Question

I have a data frame that looks like this:

_________________id ________________text______
    1   | 7821             | "some text here"
    2   | 7821             |  "here as well"
    3   | 7821             |  "and here"
    4   | 567              |   "etcetera"
    5   | 567              |    "more text"
    6   | 231              |   "other text"

And I would like to group the texts by IDs, so I can run a clustering algorithm:

________________id___________________text______
    1   | 7821             | "some text here here as well and here"
    2   | 567              |   "etcetera more text"
    3   | 231              |   "other text"

Is there any way to do this? I am importing from a database table and I have a lot of data, so I can’t do it manually.

Answer 1

You’re actually looking for aggregate, not merge, and there should be many examples on SO demonstrating different options for aggregation. Here’s the most basic and direct approach, using the formula approach to specify which columns to aggregate.

Here’s your data in a copy-and-pasteable form

mydata <- structure(list(id = c(7821L, 7821L, 7821L, 567L, 567L, 231L), 
    text = structure(c(6L, 3L, 1L, 2L, 4L, 5L), .Label = c("and here", 
    "etcetera", "here as well", "more text", "other text", "some text here"
    ), class = "factor")), .Names = c("id", "text"), class = "data.frame", 
    row.names = c(NA, -6L))

Here’s the aggregated output.

aggregate(text ~ id, mydata, paste, collapse = " ")
#     id                                 text
# 1  231                           other text
# 2  567                   etcetera more text
# 3 7821 some text here here as well and here

Of course, there is also data.table, which has nice compact syntax (and awesome speed):

> library(data.table)
> DT <- data.table(mydata)
> DT[, paste(text, collapse = " "), by = "id"]
     id                                   V1
1: 7821 some text here here as well and here
2:  567                   etcetera more text
3:  231                           other text