I have a checkpoints that leads to an unknown number of files, needing DAG reevaluation after execution. I wanted to speed it up by parallelizing the command by chromosome. Thus, after reevaluating, I would merge all chromosomes by experiment. However, snakemake is unable to infere the chromosomes.

checkpoint GenomeAnalysisTK:
    input:
        bamlist = rules.RealignerTargetCreator.output.bamlist,
        intervals = rules.RealignerTargetCreator.output.intervals,
        fasta = fasta
    output:
        temp(directory("splits/{chromosome}"))
    conda:
        "gatk3"
    wildcard_constraints:
            chromosome='|'.join([x for x in detect_chromosomes(fai)]),
    shell:
        """
        mkdir -p {output} && cd {output}
        gatk3 -Xmx24g -T IndelRealigner -I {input.bamlist} -targetIntervals {input.intervals} -L {wildcards.chromosome} -R {input.fasta} -compress 0 --nWayOut .{wildcards.chromosome}.indelrealigned.bam 
        """


def agg(wildcards):
    checkpoints.GenomeAnalysisTK.get(**wildcards).output[0]
    return expand("splits/{chromosome}/{{experiment}}.merged.{chromosome}.indelrealigned.bam", chromosome=get_chromosomes(fai))


rule merge_realigned:
    input:
        agg
    output:
        "{patient}/{sample}/{experiment}.merged.indelrealigned.bam"
    threads:
        config["other_threads"],
    params:
        compression_level = 0
    wildcard_constraints:
            chromosome='|'.join([x for x in detect_chromosomes(fai)]),
    shell:
        "samtools merge -@ {threads} -l {params.compression_level} {output} {input}"

However, I get the typical "WorkflowError: Missing wildcard values for chromosome". How can I make it infere the chromosomes? I think the main issue is that I scatter by chromosomes, but I don't use that wildcard anywhere else.