export `generate_dataset()` function #152

rcannood · 2023-12-04T08:13:43Z

Changes:

Re-exports the dataset generator function.
Reworked the arguments of the dataset generator function
Renamed internal functions
Fixed print method

Closes #136.

lazappi

I've made some comments but I think it's pretty good. I would also update the examples to use this but maybe not everyone wants to do that.

R/generate_dataset.R

lazappi · 2023-12-05T08:22:57Z

R/generate_dataset.R

+    layer_types = c(
+      "numeric_matrix", "numeric_dense", "numeric_csparse", "numeric_rsparse", "numeric_matrix_with_nas", #
+      "numeric_dense_with_nas", "numeric_csparse_with_nas", "numeric_rsparse_with_nas", "integer_matrix",
+      "integer_dense", "integer_csparse", "integer_rsparse", "integer_matrix_with_nas", "integer_dense_with_nas",
+      "integer_csparse_with_nas", "integer_rsparse_with_nas"
+    ),
+    obs_types = c(
+      "character", "integer", "factor", "factor_ordered", "logical", "numeric", "character_with_nas",
+      "integer_with_nas", "factor_with_nas", "factor_ordered_with_nas", "logical_with_nas", "numeric_with_nas"
+    ),
+    var_types = c(
+      "character", "integer", "factor", "factor_ordered", "logical", "numeric", "character_with_nas",
+      "integer_with_nas", "factor_with_nas", "factor_ordered_with_nas", "logical_with_nas", "numeric_with_nas"
+    ),
+    obsm_types = c(
+      "numeric_matrix", "numeric_dense", "numeric_csparse", "numeric_rsparse", "numeric_matrix_with_nas",
+      "numeric_dense_with_nas", "numeric_csparse_with_nas", "numeric_rsparse_with_nas", "integer_matrix",
+      "integer_dense", "integer_csparse", "integer_rsparse", "integer_matrix_with_nas", "integer_dense_with_nas",
+      "integer_csparse_with_nas", "integer_rsparse_with_nas", "character", "integer", "factor", "factor_ordered",
+      "logical", "numeric", "character_with_nas", "integer_with_nas", "factor_with_nas", "factor_ordered_with_nas",
+      "logical_with_nas", "numeric_with_nas"
+    ),
+    varm_types = c(
+      "numeric_matrix", "numeric_dense", "numeric_csparse", "numeric_rsparse", "numeric_matrix_with_nas",
+      "numeric_dense_with_nas", "numeric_csparse_with_nas", "numeric_rsparse_with_nas", "integer_matrix",
+      "integer_dense", "integer_csparse", "integer_rsparse", "integer_matrix_with_nas", "integer_dense_with_nas",
+      "integer_csparse_with_nas", "integer_rsparse_with_nas", "character", "integer", "factor", "factor_ordered",
+      "logical", "numeric", "character_with_nas", "integer_with_nas", "factor_with_nas", "factor_ordered_with_nas",
+      "logical_with_nas", "numeric_with_nas"
+    ),
+    obsp_types = c(
+      "numeric_matrix", "numeric_dense", "numeric_csparse", "numeric_rsparse", "numeric_matrix_with_nas",
+      "numeric_dense_with_nas", "numeric_csparse_with_nas", "numeric_rsparse_with_nas", "integer_matrix",
+      "integer_dense", "integer_csparse", "integer_rsparse", "integer_matrix_with_nas", "integer_dense_with_nas",
+      "integer_csparse_with_nas", "integer_rsparse_with_nas"
+    ),
+    varp_types = c(
+      "numeric_matrix", "numeric_dense", "numeric_csparse", "numeric_rsparse", "numeric_matrix_with_nas",
+      "numeric_dense_with_nas", "numeric_csparse_with_nas", "numeric_rsparse_with_nas", "integer_matrix",
+      "integer_dense", "integer_csparse", "integer_rsparse", "integer_matrix_with_nas", "integer_dense_with_nas",
+      "integer_csparse_with_nas", "integer_rsparse_with_nas"
+    ),


I like that all these are included but maybe we could make an easy way to generate a reduced list for examples?

I see. Not sure how we could go about this, other than perhaps:

generate_dummy_dataset <- function(n_obs, n_vars) { generate_dataset( n_obs = n_obs, n_vars = n_vars, obs_types = c(...), # small set of types ... ) }

Aha, wait! We could turn it around. We could set the defaults generate_dataset() to something simple, and then create a helper function in the tests which contains all of the types.

