Merge branch 'CW-7265' into 'dev'

Bambu always 1 thread [CW-7265]

Closes CW-7265

See merge request epi2melabs/workflows/wf-transcriptomes!285

bin/workflow_glue_r/R/bambu.R

@@ -64,14 +64,6 @@ bambu_arg_spec <- function() {
             default = "discover",
             choices = c("discover", "fixed_annotation")
         ),
-        list(
-            name = "threads",
-            flag = "--threads",
-            help = "Number of worker threads.",
-            type = "integer",
-            default = 1L,
-            min = 1L
-        ),
         list(
             name = "ndr",
             flag = "--ndr",
@@ -258,11 +250,13 @@ bambu_discovery_enabled <- function(args) {
 }
 
 bambu_build_args <- function(args, reads, annotation_obj, discovery, quant) {
+    # Pin bambu to ncore=1 in all modes to avoid parallel worker instability.
+    ncore <- 1L
     bambu_args <- list(
         reads = reads,
         annotations = annotation_obj,
         genome = args$genome,
-        ncore = as.integer(args$threads),
+        ncore = ncore,
         discovery = discovery,
         quant = quant,
         lowMemory = TRUE,
@@ -301,25 +295,6 @@ bambu_message_ndr <- function(args) {
         message(sprintf("  Current NDR = %.3f balances precision and recall", args$ndr))
     }
 }
-
-bambu_effective_threads <- function(args, bam_count) {
-    # bambu's low-memory mode can have issues with multiple BAMs and
-    # parallel threads due to BiocFileCache writes,
-    # so we enforce single-threading in that case.
-    threads <- as.integer(args$threads)
-    if (bam_count > 1 && threads > 1L) {
-        warning(
-            paste(
-                "Low-memory mode with multiple BAMs can fail in bambu due to",
-                "parallel BiocFileCache writes; forcing threads=1."
-            ),
-            call. = FALSE
-        )
-        return(1L)
-    }
-    threads
-}
-
 bambu_normalise_rc_file_list <- function(rc_files, aliases = NULL) {
     if (!is.list(rc_files)) {
         rc_files <- list(rc_files)
@@ -409,13 +384,12 @@ bambu_write_discovery_outputs <- function(out_dir, rc_files, discovered_annotati
 
 bambu_run_discover_mode <- function(args, analysis_fn, prepare_annotations_fn, bamfile_list_ctor) {
     inputs <- bambu_resolve_inputs(args, bamfile_list_ctor = bamfile_list_ctor)
-    args$threads <- bambu_effective_threads(args, length(inputs$bam_paths))
     annotation_obj <- prepare_annotations_fn(args$annotation)
 
     if (length(inputs$bam_paths) > 1) {
         message(sprintf("Using BamFileList yieldSize = %d", bambu_default_yield_size))
     }
-    message(sprintf("Running bambu discover setup with threads = %d", args$threads))
+    message("Running bambu discover setup with ncore = 1")
     message("Generating bambu rcFiles...")
     rc_files <- bambu_call_analysis(
         analysis_fn,

bin/workflow_glue_r/tests/testthat/test_bambu.R

@@ -101,11 +101,6 @@ testthat::test_that("invalid discovery settings rejected", {
     testthat::expect_silent(workflow_glue_r_normalise_args(args, bambu_arg_spec()))
     args$ndr <- 1
     testthat::expect_silent(workflow_glue_r_normalise_args(args, bambu_arg_spec()))
-
-    args$ndr <- NULL
-    args$threads <- "2"
-    normalised <- workflow_glue_r_normalise_args(args, bambu_arg_spec())
-    testthat::expect_identical(normalised$threads, 2L)
 })
 
 # Fail fast if --bams is empty rather than passing empty input to bambu.
@@ -259,7 +254,6 @@ testthat::test_that("bambu args include requested discovery and quant flags", {
 
     args <- list(
         genome = "genome.fa",
-        threads = 3L,
         transcriptome_mode = "discover",
         ndr = 0.2
     )
@@ -273,7 +267,7 @@ testthat::test_that("bambu args include requested discovery and quant flags", {
     testthat::expect_true(discover$discovery)
     testthat::expect_false(discover$quant)
     testthat::expect_equal(discover$NDR, 0.2)
-    testthat::expect_equal(discover$ncore, 3L)
+    testthat::expect_equal(discover$ncore, 1L)
     testthat::expect_true(discover$lowMemory)
     testthat::expect_equal(discover$yieldSize, 250000L)
 
