rlistdataframedata-manipulationbigrquery

How to flatten dataframe (loaded from BigQuery) with multiple list-columns in R


I am using R's bigrquery library to load data from a BQ database into R. bigrquery works fine, and a short example of the output I am receiving is as such:

dput(my_df_from_bq)
structure(list(season = c("2019", "2019", "2017", "2018", "2018"
), o_or_d = c("Offense", "Defense", "Offense", "Offense", "Defense"
), chances = list(list(num_ato_chances = 6L, ato_pts_scored = 4L, 
    ato_ppp = 0.66667, num_ato_chances_pctile = 0.272955974842767, 
    ato_ppp_pctile = 0.335849056603774), list(num_ato_chances = 7L, 
    ato_pts_scored = 2L, ato_ppp = 0.28571, num_ato_chances_pctile = 0.534591194968553, 
    ato_ppp_pctile = 0.913207547169811), list(num_ato_chances = 5L, 
    ato_pts_scored = 2L, ato_ppp = 0.4, num_ato_chances_pctile = 0.147118921127912, 
    ato_ppp_pctile = 0.177768696362893), list(num_ato_chances = 1L, 
    ato_pts_scored = 0L, ato_ppp = 0, num_ato_chances_pctile = 0, 
    ato_ppp_pctile = 0), list(num_ato_chances = 6L, ato_pts_scored = 8L, 
    ato_ppp = 1.33333, num_ato_chances_pctile = 0.70093839249286, 
    ato_ppp_pctile = 0.165646674826601)), dribbles = list(list(
    dribbles = 928L, dribbles_pctile = 0.437735849056604), list(
    dribbles = 1040L, dribbles_pctile = 0.113207547169811), list(
    dribbles = 771L, dribbles_pctile = 0.0469963220269718), list(
    dribbles = 735L, dribbles_pctile = 0.00489596083231334), 
    list(dribbles = 1049L, dribbles_pctile = 0.103223174214606))), row.names = c(NA, 
-5L), class = c("tbl_df", "tbl", "data.frame")) 

> my_df_from_bq
# A tibble: 5 x 4
  season o_or_d  chances          dribbles        
  <chr>  <chr>   <list>           <list>          
1 2019   Offense <named list [5]> <named list [2]>
2 2019   Defense <named list [5]> <named list [2]>
3 2017   Offense <named list [5]> <named list [2]>
4 2018   Offense <named list [5]> <named list [2]>
5 2018   Defense <named list [5]> <named list [2]>

The table I am loading from BQ contains many nested structs, and as a result, this structure for the dataframe is as expected, as the bigrquery docs themselves indicate that Nested values become list-columns containing named lists.

However, I would like to now flatten this out. You'll notice that the list my_df_from_bq$chances[[1]]$ has values including num_ato_chances, ato_pts_scored, ato_ppp, and more. I would therefore like to flatten out this dataframe, such that the column names are:

...where the list name is concatenated with the values within each list.


Solution

  • You can use unnest_wider but I think it does not unnest multiple columns in one go as referenced in this open Github issue.

    library(tidyr)
    my_df_from_bq %>%
       unnest_wider(chances, names_sep = "_") %>%
       unnest_wider(dribbles, names_sep = "_")
    
    #  season o_or_d chances_num_ato… chances_ato_pts… chances_ato_ppp chances_num_ato…
    #  <chr>  <chr>             <int>            <int>           <dbl>            <dbl>
    #1 2019   Offen…                6                4           0.667            0.273
    #2 2019   Defen…                7                2           0.286            0.535
    #3 2017   Offen…                5                2           0.4              0.147
    #4 2018   Offen…                1                0           0                0    
    #5 2018   Defen…                6                8           1.33             0.701
    # … with 3 more variables: chances_ato_ppp_pctile <dbl>, dribbles_dribbles <int>,
    #   dribbles_dribbles_pctile <dbl>