Reduce Grange and keep meta column

I am trying to reduce a Grange objcet

library(GenomicRanges)

#example data
gr <- GRanges(seqnames = c("chr1", "chr1", "chr1"),
              ranges = IRanges(c(11, 101, 101), c(20, 200, 200)),
              strand = c("+", "-", "-"),
              name = c("a1", "a2", "a2"),
              score = c(1, 2, 3))

gr
      seqnames    ranges strand |        name     score
         <Rle> <IRanges>  <Rle> | <character> <integer>
  [1]     chr1     11-20      + |          a1         1
  [2]     chr1   101-200      - |          a2         2
  [3]     chr1   101-200      - |          a2         3

reduce(gr)
      seqnames    ranges strand
         <Rle> <IRanges>  <Rle>
  [1]     chr1     11-20      +
  [2]     chr1   101-200      -

but I want keep same metacols, for example

      seqnames    ranges strand    score
         <Rle> <IRanges>  <Rle>     
  [1]     chr1     11-20      +       1
  [2]     chr1   101-200      -       2,3

Set with.revmap = TRUE, then remap mcols with collapse:

with.revmap - TRUE or FALSE. Should the mapping from output to input ranges be stored in the returned object? If yes, then it is stored as metadata column revmap of type IntegerList.

# reduce with revmap
grr <- reduce(gr, with.revmap = TRUE)

# add meta columns with collapse
mcols(grr) <- do.call(rbind,
                      lapply(mcols(grr)$revmap, function(i) {
                        data.frame(
                          name = paste(mcols(gr)$name[ i ], collapse = ","),
                          score = paste(mcols(gr)$score[ i ], collapse = ","))
                        }))

grr
#GRanges object with 2 ranges and 2 metadata columns:
#      seqnames    ranges strand |        name       score
#         <Rle> <IRanges>  <Rle> | <character> <character>
#  [1]     chr1     11-20      + |          a1           1
#  [2]     chr1   101-200      - |       a2,a2         2,3
#  -------
#  seqinfo: 1 sequence from an unspecified genome; no seqlengths