I implemented a solution for this in 206f11a

lazappi · 2023-12-05T08:24:33Z

R/generate_dataset.R

-  X <- t(dummy$layers[["integer_csparse"]])
-  colnames(X) <- dummy$obs_names
-  rownames(X) <- dummy$var_names
+  X <- t(list$layers[["integer_csparse"]])


Should this be list$X in case integer_csparse wasn't generated? Same with the ones below.

True -- though I think the Seurat conversion should be revised in light of Seuratv5. Maybe we can just pass all layers to Seurat without needing to be selective.

lazappi · 2023-12-05T08:27:54Z

R/get_generator.R

@@ -10,6 +12,8 @@ list_generators <- function() {
 #'
 #' @param class Name of the AnnData class. Must be one of `"HDF5AnnData"`
 #' or `"InMemoryAnnData"`.
+#'
+#' @noRd
 get_generator <- function(class = c("HDF5AnnData", "InMemoryAnnData")) {
  # TODO: also support directly passing the correct class?


I'm not quite sure what the purpose of this is. Is generator here different to the dataset generators?

Yes -- the generator function is used by the from_Seurat and from_SingleCellExperiment to create the correct type of AnnData object:

from_SingleCellExperiment <- function(sce, output_class = c("InMemory", "HDF5AnnData"), ...) { # nolint stopifnot( inherits(sce, "SingleCellExperiment") ) # fetch generator generator <- get_generator(output_class) # ... create obs, var, ... generator$new( X = x, obs = obs, var = var, obs_names = obs_names, var_names = var_names, layers = layers, ... ) }

I agree, however, that the term generator is now somewhat confusing, given that there is a generate_dataset() function. Fortunately, the get_generator function is an internal one, so we can rename it to whatever we like :)

Yeah, it might be better to have a more specific name. object_generator or converter maybe?

Good point! I renamed the generator to anndata constructor in 5093884

tests/testthat/test-generate_dataset.R

Co-authored-by: Luke Zappia <[email protected]>

* origin/main: Version update (#161) re-enable tests for writing compressed files (#140) re-enable tests (#139) re-enable tests (#138) Add stripTrailingWhitespace option to .Rproj (#157) export `generate_dataset()` function (#152)

rcannood added 8 commits December 4, 2023 09:11

export generate_dataset() function

f57a410

no need to export these docs

e288779

update docs

5e823ed

rename / restructure functions

86a515e

store types directly in generate_dataset function

c2cf9e8

fix ci

98b5e45

actually export the function as well

029fb44

add missing types

9918aa4

rcannood requested a review from lazappi December 4, 2023 10:06

Merge remote-tracking branch 'origin/main' into export_dataset_generator

311bf07

lazappi approved these changes Dec 5, 2023

View reviewed changes

rcannood and others added 2 commits December 5, 2023 10:53

Rename list to dataset_list to avoid confusion

dae2561

Co-authored-by: Luke Zappia <[email protected]>

add example and uns_types arguments to generate_dataset()

206f11a

rcannood mentioned this pull request Dec 5, 2023

Pass all R CMD and BiocChecks #135

Closed

rcannood added 7 commits December 5, 2023 15:54

rename 'generator' to 'anndata_constructor'

5093884

fix tests and documentation

d34e772

move anndata constructor functions

9efa92e

don't generate Rd for helper functions

ad54992

fix print method

6310774

add obs_names and var_names

b052230

fix spaces

d524616

rcannood merged commit 1fd62d5 into main Dec 5, 2023
7 checks passed

rcannood deleted the export_dataset_generator branch December 5, 2023 15:28

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

export `generate_dataset()` function #152

export `generate_dataset()` function #152

rcannood commented Dec 4, 2023 •

edited

Loading

lazappi left a comment

lazappi Dec 5, 2023

rcannood Dec 5, 2023

rcannood Dec 5, 2023

rcannood Dec 5, 2023

lazappi Dec 5, 2023

rcannood Dec 5, 2023

lazappi Dec 5, 2023

rcannood Dec 5, 2023

lazappi Dec 5, 2023

rcannood Dec 5, 2023

export generate_dataset() function #152

export generate_dataset() function #152

Conversation

rcannood commented Dec 4, 2023 • edited Loading

lazappi left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

export `generate_dataset()` function #152

export `generate_dataset()` function #152

rcannood commented Dec 4, 2023 •

edited

Loading