_R1_001.fastq.gz. These might look cryptic, but they actually encode useful metadata like sample ID, lane number, and read direction. We're going to use Groovy's regex capabilities to parse these names.
Make the following change to your existing main.nf workflow:
This demonstrates key Groovy string processing concepts:
- Regular expression literals using
~/pattern/syntax - this creates a regex pattern without needing to escape backslashes - Pattern matching with the
=~operator - this attempts to match a string against a regex pattern - Matcher objects that capture groups with
[0][1],[0][2], etc. -[0]refers to the entire match,[1],[2], etc. refer to captured groups in parentheses
Let's break down the regex pattern ^(.+)_S(\d+)_L(\d{3})_(R[12])_(\d{3})\.fastq(?:\.gz)?$:
| Pattern | Matches | Captures |
|---|---|---|
^(.+) |
Sample name from start | Group 1: sample name |
_S(\d+) |
Sample number _S1, _S2, etc. |
Group 2: sample number |
_L(\d{3}) |
Lane number _L001 |
Group 3: lane (3 digits) |
_(R[12]) |
Read direction _R1 or _R2 |
Group 4: read direction |
_(\d{3}) |
Chunk number _001 |
Group 5: chunk (3 digits) |
\.fastq(?:\.gz)?$ |
File extension .fastq or .fastq.gz |
Not captured (?: is non-capturing) |
This parses Illumina-style naming conventions to extract metadata automatically.
Run the modified workflow:
You should see output with metadata enriched from the file names, like
N E X T F L O W ~ version 25.04.6
Launching `main.nf` [clever_pauling] DSL2 - revision: 605d2058b4
[[id:sample_001, organism:human, tissue:liver, depth:30000000, quality:38.5, sample_num:1, lane:001, read:R1, chunk:001, priority:normal], /workspaces/training/side-quests/groovy_essentials/data/sequences/SAMPLE_001_S1_L001_R1_001.fastq]
[[id:sample_002, organism:mouse, tissue:brain, depth:25000000, quality:35.2, sample_num:2, lane:001, read:R1, chunk:001, priority:normal], /workspaces/training/side-quests/groovy_essentials/data/sequences/SAMPLE_002_S2_L001_R1_001.fastq]
[[id:sample_003, organism:human, tissue:kidney, depth:45000000, quality:42.1, sample_num:3, lane:001, read:R1, chunk:001, priority:high], /workspaces/training/side-quests/groovy_essentials/data/sequences/SAMPLE_003_S3_L001_R1_001.fastq]
2.2. Dynamic Script Generation in Processes¶
Process script blocks are essentially multi-line strings that get passed to the shell. You can use Groovy conditional logic (if/else, ternary operators) to dynamically generate different script strings based on input characteristics. This is essential for handling diverse input types—like single-end vs paired-end sequencing reads—without duplicating process definitions.
Let's add a process to our workflow that demonstrates this pattern. Open modules/fastp.nf and take a look:
The process takes FASTQ files as input and runs the fastp tool to trim adapters and filter low-quality reads. Unfortunately, the person who wrote this process didn't allow for the single-end reads we have in our example dataset. Let's add it to our workflow and see what happens:
First, include the module at the very first line of your main.nf workflow:
| main.nf | |
|---|---|
Then modify the workflow block to connect the ch_samples channel to the FASTP process:
Run this modified workflow:
You'll see a long error trace with some content like:
ERROR ~ Error executing process > 'FASTP (3)'
Caused by:
Process `FASTP (3)` terminated with an error exit status (255)
Command executed:
fastp \
--in1 SAMPLE_003_S3_L001_R1_001.fastq \
--in2 null \
--out1 sample_003_trimmed_R1.fastq.gz \
--out2 sample_003_trimmed_R2.fastq.gz \
--json sample_003.fastp.json \
--html sample_003.fastp.html \
--thread 2
Command exit status:
255
Command output:
(empty)
You can see that the process is trying to run fastp with a null value for the second input file, which is causing it to fail. This is because our dataset contains single-end reads, but the process is hardcoded to expect paired-end reads (two input files at a time).
