Chunk RDS from disk if provided a path [CW-7338]

CHANGELOG.md

@@ -7,11 +7,12 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [v2.0.1]
 
-This patch release of `wf-transcriptomes` handles an additional quantification failure case that was not observed before release.
-Users of wf-transcriptomes v2.0.0 who have encountered issues during quantification should adopt this release.
+This patch release of `wf-transcriptomes` handles an additional quantification failure edge case that was not observed before release, and fixes an issue encountered during joint discovery for many samples.
+Users of wf-transcriptomes v2.0.0 who have encountered issues during discovery and quantification should adopt this release.
 
 ### Fixed
-- Error in full_join encountered during runPerSampleBambuQuant when all read classes have no compatible transcript assignment.
+- "Error in full_join" encountered during `runPerSampleBambuQuant` when all read classes have no compatible transcript assignment. An empty quant table is correctly emitted instead.
+- "unable to find an inherited method for function 'rowData'" encountered during `runJointBambuDiscover` when providing many samples. The workflow now correctly handles data spilled to disk by bambu discover.
 
 
 ## [v2.0.0]

bin/workflow_glue_r/R/bambu.R

@@ -280,6 +280,22 @@ bambu_normalise_rc_file_list <- function(rc_files, aliases = NULL) {
     rc_files
 }
 
+bambu_load_rc_file <- function(rc_file) {
+    if (is.character(rc_file) && length(rc_file) == 1L) {
+        rc_file <- readRDS(rc_file)
+    }
+    if (!methods::is(rc_file, "RangedSummarizedExperiment")) {
+        stop(
+            sprintf(
+                "Expected bambu read-class output to be a RangedSummarizedExperiment or RDS path, got '%s'.",
+                paste(class(rc_file), collapse = ", ")
+            ),
+            call. = FALSE
+        )
+    }
+    rc_file
+}
+
 bambu_chunk_id_for_seqname <- function(seqname) {
     seqname <- as.character(seqname)
     seqname <- gsub("[^A-Za-z0-9._-]+", "_", seqname)
@@ -293,6 +309,8 @@ bambu_chunk_id_for_seqname <- function(seqname) {
 
 bambu_chunk_rc_files <- function(rc_files, aliases, sample_df) {
     rc_files <- bambu_normalise_rc_file_list(rc_files, aliases = aliases)
+    # bambu may provide a RangedSummarizedExperiment or a path to an RDS spilled to disk [CW-7338]
+    rc_files <- lapply(rc_files, bambu_load_rc_file)
     seqnames <- unique(unlist(lapply(rc_files, function(rcf) {
         as.character(SummarizedExperiment::rowData(rcf)$chr.rc)
     })))

bin/workflow_glue_r/tests/testthat/test_bambu.R

@@ -383,6 +383,33 @@ testthat::test_that("discover mode writes chunked rc outputs", {
     testthat::expect_equal(chunk_bundle$aliases, c("sampleA", "sampleB"))
 })
 
+testthat::test_that("cw-7338 chunking accepts path-backed rc outputs", {
+    fixture_dir <- tempfile("bambu-path-backed-rc-")
+    dir.create(fixture_dir)
+
+    make_rc_sample <- function(alias) {
+        rcf <- make_test_tx_se(sample_names = alias)
+        S4Vectors::mcols(SummarizedExperiment::rowRanges(rcf))$chr.rc <- c("chr1", "chr1", "chr2", "chr2")
+        rcf
+    }
+
+    rc_paths <- file.path(fixture_dir, paste0(c("sampleA", "sampleB"), "_readClassSe.rds"))
+    saveRDS(make_rc_sample("sampleA"), rc_paths[[1]])
+    saveRDS(make_rc_sample("sampleB"), rc_paths[[2]])
+    names(rc_paths) <- c("sampleA", "sampleB")
+
+    chunk_bundles <- bambu_chunk_rc_files(
+        as.list(rc_paths),
+        aliases = c("sampleA", "sampleB"),
+        sample_df = data.frame(alias = c("sampleA", "sampleB"), stringsAsFactors = FALSE)
+    )
+
+    testthat::expect_equal(unname(sort(vapply(chunk_bundles, `[[`, character(1), "seqname"))), c("chr1", "chr2"))
+    testthat::expect_equal(chunk_bundles[[1]]$aliases, c("sampleA", "sampleB"))
+    testthat::expect_s4_class(chunk_bundles[[1]]$rc_files$sampleA, "RangedSummarizedExperiment")
+    testthat::expect_s4_class(chunk_bundles[[1]]$rc_files$sampleB, "RangedSummarizedExperiment")
+})
+
 testthat::test_that("annotation tx counts are computed once per seqname", {
     discovered_annotations <- GenomicRanges::GRangesList(
         tx1 = GenomicRanges::GRanges(seqnames = "chr1", ranges = IRanges::IRanges(c(1, 10), width = 5)),