Reticulate backend

[rcannood]

This PR intends to add a reticulate backend to the package. If done correctly, it could be used to better test our own HDF5 AnnData reader, since it'll allow to create h5ad files from within R.

Status

Working:

• Constructor: X, obs, var, var_names, obs_names, layers
• Getter: X, obs, var, var_names, obs_names, layers
• Setters: X, obs, var, var_names, obs_names, layers

Example

X <- Matrix::rsparsematrix(nrow = 10, ncol = 20, density = .1)
obs <- data.frame(
  cell_type = sample(c("tcell", "bcell"), 10, replace = TRUE),
  cluster = sample.int(3, 10, replace = TRUE)
)
obs_names <- paste0("cell_", seq_len(10))
var <- data.frame(
  geneinfo = sample(c("a", "b", "c"), 20, replace = TRUE)
)
var_names <- paste0("gene", seq_len(20))
layers <- list(
  X2 = matrix(2L, nrow = 10, ncol = 20),
  X3 = matrix(2.5, nrow = 10, ncol = 20),
  X4 = Matrix::rsparsematrix(10, 20, .2)
)
ad <- ReticulateAnnData$new(
  X = X,
  obs = obs,
  var = var,
  obs_names = obs_names,
  var_names = var_names,
  layers = layers
)
ad

class: ReticulateAnnData
dim: 10 obs x 20 var
X: dgCMatrix
layers: X2 X3 X4
obs: cell_type cluster
var: geneinfo

[lazappi]

I'm not sure how I feel about this, probably mostly because it is similar to what {zellkonverter} and {anndata} already do. Maybe we need to discuss how they fit together.

[rcannood]

If this backend would be integrated into anndataR, I'd be able to start deprecating {anndata} ☺️

The ReticulateAnnData can also be used to validate HDF5AnnData and vice versa.