Fix this by adding Groovy logic to the FASTP process script: block. An if/else statement checks read file count and adjusts the command accordingly.
Now the workflow can handle both single-end and paired-end reads gracefully. The Groovy logic checks the number of input files and constructs the appropriate command for fastp. Let's see if it works:
N E X T F L O W ~ version 25.04.6
Launching `main.nf` [adoring_rosalind] DSL2 - revision: 04b1cd93e9
executor > local (3)
[31/a8ad4d] process > FASTP (3) [100%] 3 of 3 ✔
Looks good! If we check the actual commands that were run (customise for your task hash):
We can see that Nextflow correctly picked the right command for single-end reads:
#!/bin/bash -ue
fastp \
--in1 SAMPLE_003_S3_L001_R1_001.fastq \
--out1 sample_003_trimmed.fastq.gz \
--json sample_003.fastp.json \
--html sample_003.fastp.html \
--thread 2
Another common usage of dynamic script logic can be seen in the Nextflow for Science Genomics module. In that module, the GATK process being called can take multiple input files, but each must be prefixed with -V to form a correct command line. The process uses Groovy logic to transform a collection of input files (all_gvcfs) into the correct command arguments:
| command line manipulation for GATK | |
|---|---|
These patterns of using Groovy logic in process script blocks are extremely powerful and can be applied in many scenarios - from handling variable input types to building complex command-line arguments from file collections, making your processes truly adaptable to the diverse requirements of real-world data.
2.3. Variable Interpolation: Groovy, Bash, and Shell Variables¶
Process scripts mix Nextflow variables, shell variables, and command substitutions, each with different interpolation syntax. Using the wrong syntax causes errors. Let's explore these with a process that creates a processing report.
Take a look a the module file modules/generate_report.nf:
| modules/generate_report.nf | |
|---|---|
This process writes a simple report with the sample ID and filename. Now let's run it to see what happens when we need to mix different types of variables.
Include the process in your main.nf and add it to the workflow:
Now run the workflow and check the generated reports in results/reports/. They should contain basic information about each sample.
But what if we want to add information about when and where the processing occurred? Let's modify the process to use shell variables and a bit of command substitution to include the current user, hostname, and date in the report:
| modules/generate_report.nf | |
|---|---|
If you run this, you'll notice an error or unexpected behavior - Nextflow tries to interpret $(hostname) as a Groovy variable that doesn't exist:
unknown recognition error type: groovyjarjarantlr4.v4.runtime.LexerNoViableAltException
ERROR ~ Module compilation error
- file : /workspaces/training/side-quests/groovy_essentials/modules/generate_report.nf
- cause: token recognition error at: '(' @ line 16, column 22.
echo "Hostname: $(hostname)" >> ${meta.id}_report.txt
^
1 error
We need to escape it so Bash can handle it instead.
Fix this by escaping the shell variables and command substitutions with a backslash (\):
| modules/generate_report.nf | |
|---|---|
| modules/generate_report.nf | |
|---|---|
Now it works! The backslash (\) tells Nextflow "don't interpret this, pass it through to Bash."
Takeaway¶
In this section, you've learned Groovy string processing techniques:
- Regular expressions for file parsing: Using Groovy's
=~operator and regex patterns (~/pattern/) to extract metadata from complex file naming conventions - Dynamic script generation: Using Groovy conditional logic (if/else, ternary operators) to generate different script strings based on input characteristics
- Variable interpolation: Understanding when Nextflow interprets strings vs when the shell does
${var}- Groovy/Nextflow variables (interpolated by Nextflow at workflow compile time)\${var}- Shell environment variables (escaped, passed to bash at runtime)\$(cmd)- Shell command substitution (escaped, executed by bash at runtime)
These string processing and generation patterns are essential for handling the diverse file formats and naming conventions you'll encounter in real-world bioinformatics workflows.
3. Creating Reusable Functions¶
Complex workflow logic inline in channel operators or process definitions reduces readability and maintainability. Groovy functions let you extract this logic into named, reusable components—this is core Groovy programming, not Nextflow-specific syntax.
