Support optional DuckDB-backed inputs

DESCRIPTION

@@ -50,7 +50,9 @@ Suggests:
     tidyseurat,
     tidySingleCellExperiment,
     bayesplot,
-    remotes
+    remotes,
+    duckdb,
+    dbplyr
 Additional_repositories:
     https://mc-stan.org/r-packages/
 SystemRequirements: CmdStan (https://mc-stan.org/users/interfaces/cmdstan), C++14

NAMESPACE

@@ -6,6 +6,7 @@ S3method(sccomp_calculate_residuals,sccomp_tbl)
 S3method(sccomp_estimate,DFrame)
 S3method(sccomp_estimate,Seurat)
 S3method(sccomp_estimate,SingleCellExperiment)
+S3method(sccomp_estimate,tbl_duckdb_connection)
 S3method(sccomp_estimate,data.frame)
 S3method(sccomp_predict,sccomp_tbl)
 S3method(sccomp_proportional_fold_change,sccomp_tbl)

R/sccomp_estimate.R

@@ -532,6 +532,13 @@ sccomp_estimate.DFrame <- function(.data,
     )
 }
 
+#' @export
+sccomp_estimate.tbl_duckdb_connection <- function(.data, ...) {
+  check_and_install_packages(c("duckdb", "dbplyr"))
+
+  sccomp_estimate.data.frame(dplyr::collect(.data), ...)
+}
+
 
 #' @importFrom purrr when
 #' @importFrom rlang is_symbolic

tests/testthat/test-duckdb.R

@@ -0,0 +1,46 @@
+test_that("sccomp_estimate has a DuckDB table method", {
+  expect_true(is.function(getS3method("sccomp_estimate", "tbl_duckdb_connection")))
+})
+
+test_that("sccomp_estimate gives the same results for DuckDB tables", {
+  skip_if_not_installed("duckdb")
+  skip_if_not_installed("dbplyr")
+  skip_cmdstan()
+
+  data("counts_obj", package = "sccomp", envir = environment())
+  con <- DBI::dbConnect(duckdb::duckdb(), dbdir = ":memory:")
+  on.exit(DBI::dbDisconnect(con, shutdown = TRUE), add = TRUE)
+  DBI::dbWriteTable(con, "counts_obj", counts_obj)
+
+  estimate_args <- list(
+    formula_composition = ~type,
+    sample = "sample",
+    cell_group = "cell_group",
+    abundance = "count",
+    inference_method = "pathfinder",
+    cores = 1,
+    mcmc_seed = 12345,
+    max_sampling_iterations = 1000,
+    verbose = FALSE
+  )
+
+  data_frame_estimate <- do.call(
+    sccomp_estimate,
+    c(list(counts_obj), estimate_args)
+  )
+  duckdb_estimate <- do.call(
+    sccomp_estimate,
+    c(list(dplyr::tbl(con, "counts_obj")), estimate_args)
+  )
+
+  data_frame_results <- data_frame_estimate |>
+      dplyr::arrange(cell_group, parameter) |>
+      dplyr::select(cell_group, parameter, c_effect, c_lower, c_upper) |>
+      as.data.frame()
+  duckdb_results <- duckdb_estimate |>
+      dplyr::arrange(cell_group, parameter) |>
+      dplyr::select(cell_group, parameter, c_effect, c_lower, c_upper) |>
+      as.data.frame()
+
+  expect_equal(data_frame_results, duckdb_results, tolerance = 1e-2)
+})