Using Reduce to merge multiple data frames with passing arguments and without defining function outside the Reduce (syntax)

Question

I've a a list with a number of data set that corresponds to the object generated below:

data("AirPassengers"); data("mtcars")
lstDta <- list(dtaCars1 = mtcars, dtaCars2 = mtcars,
               dtaCars3 = mtcars, dtaOtherStuff = AirPassengers)

I would like to merge the cars data sets on row.names

Results

The results should correspond to the data frame:

res <- merge(
    x = merge(x = lstDta$dtaCars1, y = lstDta$dtaCars2, by = "row.names"),
    y = lstDta$dtaCars3, by.y = "row.names",
    by.x = "Row.names")

where the columns are joined using row.names (ideally, I would drop the Row.names variable but this doesn't bother me):

> dim(res)
[1] 32 34 

Problem

I want to achieve the same results making use of Reduce, in particular I am interested in:

• Merging the data frames on the row.names
• Filtering the list. For example, I want to merge the cars data only and ignore the other data set

Additional requirements

Very useful answer suggests defining a function outside reduce, on the lines of the code:

merge.all <- function(x, y) {
    merge(x, y, all=TRUE, by="Sample")
}

output <- Reduce(merge.all, DataList)

I would like to avoid defining the function outside the Reduce syntax.

Attempt

As shown in the attempt below, I would like to cram everything inside the Reduce:

dtaMrgd <- Reduce(f = function(x,y) {merge(x,y, by = "row.names")},
              lapply(lstDta[grepl("Cars", names(lstDta)) == TRUE]))

so the Reduce does two things:

1. Filters the passed list using string according to matching names
2. Uses the filtered object to apply merge function with the desired characteristics

Needles to say, the code above fails.

---

Notes

I'm specifically interested in a solution that would be of format res <- Reduce( ... ). I'm not interested in creating some additional objects/functions outside the Reduce().

Answer 1

I think this is one way to achieve what you want:

res2 <- Reduce(function(x, y) {
  data.frame(merge(x, y, by = 0), row.names = row.names(x))[,-1]
}, lstDta[grep("Cars", names(lstDta))])

dim(res2)
#[1] 32 33

names(res2)
#[1] "mpg.x"  "cyl.x"  "disp.x" "hp.x"   "drat.x" "wt.x"   "qsec.x" "vs.x"   "am.x"   "gear.x" "carb.x"
#[12] "mpg.y"  "cyl.y"  "disp.y" "hp.y"   "drat.y" "wt.y"   "qsec.y" "vs.y"   "am.y"   "gear.y" "carb.y"
#[23] "mpg"    "cyl"    "disp"   "hp"     "drat"   "wt"     "qsec"   "vs"     "am"     "gear"   "carb"

---

• To filter the input list, I'm using lstDta[grep("Cars", names(lstDta))]
• I drop the first column of each merged result (Row.names) with [,-1]
• You can use by = 0 as a synonym for row.names to avoid any problems trying to merge on Row.names and row.names
• Without the explicit data.frame(..., row.names = row.names(x)), merge will drop the original mtcars row names and replace them with the default 1:nrow(x). This would be problematic for ensuing merge calls.

---

From the helpfile ?merge,

Columns to merge on can be specified by name, number or by a logical vector: the name "row.names" or the number 0 specifies the row names. If specified by name it must correspond uniquely to a named column in the input.