Make BBSPLIT indexing behave itself (#5005)

modules/nf-core/bbmap/bbsplit/main.nf

@@ -20,6 +20,7 @@ process BBMAP_BBSPLIT {
     tuple val(meta), path('*primary*fastq.gz'), optional:true, emit: primary_fastq
     tuple val(meta), path('*fastq.gz')        , optional:true, emit: all_fastq
     tuple val(meta), path('*txt')             , optional:true, emit: stats
+    tuple val(meta), path('*.log')            , optional:true, emit: log
     path "versions.yml"                       , emit: versions
 
     when:
@@ -40,51 +41,69 @@ process BBMAP_BBSPLIT {
     other_ref_names.eachWithIndex { name, index ->
         other_refs << "ref_${name}=${other_ref_paths[index]}"
     }
+
+    def fastq_in=''
+    def fastq_out=''
+    def index_files=''
+    def refstats_cmd=''
+
     if (only_build_index) {
+        println("only building index")
         if (primary_ref && other_ref_names && other_ref_paths) {
-            """
-            bbsplit.sh \\
-                -Xmx${avail_mem}M \\
-                ref_primary=$primary_ref \\
-                ${other_refs.join(' ')} \\
-                path=bbsplit \\
-                threads=$task.cpus \\
-                $args
-
-            cat <<-END_VERSIONS > versions.yml
-            "${task.process}":
-                bbmap: \$(bbversion.sh | grep -v "Duplicate cpuset")
-            END_VERSIONS
-            """
+            index_files = 'ref_primary=' +primary_ref + ' ' + other_refs.join(' ') + ' path=bbsplit'
         } else {
             log.error 'ERROR: Please specify as input a primary fasta file along with names and paths to non-primary fasta files.'
         }
     } else {
-        def index_files = ''
+        index_files = ''
         if (index) {
             index_files = "path=$index"
         } else if (primary_ref && other_ref_names && other_ref_paths) {
             index_files = "ref_primary=${primary_ref} ${other_refs.join(' ')}"
         } else {
             log.error 'ERROR: Please either specify a BBSplit index as input or a primary fasta file along with names and paths to non-primary fasta files.'
         }
-        def fastq_in  = meta.single_end ? "in=${reads}" : "in=${reads[0]} in2=${reads[1]}"
-        def fastq_out = meta.single_end ? "basename=${prefix}_%.fastq.gz" : "basename=${prefix}_%_#.fastq.gz"
-        """
-        bbsplit.sh \\
-            -Xmx${avail_mem}M \\
-            $index_files \\
-            threads=$task.cpus \\
-            $fastq_in \\
-            $fastq_out \\
-            refstats=${prefix}.stats.txt \\
-            $args
-
-        cat <<-END_VERSIONS > versions.yml
-        "${task.process}":
-            bbmap: \$(bbversion.sh | grep -v "Duplicate cpuset")
-        END_VERSIONS
-        """
+        fastq_in  = meta.single_end ? "in=${reads}" : "in=${reads[0]} in2=${reads[1]}"
+        fastq_out = meta.single_end ? "basename=${prefix}_%.fastq.gz" : "basename=${prefix}_%_#.fastq.gz"
+        refstats_cmd = 'refstats=' + prefix + '.stats.txt'
     }
+    """
+
+    # When we stage in the index files the time stamps get disturbed, which
+    # bbsplit doesn't like. Fix the time stamps in its summaries. This needs to
+    # be done via Java to match what bbmap does
+
+    if [ $index ]; then
+        for summary_file in \$(find $index/ref/genome -name summary.txt); do
+            src=\$(grep '^source' "\$summary_file" | cut -f2- -d\$'\\t' | sed 's|.*/bbsplit|bbsplit|')
+            mod=\$(echo "System.out.println(java.nio.file.Files.getLastModifiedTime(java.nio.file.Paths.get(\\"\$src\\")).toMillis());" | jshell -J-Djdk.lang.Process.launchMechanism=vfork -)
+            sed "s|^last modified.*|last modified\\t\$mod|" "\$summary_file" > \${summary_file}.tmp && mv \${summary_file}.tmp \${summary_file}
+        done
+    fi
+
+    # Run BBSplit
+
+    bbsplit.sh \\
+        -Xmx${avail_mem}M \\
+        $index_files \\
+        threads=$task.cpus \\
+        $fastq_in \\
+        $fastq_out \\
+        $refstats_cmd \\
+        $args 2> >(tee ${prefix}.log >&2)
+
+    # Summary files will have an absolute path that will make the index
+    # impossible to use in other processes- we can fix that
+
+    for summary_file in \$(find bbsplit/ref/genome -name summary.txt); do
+        src=\$(grep '^source' "\$summary_file" | cut -f2- -d\$'\\t' | sed 's|.*/bbsplit|bbsplit|')
+        sed "s|^source.*|source\\t\$src|" "\$summary_file" > \${summary_file}.tmp && mv \${summary_file}.tmp \${summary_file}
+    done
+
+    cat <<-END_VERSIONS > versions.yml
+    "${task.process}":
+        bbmap: \$(bbversion.sh | grep -v "Duplicate cpuset")
+    END_VERSIONS
+    """
 
 }

modules/nf-core/bbmap/bbsplit/tests/main.nf.test

@@ -45,7 +45,7 @@ nextflow_process {
                         [ 'human' ], // meta map
                         file(params.modules_testdata_base_path + 'genomics/homo_sapiens/genome/chr22/sequence/chr22_23800000-23980000.fa', checkIfExists: true)
                     ])
-                    input[4] = true
+                    input[4] = false
                     """
                 }
         }
@@ -81,6 +81,7 @@ nextflow_process {
 
             assertAll(
                 { assert process.success },
+                { assert path(process.out.log[0][1]).text.contains("If you wish to regenerate the index") },
                 { assert snapshot(filteredFiles).match("bbsplit_index_filtered_files")},
                 { assert filesExist : "One or more files to exclude do not exist" },
                 { assert snapshot(process.out.versions).match("versions")}