Merging multiple data.tables

Question

I am aware that one can merge (join) two data.table with the merge function or the [.data.table function. However, if I have say 10, data.tables and want to use do.call to merge them all together, is there a function that would do this? Currently I resort to do.call(cbind, ...) which works for only very special cases.

score 62 · Accepted Answer · answered Nov 07 '12 at 16:47

62

Not sure, but maybe (untested) :

Reduce(merge,list(DT1,DT2,DT3,...))

answered Nov 07 '12 at 16:47

Matt Dowle

57,542
22
163
221

3

The reduce solution works nice, but I don't know how to specify all=TRUE. Any idea? – sdaza Feb 26 '13 at 19:15
5

@sdaza Hm good question. All I can think currently is to create a wrapper like `mymerge = function(x,y) merge.data.table(x,y,all=TRUE)` and pass that wrapper to `Reduce` instead, i.e. `Reduce(mymerge,list(DT1,DT2,DT3,...))`. – Matt Dowle Feb 26 '13 at 19:35
1

[the all = T](https://stackoverflow.com/questions/8091303/simultaneously-merge-multiple-data-frames-in-a-list) worked for me in data.table. – Henk Feb 18 '16 at 14:59
4

If your `data.table`s are unkeyed, the wrapper should look like `function(x, y) x[y, on = "merge_column_name"]` – MichaelChirico Jan 13 '17 at 21:34

Michael Ohlrogge · Answer 2 · 2017-07-29T18:51:58.263

36

To flesh out the solution suggested by @Henk in comments, here is how to use this to select the all = TRUE option in a concise formulation:

MergedDT = Reduce(function(...) merge(..., all = TRUE), List_of_DTs)

edited Jul 29 '17 at 18:51

answered Sep 18 '16 at 18:14

Michael Ohlrogge

9,801
5
45
73

score 7 · Answer 3 · answered Jun 22 '17 at 07:27

7

The latest version of data.table (1.10.5) has an easy syntax for table join, which can also be chained.

 set(DT1, key)
 set(DT2, key) 
  ...
 set(DT10, key)

 DT_cmb <- DT1[DT2,][DT3,]...[DT10,]

answered Jun 22 '17 at 07:27

HappyCoding

4,869
7
28
47

1

This solution seems good, but it can be exhaustive if you have several `data.table`'s. Probably, there is an easy (and more generic) way to do it using `lapply` and `ls(pattern = '^DT')` instead. – lcgodoy Feb 07 '19 at 00:57
@lcgodoy Yeah, I like to use `lapply()` along with `setkeyv( )` in these situations, especially since I'll already want the `data.table`s in a list to use my preferred [merge syntax](https://stackoverflow.com/a/39561045/3541976) – Michael Ohlrogge Feb 07 '19 at 21:22
this syntax is there for ages, and chaining works for ages as well, the problem is that it does not scale, as you might have a list of data.table that varies in quantity – jangorecki Apr 10 '20 at 08:53

score 0 · Answer 4 · answered Sep 07 '21 at 06:51

To fresh answer from Michael Ohlrogge, here I create a function to do this task. You can inherit more arguments from merge.data.table()

mergeDTs <- function(dt_list, by = NULL, sort = FALSE) {
  Reduce(
    function(...) {
      merge(..., by = by, all = TRUE, sort = sort)
    }, dt_list)
}

Merging multiple data.tables

4 Answers4

Linked

Related