CODEWITHSUNDEEP
rdataframe
user4078581
user4078581

Reputation:

Filter a dataframe by values in a column of type list

I'm aware there are a lot of answered similar questions. For example: Filter data.frame rows by a logical condition . The problem is these answers don't work when the type of the column is a list.

In fact, I'm using the Yelp businesses dataset that I've loaded using the library jsonlite (flattening the result). One of the columns, the categories of the business, is a list of strings.

> typeof(business_df["categories"])
[1] "list"
> business_df[1:3, "categories"]
[[1]]
[1] "Shopping"         "Shopping Centers"

[[2]]
[1] "Food"               "Soul Food"          "Convenience Stores" "Restaurants"       

[[3]]
[1] "Food"         "Coffee & Tea"

For now, I have this horrible solution:

filterByCategory <- function(category) {
  filtered_df <- cbind(businesses_df)
  if (category != "All") {
    filtered_df[, "belongs"] <-
      apply(filtered_df["categories"], 1, function(x)
        is.element(category, x[[1]]))
    filtered_df <<- subset(filtered_df, belongs)
  }
}

As you can see, I need to access the column with the [[1]] syntax. This is why I think none of these solutions actually work:

# All rows returned 
business_df[category %in% business_df$categories]
subset(business_df, category %in% business_df$categories)
# No rows returned
business_df %>% filter(category %in% categories)

Upvotes: 3

Views: 3667

Answers (1)

Paul
Paul

Reputation: 9087

It sounds like you are trying to filter a data frame where a list column contains a specific value.

categories is a list of vectors. map_lgl will map each element (vector) of the list into a logical.

library('tidyverse')

df <- tribble(
  ~rownum, ~categories,
        1, c('a', 'b'),
        2, c('c', 'd'),
        3, c('d', 'e')
)

# All rows containing the 'd' category
df %>%
  filter(map_lgl(categories, ~'d' %in% .)) %>%
  str
#> Classes ‘tbl_df’, ‘tbl’ and 'data.frame':    2 obs. of  2 variables:
#>  $ rownum    : num  2 3
#>  $ categories:List of 2
#>   ..$ : chr  "c" "d"
#>   ..$ : chr  "d" "e"

Upvotes: 4

Related Questions