@@ -376,8 +370,7 @@ testthat::test_that("discover mode writes chunked rc outputs", {
             aliases = "sampleA,sampleB",
             sample_sheet = sample_sheet,
             transcriptome_mode = "discover",
-            ndr = 0.25,
+            ndr = 0.25
-            threads = 2
         ),
         bambu_arg_spec()
     )
@@ -497,8 +490,7 @@ testthat::test_that("quant mode writes chunk quantification outputs", {
             chunk_rds = chunk_rds,
             discovered_annotation_rds = discovered_annotation_rds,
             transcriptome_mode = "discover",
-            ndr = NULL,
+            ndr = NULL
-            threads = 2
         ),
         bambu_arg_spec()
     )
@@ -562,8 +554,7 @@ testthat::test_that("quant mode skips chunks with no discovered annotations on t
             chunk_rds = chunk_rds,
             discovered_annotation_rds = discovered_annotation_rds,
             transcriptome_mode = "discover",
-            ndr = NULL,
+            ndr = NULL
-            threads = 2
         ),
         bambu_arg_spec()
     )
@@ -1000,7 +991,6 @@ testthat::test_that("CLI discover writes reusable chunk artifacts", {
             "--annotation", annotation,
             "--genome", reference,
             "--transcriptome_mode", "fixed_annotation",
-            "--threads", "1",
             "--out_dir", out_dir
         )
     )
@@ -1064,7 +1054,6 @@ testthat::test_that("CLI quant consumes a discover chunk", {
             "--annotation", annotation,
             "--genome", reference,
             "--transcriptome_mode", "fixed_annotation",
-            "--threads", "1",
             "--out_dir", discover_out_dir
         )
     )
@@ -1089,7 +1078,6 @@ testthat::test_that("CLI quant consumes a discover chunk", {
             "--chunk_rds", manifest$rds_path[[1]],
             "--discovered_annotation_rds", file.path(discover_out_dir, "bambu_discovered_annotations.rds"),
             "--genome", reference,
-            "--threads", "1",
             "--out_dir", out_dir
         )
     )
@@ -1147,7 +1135,6 @@ testthat::test_that("CLI collate consumes quant chunk directories", {
             "--annotation", annotation,
             "--genome", reference,
             "--transcriptome_mode", "fixed_annotation",
-            "--threads", "1",
             "--out_dir", discover_out_dir
         )
     )
@@ -1172,7 +1159,6 @@ testthat::test_that("CLI collate consumes quant chunk directories", {
             "--chunk_rds", manifest$rds_path[[1]],
             "--discovered_annotation_rds", file.path(discover_out_dir, "bambu_discovered_annotations.rds"),
             "--genome", reference,
-            "--threads", "1",
             "--out_dir", chunk_out_dir
         )
     )

modules/local/bambu_chunked.nf

@@ -5,11 +5,7 @@ OPTIONAL_FILE = file("$projectDir/data/OPTIONAL_FILE")
 
 process bambuDiscover {
     label "wf_transcriptomes"
-    cpus {
+    cpus 1
-        int requested = (params.threads ?: 4) as int
-        int sampleCount = aliases instanceof Collection ? aliases.size() : 1
-        sampleCount > 1 ? requested : 1
-    }
     memory "60 GB"
     input:
         tuple val(meta), val(aliases), path(bams, stageAs: "bams/??.bam"), path(bais, stageAs: "bams/??.bam.bai"), path(sample_sheet)
@@ -32,7 +28,6 @@ process bambuDiscover {
         --annotation "${annotation}" \
         --genome "${reference}" \
         --transcriptome_mode "${params.transcriptome_mode}" \
-        --threads ${task.cpus} \
         ${ndr_arg} \
         --out_dir discover
     """
@@ -41,11 +36,7 @@ process bambuDiscover {
 
 process bambuQuant {
     label "wf_transcriptomes"
-    cpus {
+    cpus 1
-        int requested = (params.threads ?: 4) as int
-        boolean isJoint = meta instanceof Map && meta.alias == 'cohort'
-        isJoint ? requested : 1
-    }
     memory { ["8.GB", "16.GB", "48.GB"][task.attempt - 1] }
     maxRetries 2
     errorStrategy 'retry'
@@ -60,7 +51,6 @@ process bambuQuant {
         --chunk_rds "${chunk_rds}" \
         --discovered_annotation_rds "${discovered_annotation}" \
         --genome "${reference}" \
-        --threads ${task.cpus} \
         --out_dir "${chunk_id}"
     """
 }