/dev/shm/

#!/usr/bin/env nextflow
// Standalone wrapper around FCSGX_RUNGX that stages the GX database into a RAM disk.
//
//   nextflow run <http://rungx_wrapper.nf|rungx_wrapper.nf> -c rungx_wrapper.config \
//       --assembly 'asm/*.fasta.gz' --taxid 9606 --gxdb /path/to/gxdb
//
include { FCSGX_RUNGX } from "./modules/local/fcsgx/rungx"

params.assembly    = null          // FASTA file or glob
params.taxid       = null          // NCBI taxid of the sample
params.gxdb        = null          // on-disk GX database directory (source for the RAM-disk copy)
params.ramdisk_path = '/dev/shm/gxdb'  // destination in RAM; removed when the task exits
params.outdir      = 'results'

workflow {
    main:
    if( !params.assembly || !params.taxid || !params.gxdb ) {
        error "Required: --assembly, --taxid, --gxdb"
    }
    ch_assembly = channel.fromPath( params.assembly, checkIfExists: true )
        .map { fasta -> [ [ id: fasta.name.replaceFirst(/\.(fa|fasta|fna)(\.gz)?$/, '') ], params.taxid, fasta ] }
    ch_gxdb = channel.fromPath( params.gxdb, checkIfExists: true, type: 'dir' ).collect()

    FCSGX_RUNGX( ch_assembly, ch_gxdb, params.ramdisk_path )

    publish:
    reports = FCSGX_RUNGX.out.fcs_gx_report
    taxonomy = FCSGX_RUNGX.out.taxonomy_report
    hits = FCSGX_RUNGX.out.hits
}

output {
    reports  { path 'fcsgx' }
    taxonomy { path 'fcsgx' }
    hits     { path 'fcsgx' }
}

outputDir = params.outdir

process {
    withName: 'FCSGX_RUNGX' {
        // One task at a time: all tasks share the single RAM-disk copy of the DB.
        maxForks = 1
        cpus     = 20
        memory   = 600.GB   // must cover the ~470 GB DB plus run_gx working memory
        time     = 1.h
    }
}

FAST_TEMP_DIR

/dev/shm