Nextflow: Proof of concept example for repeated function call with expected static output

nextflow_sample_map_str.nf

#!/usr/bin/env nextflow

/* this is a minimal working example for a bug we recently ran into.
 * in a nutshell: we can function (gen_sample_map_str) in every call to
 * process GenomicsDB.  * the function has static input and should
 * produce identical output each time, but it doesn't. the question is
 * why.
 *
 * yes, we can do tihs different, i.e. construct the string once and yes
 * some things don't make sense in this current minimalistic form.
 * but the code should in theory nevertheless work, yet doesn't.
 */


params.publishdir = 'out'


// a sample map akin to what's used in GATK's GenomicsDB
params.sample_name_map = [
    'sample1': 's1.g.vcf.gz',
    'sample2': 's2.g.vcf.gz',
    'sample3': 's3.g.vcf.gz',
    'sample4': 's4.g.vcf.gz',
    'sample5': 's5.g.vcf.gz',
    'sample6': 's6.g.vcf.gz',
    'sample7': 's7.g.vcf.gz',
    'sample8': 's8.g.vcf.gz',
    'sample9': 's9.g.vcf.gz',
    'sample10': 's10.g.vcf.gz'
]


// fake regions
region_ch = Channel.from([1, 2, 3, 4, 5, 6, 7, 8, 9, 10])


// constructing a sample map string from params.sample_name_map.the
// question is, why do repeated calls give different results?
def gen_sample_map_str() {
    str = ""
    params.sample_name_map.each{ k, v ->
        str += "${k}\t${v}\n"
    }
    return str;
}


// just prints output of gen_sample_map_str() to a file. in theory all
// files should have at least identical number of lines, but they
// don't (see output generated in Validate)
//
process GenomicsDBImport {
    publishDir params.publishdir

    input:
    val reg from region_ch
    
    output:
    file("${reg}.txt") into final_ch

    script:
    sample_name_map_str = gen_sample_map_str()
    """
    echo "${sample_name_map_str}" > ${reg}.txt;
    """
}


process Validate {
    input:
    file(all) from final_ch.collect()

    output:
    stdout wc_ch
    
    script:
    """
    wc -l ${all}
    """
}


wc_ch.subscribe { print "Line numbers should be identical but are not (forget about order):\n  $it" }

Adding -process.cpus=1 or -qs 1 as arguments doesn't fix the behaviour