Our map operation has grown long and complex. Let's extract it into a reusable Groovy function using the def keyword.
To illustrate what that looks like with our existing workflow, make the modification below, using def to define a reusable function called separateMetadata:
By extracting this logic into a function, we've reduced the actual workflow logic down to something much cleaner:
ch_samples = Channel.fromPath("./data/samples.csv")
.splitCsv(header: true)
.map{ row -> separateMetadata(row) }
ch_fastp = FASTP(ch_samples)
GENERATE_REPORT(ch_samples)
This makes the workflow logic much easier to read and understand at a glance. The function separateMetadata encapsulates all the complex logic for parsing and enriching metadata, making it reusable and testable.
Run the workflow to make sure it still works:
N E X T F L O W ~ version 25.04.6
Launching `main.nf` [admiring_panini] DSL2 - revision: 8cc832e32f
executor > local (6)
[8c/2e3f91] process > FASTP (3) [100%] 3 of 3 ✔
[7a/1b4c92] process > GENERATE_REPORT (3) [100%] 3 of 3 ✔
The output should show both processes completing successfully. The workflow is now much cleaner and easier to maintain, with all the complex metadata processing logic encapsulated in the separateMetadata function.
Takeaway¶
In this section, you've learned core Groovy programming concepts:
- Defining functions with
def: Groovy's keyword for creating named functions (likedefin Python orfunctionin JavaScript) - Function scope: Functions defined at the script level are accessible throughout your Nextflow workflow
- Return values: Functions automatically return the last expression, or use explicit
return - Cleaner code: Extracting complex logic into functions is a fundamental software engineering practice in any language, including Groovy
Next, we'll explore how to use Groovy closures in process directives for dynamic resource allocation.
4. Dynamic Resource Directives with Closures¶
So far we've used Groovy in the script block of processes. But Groovy closures (introduced in Section 1.1) are also incredibly useful in process directives, especially for dynamic resource allocation. Let's add resource directives to our FASTP process that adapt based on the sample characteristics.
Currently, our FASTP process uses default resources. Let's make it smarter by allocating more CPUs for high-depth samples. Edit modules/fastp.nf to include a dynamic cpus directive and a static memory directive:
The closure { meta.depth > 40000000 ? 4 : 2 } uses the Groovy ternary operator (covered in Section 1.1) and is evaluated for each task, allowing per-sample resource allocation. High-depth samples (>40M reads) get 4 CPUs, while others get 2 CPUs.
Accessing Input Variables in Directives
The closure can access any input variables (like meta here) because Nextflow evaluates these closures in the context of each task execution.
Run the workflow again:
We're using the -no-ansi-log option to make it easier to see the task hashes.
N E X T F L O W ~ version 25.04.6
Launching `main.nf` [fervent_albattani] DSL2 - revision: fa8f249759
[bd/ff3d41] Submitted process > FASTP (2)
[a4/a3aab2] Submitted process > FASTP (1)
[48/6db0c9] Submitted process > FASTP (3)
[ec/83439d] Submitted process > GENERATE_REPORT (3)
[bd/15d7cc] Submitted process > GENERATE_REPORT (2)
[42/699357] Submitted process > GENERATE_REPORT (1)
You can check the exact docker command that was run to see the CPU allocation for any given task:
You should see something like:
docker run -i --cpu-shares 4096 --memory 2048m -e "NXF_TASK_WORKDIR" -v /workspaces/training/side-quests/groovy_essentials:/workspaces/training/side-quests/groovy_essentials -w "$NXF_TASK_WORKDIR" --name $NXF_BOXID community.wave.seqera.io/library/fastp:0.24.0--62c97b06e8447690 /bin/bash -ue /workspaces/training/side-quests/groovy_essentials/work/48/6db0c9e9d8aa65e4bb4936cd3bd59e/.command.sh
In this example we've chosen an example that requested 4 CPUs (--cpu-shares 4096), because it was a high-depth sample, but you should see different CPU allocations depending on the sample depth. Try this for the other tasks as well.
