Implement coercion using as()

[rcannood]

Related to: Fixes #349

Description

• Added S4 as() coercions linking AnnData implementations with SingleCellExperiment and Seurat objects
• Fix inconsistency in output_class argument values across conversion functions

When as(...) is possible, the conversion function displays a warning. If it's not possible (e.g. HDF5AnnData, because an additional path argument is needed), an error is thrown instead. In both cases, the message lets the user know which function to use instead.

Example:

library(anndataR)
dummy_anndata <- generate_dataset(format = "AnnData", example = TRUE)

as(dummy_anndata, "Seurat")
## Warning: Data is of class matrix. Coercing to dgCMatrix.
## An object of class Seurat 
## 20 features across 10 samples within 1 assay 
## Active assay: RNA (20 features, 0 variable features)
##  4 layers present: numeric_matrix, numeric_dense, numeric_csparse, X
##  3 dimensional reductions calculated: numeric_matrix, numeric_dense, numeric_csparse
## Warning messages:
## 1: In asMethod(object) :
##   Using `as()` limits control over data mapping.
## ℹ Prefer `adata$as_Seurat(...)` for fine-grained control over data mapping.
## 2: No "counts" or "data" layer found in `names(layers_mapping)`, this may lead to unexpected results when using the resulting <Seurat> object. 

as(dummy_anndata, "SingleCellExperiment")
## class: SingleCellExperiment 
## dim: 20 10 
## metadata(4): scalar_character vec_integer df_logical mat_numeric_matrix
## assays(4): numeric_matrix numeric_dense numeric_csparse X
## rownames(20): gene1 gene2 ... gene19 gene20
## rowData names(3): character integer factor
## colnames(10): cell1 cell2 ... cell9 cell10
## colData names(3): character integer factor
## reducedDimNames(3): numeric_matrix numeric_dense numeric_csparse
## mainExpName: NULL
## altExpNames(0):
## Warning message:
## In asMethod(object) : Using `as()` limits control over data mapping.
## ℹ Prefer `adata$as_SingleCellExperiment(...)` for fine-grained control over data mapping.

as(dummy_anndata, "ReticulateAnnData")
## Error in py_set_attr(x, name, value) : 
##   AttributeError: 'PyCapsule' object has no attribute 'shape'
## Run `reticulate::py_last_error()` for details.
## In addition: Warning message:
## In asMethod(object) : Using `as()` limits control over data mapping.
## ℹ Prefer `adata$as_ReticulateAnnData(...)` for fine-grained control over data mapping.

as(dummy_anndata, "HDF5AnnData")
## Error in `asMethod()`:
## ! Can't coerce <AbstractAnnData> to <HDF5AnnData> with `as()()`.
## ℹ Use `adata$as_HDF5AnnData(file = <path>)` to provide the output file.
## Run `rlang::last_trace()` to see where the error occurred.

Edit: currently checking why as(dummy_anndata, "ReticulateAnnData") fails!

Checklist

Before review

• Update and regenerate man pages
• Add/update tests
• Add/update examples in vignettes
• Pass CI checks

Before merge

• Update NEWS
• Bump devel version

[lazappi]

Few formatting suggestions but looks good

[lazappi]

Why do you need this (just curious)?

[rcannood]

It causes these arguments to be evaluated eagerly instead of lazily

[lazappi]

I figured that, I just wasn't sure why that was needed 😸

[lazappi]

For AnnData the argument is file but here it is filename. Maybe we should make that consistent?

[lazappi]

I thought the tests might be failing because the classes aren't exported but I tried that in #365 and got the same errors 😿