Thanks! In the meantime I did this and seem to have worked. Although with larger sets I guess the sapply would be faster

df$ensembl_gene_id <- gsub('\\..+$', '', df$ensembl_gene_id)

library(biomaRt)
mart <- useDataset("hsapiens_gene_ensembl", useMart("ensembl"))
genes <- df$ensembl_gene_id
symbol <- getBM(filters = "ensembl_gene_id",
                attributes = c("ensembl_gene_id","hgnc_symbol"),
                values = genes, 
                mart = mart)
df <- merge(x = symbol, 
              y = df, 
              by.x="ensembl_gene_id",
              by.y="ensembl_gene_id")

Yes, you will likely need to trim off the version before merging with data from biomart, which will lack those. I don't know if it's the "most efficient" method, but at least a convenient and "good enough" one is:

> foo = as.character(c("ENSG00000000003.14", "ENSG00000000457.13", "ENSG00000000005.5"))
> fixed_names = sapply(strsplit(foo, ".", fixed=T), function(x) x[1])

You can replace foo with the row.names() accessor.