Another powerful pattern is using task.attempt for retry strategies. To show why this is useful, we're going to start by reducing the memory allocation to FASTP to less than it needs. Change the memory directive in modules/fastp.nf to 1.GB:
... and run the workflow again:
You'll see an error indicating that the process was killed for exceeding memory limits:
Command exit status:
137
Command output:
(empty)
Command error:
Detecting adapter sequence for read1...
No adapter detected for read1
.command.sh: line 7: 101 Killed fastp --in1 SAMPLE_002_S2_L001_R1_001.fastq --out1 sample_002_trimmed.fastq.gz --json sample_002.fastp.json --html sample_002.fastp.html --thread 2
This is a very common scenario in real-world workflows - sometimes you just don't know how much memory a task will need until you run it. To make our workflow more robust, we can implement a retry strategy that increases memory allocation on each attempt, once again using a Groovy closure. Modify the memory directive to multiply the base memory by task.attempt, and add errorStrategy 'retry' and maxRetries 2 directives:
Now if the process fails due to insufficient memory, Nextflow will retry with more memory:
- First attempt: 1 GB (task.attempt = 1)
- Second attempt: 2 GB (task.attempt = 2)
... and so on, up to the maxRetries limit.
Takeaway¶
Dynamic directives with Groovy closures let you:
- Allocate resources based on input characteristics
- Implement automatic retry strategies with increasing resources
- Combine multiple factors (metadata, attempt number, priorities)
- Use Groovy logic for complex resource calculations
This makes your workflows both more efficient (not over-allocating) and more robust (automatic retry with more resources).
5. Conditional Logic and Process Control¶
Previously, we used .map() with Groovy to transform channel data. Now we'll use Groovy to control which processes execute based on data—essential for flexible workflows adapting to different sample types.
Nextflow's flow control operators take closures evaluated at runtime, enabling Groovy logic to drive workflow decisions based on channel content.
5.1. Routing with .branch()¶
For example, let's pretend that our sequencing samples need to be trimmed with FASTP only if they're human samples with a coverage above a certain threshold. Mouse samples or low-coverage samples should be run with Trimgalore instead (this is a contrived example, but it illustrates the point).
We've provided a simple Trimgalore process in modules/trimgalore.nf, take a look if you like, but the details aren't important for this exercise. The key point is that we want to route samples based on their metadata.
Include the new from in modules/trimgalore.nf:
... and then modify your main.nf workflow to branch samples based on their metadata and route them through the appropriate trimming process, like this:
Run this modified workflow:
N E X T F L O W ~ version 25.04.6
Launching `main.nf` [adoring_galileo] DSL2 - revision: c9e83aaef1
executor > local (6)
[1d/0747ac] process > FASTP (2) [100%] 2 of 2 ✔
[cc/c44caf] process > TRIMGALORE (1) [100%] 1 of 1 ✔
[34/bd5a9f] process > GENERATE_REPORT (1) [100%] 3 of 3 ✔
Here, we've used small but mighty Groovy expressions inside the .branch{} operator to route samples based on their metadata. Human samples with high coverage go through FASTP, while all other samples go through TRIMGALORE.
5.2. Using .filter() with Groovy Truth¶
Another powerful pattern for controlling workflow execution is the .filter() operator, which uses a closure to determine which items should continue down the pipeline. Inside the filter closure, you'll write Groovy boolean expressions that decide which items pass through.
Groovy has a concept called "Groovy Truth" that determines what values evaluate to true or false in boolean contexts:
- Truthy: Non-null values, non-empty strings, non-zero numbers, non-empty collections
- Falsy:
null, empty strings"", zero0, empty collections[]or[:],false
This means meta.id alone (without explicit != null) checks if the ID exists and isn't empty. Let's use this to filter out samples that don't meet our quality requirements.
