fix match_combine cache

conf/modules.config

@@ -33,11 +33,6 @@ process {
         ext.args2 = "zs"  // compress .sscore with zstd by default
     }
 
-    withName: MATCH_COMBINE {
-        // don't recombine unless the file content has changed
-        cache = 'deep'
-    }
-
     // container configuration    
     withLabel: pgscatalog_utils {
         ext.conda = "$projectDir/environments/pgscatalog_utils/environment.yml"

modules/local/match_combine.nf

@@ -28,9 +28,9 @@ process MATCH_COMBINE {
     def ambig = params.keep_ambiguous       ? '--keep_ambiguous'    : ''
     def multi = params.keep_multiallelic    ? '--keep_multiallelic' : ''
     // output one (or more) scoring files per chromosome?
-    def split_output = !meta.chrom.contains("ALL") ? '--split' : ''
+    def split_output = meta.split ? '--split' : ''
     // output one (or more) scoring file per sampleset?
-    def combined_output = (meta.chrom.contains("ALL") || shared.name != 'NO_FILE') ? '--combined' : ''
+    def combined_output = (!meta.split || shared.name != 'NO_FILE') ? '--combined' : ''
     // filter match candidates to intersect with reference:
     // omit multi-allelic variants in reference because these will cause errors with relabelling!...
     // ... unclear whether we should remove them from target with '&& ($9 == 0') as well?

subworkflows/local/match.nf

@@ -42,9 +42,14 @@ workflow MATCH {
         }
         .set { ch_intersection_grouped }
 
-    // grab first() meta object for MATCH_COMBINE
     // only meta.chrom is checked to see if it's set to 'ALL' or not
-    ch_matches.meta.first()
+    // but using chrom values directly in meta map breaks cache because chrom order can differ across runs
+    ch_matches.meta.first().map { it -> 
+        def split = it.chrom == "ALL"
+        return [split:split, id: it.id]
+    }.set { combine_meta }
+
+    combine_meta
         .concat( ch_matches.matches.collect() )
         .concat( scorefile )
         .concat( ch_intersection_grouped.intersections.collect() )