Add the following before the branch operation:
Run the workflow again:
Because we've chosen a filter that excludes some samples, you should see fewer tasks executed:
N E X T F L O W ~ version 25.04.6
Launching `main.nf` [deadly_woese] DSL2 - revision: 9a6044a969
executor > local (5)
[01/7b1483] process > FASTP (2) [100%] 2 of 2 ✔
[- ] process > TRIMGALORE -
[07/ef53af] process > GENERATE_REPORT (3) [100%] 3 of 3 ✔
The filter expression meta.id && meta.organism && meta.depth >= 25000000 combines Groovy Truth with explicit comparisons:
meta.id && meta.organismchecks that both fields exist and are non-empty (using Groovy Truth)meta.depth >= 25000000ensures sufficient sequencing depth with an explicit comparison
Groovy Truth in Practice
The expression meta.id && meta.organism is more concise than writing:
This makes filtering logic much cleaner and easier to read.
Takeaway¶
In this section, you've learned to use Groovy logic to control workflow execution using the closure interfaces of Nextflow operators like .branch{} and .filter{}, leveraging Groovy Truth to write concise conditional expressions.
Our pipeline now intelligently routes samples through appropriate processes, but production workflows need to handle invalid data gracefully. Let's make our workflow robust against missing or null values.
6. Safe Navigation and Elvis Operators¶
Our separateMetadata function currently assumes all CSV fields are present and valid. But what happens with incomplete data? Let's find out.
6.1. The Problem: Accessing Properties That Don't Exist¶
Let's say we want to add support for optional sequencing run information. In some labs, samples might have an additional field for the sequencing run ID or batch number, but our current CSV doesn't have this column. Let's try to access it anyway.
Modify the separateMetadata function to include a run_id field:
| main.nf | |
|---|---|
Now run the workflow:
It crashes with a NullPointerException:
N E X T F L O W ~ version 25.04.6
Launching `main.nf` [trusting_torvalds] DSL2 - revision: b56fbfbce2
ERROR ~ Cannot invoke method toUpperCase() on null object
-- Check script 'main.nf' at line: 13 or see '.nextflow.log' file for more details
The problem is that row.run_id returns null because the run_id column doesn't exist in our CSV. When we try to call .toUpperCase() on null, it crashes. This is where Groovy's safe navigation operator saves the day.
6.2. Safe Navigation Operator (?.)¶
The safe navigation operator (?.) returns null instead of throwing an exception when called on a null value. If the object before ?. is null, the entire expression evaluates to null without executing the method.
Update the function to use safe navigation:
| main.nf | |
|---|---|
| main.nf | |
|---|---|
Run again:
No crash! The workflow now handles the missing field gracefully. When row.run_id is null, the ?. operator prevents the .toUpperCase() call, and run_id becomes null instead of causing an exception.
6.3. Elvis Operator (?:) for Defaults¶
The Elvis operator (?:) provides default values when the left side is null (or empty, in Groovy's "truth" evaluation). It's named after Elvis Presley because ?: looks like his famous hair and eyes when viewed sideways!
Now that we're using safe navigation, run_id will be null for samples without that field. Let's use the Elvis operator to provide a default value and add it to our sample_meta map:
| main.nf | |
|---|---|
Also add a view() operator in the workflow to see the results:
and run the workflow:
You'll see output like this:
[[id:sample_001, organism:human, tissue:liver, depth:30000000, quality:38.5, run:UNSPECIFIED, sample_num:1, lane:001, read:R1, chunk:001, priority:normal], /workspaces/training/side-quests/groovy_essentials/data/sequences/SAMPLE_001_S1_L001_R1_001.fastq]
[[id:sample_002, organism:mouse, tissue:brain, depth:25000000, quality:35.2, run:UNSPECIFIED, sample_num:2, lane:001, read:R1, chunk:001, priority:normal], /workspaces/training/side-quests/groovy_essentials/data/sequences/SAMPLE_002_S2_L001_R1_001.fastq]
[[id:sample_003, organism:human, tissue:kidney, depth:45000000, quality:42.1, run:UNSPECIFIED, sample_num:3, lane:001, read:R1, chunk:001, priority:high], /workspaces/training/side-quests/groovy_essentials/data/sequences/SAMPLE_003_S3_L001_R1_001.fastq]
Perfect! Now all samples have a run field with either their actual run ID (in uppercase) or the default value 'UNSPECIFIED'. The combination of ?. and ?: provides both safety (no crashes) and sensible defaults.
Take out the .view() operator now that we've confirmed it works.
Combining Safe Navigation and Elvis
The pattern value?.method() ?: 'default' is common in production Nextflow:
value?.method()- Safely calls method, returnsnullifvalueisnull?: 'default'- Provides fallback if result isnull
This pattern handles missing/incomplete data gracefully.
Use these operators consistently in functions, operator closures (.map{}, .filter{}), process scripts, and config files. They prevent crashes when handling real-world data.
Takeaway¶
- Safe navigation (
?.): Prevents crashes on null values - returns null instead of throwing exception - Elvis operator (
?:): Provides defaults -value ?: 'default' - Combining:
value?.method() ?: 'default'is the common pattern
These operators make workflows resilient to incomplete data - essential for real-world bioinformatics.
7. Validation with error() and log.warn¶
Sometimes you need to stop the workflow immediately if input parameters are invalid. While error() and log.warn are Nextflow-provided functions, the validation logic itself is pure Groovy—using conditionals (if, !), boolean logic, and methods like .exists(). Let's add validation to our workflow.
Create a validation function before your workflow block, call it from the workflow, and change the channel creation to use a parameter for the CSV file path. If the parameter is missing or the file doesn't exist, call error() to stop execution with a clear message.
Now try running without the CSV file:
The workflow stops immediately with a clear error message instead of failing mysteriously later!
N E X T F L O W ~ version 25.04.6
Launching `main.nf` [confident_coulomb] DSL2 - revision: 07059399ed
WARN: Access to undefined parameter `input` -- Initialise it to a default value eg. `params.input = some_value`
Input CSV file path not provided. Please specify --input <file.csv>
You can also add validation within the separateMetadata function. Let's use the non-fatal log.warn to issue warnings for samples with low sequencing depth, but still allow the workflow to continue:
Run the workflow again with the original CSV:
... and you'll see a warning about low sequencing depth for one of the samples:
N E X T F L O W ~ version 25.04.6
Launching `main.nf` [awesome_goldwasser] DSL2 - revision: a31662a7c1
executor > local (5)
[ce/df5eeb] process > FASTP (2) [100%] 2 of 2 ✔
[- ] process > TRIMGALORE -
[d1/7d2b4b] process > GENERATE_REPORT (3) [100%] 3 of 3 ✔
WARN: Low sequencing depth for sample_002: 25000000
Takeaway¶
error(): Stops workflow immediately with clear messagelog.warn: Issues warnings without stopping workflow- Early validation: Check inputs before processing to fail fast with helpful errors
- Validation functions: Create reusable validation logic that can be called at workflow start
Proper validation makes workflows more robust and user-friendly by catching problems early with clear error messages.
8. Groovy in Configuration: Workflow Event Handlers¶
Up until now, we've been writing Groovy code in our workflow scripts and process definitions. But there's one more important place where Groovy is essential: workflow event handlers in your nextflow.config file (or other places you write configuration).
Event handlers are Groovy closures that run at specific points in your workflow's lifecycle. They're perfect for adding logging, notifications, or cleanup operations without cluttering your main workflow code.
8.1. The onComplete Handler¶
The most commonly used event handler is onComplete, which runs when your workflow finishes (whether it succeeded or failed). Let's add one to summarize our pipeline results.
Your nextflow.config file already has Docker enabled. Add an event handler after the existing configuration:
This is a Groovy closure being assigned to workflow.onComplete. Inside, you have access to the workflow object which provides useful properties about the execution.
Run your workflow and you'll see this summary appear at the end!
N E X T F L O W ~ version 25.04.6
Launching `main.nf` [marvelous_boltzmann] DSL2 - revision: a31662a7c1
WARN: Low sequencing depth for sample_002: 25000000
[9b/d48e40] Submitted process > FASTP (2)
[6a/73867a] Submitted process > GENERATE_REPORT (2)
[79/ad0ac5] Submitted process > GENERATE_REPORT (1)
[f3/bda6cb] Submitted process > FASTP (1)
[34/d5b52f] Submitted process > GENERATE_REPORT (3)
Pipeline execution summary:
==========================
Completed at: 2025-10-10T12:14:24.885384+01:00
Duration : 2.9s
Success : true
workDir : /Users/jonathan.manning/projects/training/side-quests/groovy_essentials/work
exit status : 0
Let's make it more useful by adding conditional logic:
Now we get an even more informative summary, including a success/failure message and the output directory if specified:
N E X T F L O W ~ version 25.04.6
Launching `main.nf` [boring_linnaeus] DSL2 - revision: a31662a7c1
WARN: Low sequencing depth for sample_002: 25000000
[e5/242efc] Submitted process > FASTP (2)
[3b/74047c] Submitted process > GENERATE_REPORT (3)
[8a/7a57e6] Submitted process > GENERATE_REPORT (1)
[a8/b1a31f] Submitted process > GENERATE_REPORT (2)
[40/648429] Submitted process > FASTP (1)
Pipeline execution summary:
==========================
Completed at: 2025-10-10T12:16:00.522569+01:00
Duration : 3.6s
Success : true
workDir : /Users/jonathan.manning/projects/training/side-quests/groovy_essentials/work
exit status : 0
✅ Pipeline completed successfully!
You can also write the summary to a file using Groovy file operations:
workflow.onComplete = {
def summary = """
Pipeline Execution Summary
===========================
Completed: ${workflow.complete}
Duration : ${workflow.duration}
Success : ${workflow.success}
Command : ${workflow.commandLine}
"""
println summary
// Write to a log file
def log_file = file("${workflow.launchDir}/pipeline_summary.txt")
log_file.text = summary
}
8.2. Other Useful Event Handlers¶
Besides onComplete, there are other event handlers you can use:
onStart - Runs when the workflow begins:
workflow.onStart = {
println "="* 50
println "Starting pipeline: ${workflow.runName}"
println "Project directory: ${workflow.projectDir}"
println "Launch directory: ${workflow.launchDir}"
println "Work directory: ${workflow.workDir}"
println "="* 50
}
onError - Runs only if the workflow fails:
workflow.onError = {
println "="* 50
println "Pipeline execution failed!"
println "Error message: ${workflow.errorMessage}"
println "="* 50
// Write detailed error log
def error_file = file("${workflow.launchDir}/error.log")
error_file.text = """
Workflow Error Report
=====================
Time: ${new Date()}
Error: ${workflow.errorMessage}
Error report: ${workflow.errorReport ?: 'No detailed report available'}
"""
println "Error details written to: ${error_file}"
}
You can use multiple handlers together:
workflow.onStart = {
println "Starting ${workflow.runName} at ${workflow.start}"
}
workflow.onError = {
println "Workflow failed: ${workflow.errorMessage}"
}
workflow.onComplete = {
def duration_mins = workflow.duration.toMinutes().round(2)
def status = workflow.success ? "SUCCESS ✅" : "FAILED ❌"
println """
Pipeline finished: ${status}
Duration: ${duration_mins} minutes
"""
}
Takeaway¶
In this section, you've learned:
- Event handler closures: Groovy closures in
nextflow.configthat run at different lifecycle points onCompletehandler: For execution summaries and result reportingonStarthandler: For logging pipeline initializationonErrorhandler: For error handling and logging failures- Workflow object properties: Accessing
workflow.success,workflow.duration,workflow.errorMessage, etc.
Event handlers are pure Groovy code running in your config file, demonstrating that Nextflow configuration is actually a Groovy script with access to the full language.
Summary¶
Throughout this side quest, you've built a comprehensive sample processing pipeline that evolved from basic metadata handling to a sophisticated, production-ready workflow. Each section built upon the previous, demonstrating how Groovy transforms simple Nextflow workflows into powerful data processing systems.
Here's how we progressively enhanced our pipeline:
-
Nextflow vs Groovy Boundaries: You learned to distinguish between workflow orchestration (Nextflow) and programming logic (Groovy), including the crucial differences between constructs like
collect. -
Advanced String Processing: You mastered regular expressions for parsing file names, dynamic script generation in processes, and variable interpolation (Groovy vs Bash vs Shell).
-
Creating Reusable Functions: You learned to extract complex logic into named functions that can be called from channel operators, making workflows more readable and maintainable.
-
Dynamic Resource Directives with Closures: You explored using Groovy closures in process directives for adaptive resource allocation based on input characteristics.
-
Conditional Logic and Process Control: You added intelligent routing using
.branch()and.filter()operators, leveraging Groovy Truth for concise conditional expressions. -
Safe Navigation and Elvis Operators: You made the pipeline robust against missing data using
?.for null-safe property access and?:for providing default values. -
Validation with error() and log.warn: You learned to validate inputs early and fail fast with clear error messages.
-
Groovy in Configuration: You learned to use workflow event handlers (
onComplete,onStart,onError) for logging, notifications, and lifecycle management.
Key Benefits¶
- Clearer code: Understanding when to use Nextflow and Groovy helps you write more organized workflows
- Robust handling: Safe navigation and Elvis operators make workflows resilient to missing data
- Flexible processing: Conditional logic lets your workflows process different sample types appropriately
- Adaptive resources: Dynamic directives optimize resource usage based on input characteristics
From Simple to Sophisticated¶
This pipeline evolved from basic data processing to production-ready workflows:
- Simple: CSV processing and metadata extraction (Nextflow vs Groovy boundaries)
- Intelligent: Regex parsing, variable interpolation, dynamic script generation
- Maintainable: Reusable functions for cleaner, testable code
- Efficient: Dynamic resource allocation and retry strategies
- Adaptive: Conditional routing based on sample characteristics
- Robust: Safe navigation, Elvis operators, early validation
- Observable: Event handlers for logging and lifecycle management
This progression mirrors the real-world evolution of bioinformatics pipelines - from research prototypes handling a few samples to production systems processing thousands of samples across laboratories and institutions. Every challenge you solved and pattern you learned reflects actual problems developers face when scaling Nextflow workflows.
Next Steps¶
With these Groovy fundamentals mastered, you're ready to:
- Write cleaner workflows with proper separation between Nextflow and Groovy logic
- Master variable interpolation to avoid common pitfalls with Groovy, Bash, and shell variables
- Use dynamic resource directives for efficient, adaptive workflows
- Transform file collections into properly formatted command-line arguments
- Handle different file naming conventions and input formats gracefully using regex and string processing
- Build reusable, maintainable code using advanced closure patterns and functional programming
- Process and organize complex datasets using collection operations
- Add validation, error handling, and logging to make your workflows production-ready
- Implement workflow lifecycle management with event handlers
Continue practicing these patterns in your own workflows, and refer to the Groovy documentation when you need to explore more advanced features.
Key Concepts Reference¶
- Language Boundaries
// Nextflow: workflow orchestration
Channel.fromPath('*.fastq').splitCsv(header: true)
// Groovy: data processing
sample_data.collect { it.toUpperCase() }
- String Processing
// Pattern matching
filename =~ ~/^(\w+)_(\w+)_(\d+)\.fastq$/
// Function with conditional return
def parseSample(filename) {
def matcher = filename =~ pattern
return matcher ? [valid: true, data: matcher[0]] : [valid: false]
}
// File collection to command arguments (in process script block)
script:
def file_args = input_files.collect { file -> "--input ${file}" }.join(' ')
"""
analysis_tool ${file_args} --output results.txt
"""
- Error Handling
try {
def errors = validateSample(sample)
if (errors) throw new RuntimeException("Invalid: ${errors.join(', ')}")
} catch (Exception e) {
println "Error: ${e.message}"
}
- Essential Groovy Operators
// Safe navigation and Elvis operators
def id = data?.sample?.id ?: 'unknown'
if (sample.files) println "Has files" // Groovy Truth
// Slashy strings for regex
def pattern = /^\w+_R[12]\.fastq$/
def script = """
echo "Processing ${sample.id}"
analysis --depth ${depth ?: 1_000_000}
"""
- Advanced Closures
// Named closures and composition
def enrichData = normalizeId >> addQualityCategory >> addFlags
def processor = generalFunction.curry(fixedParam)
// Closures with scope access
def collectStats = { data -> stats.count++; return data }
- Collection Operations