CODEWITHSUNDEEP
rdatedata.tabletidyversebetween
JFG123
JFG123

Reputation: 597

Check date in one dataframe is in date range in another dataframe and return rows when true

if (!require("pacman")) install.packages("pacman")
pacman::p_load(tidyverse, lubridate)

# Example of sample dates - these are to be used to cross check if date exists within the range
Sample.Dates = tibble(
  ID = "ID",
  Round = 1:3,
  Start.Date = dmy(c("03/12/2018","10/12/2018","17/12/2018")),
  End.Date = dmy(c("09/12/2018","16/12/2018","23/12/2018")))

# Reference dates for a particular player - "John". Need to cross check the date against Sample.Dates and attach round, start and end date columns
Ref.Dates = tibble(
  ID= "ID",
  Date = seq.Date(ymd("2018-12-05"), ymd("2018-12-31") ,  by = "day"),
  Player = "John",
  Rows = row_number(Date))


# Function for checking if date exists within range and then returns the round, start and end date values

Dates.Check.YN.Func = function(x){

  Date = x %>% pull(Date)

  Cross.Check = Sample.Dates %>% rowwise()%>%
    dplyr::mutate(Match = ifelse(between(Date, Start.Date, End.Date),1,0))%>%
    filter(Match == 1)%>%
    ungroup()%>%
    select(-Match)

  left_join(x, Cross.Check, by = "ID")

}

# Applying function to each row/date using nest()
Data.Nest = Ref.Dates %>%
  nest(-Rows)%>%
  mutate(out = map(data,Dates.Check.YN.Func)) %>%
  unnest(out) %>%
  select(-data)

Now this code works with no problems. However this is just a dummy data set and in actual fact I want to cross check over 100,000 dates. When doing this with my real data set this takes ~30mins. Searching to see if anyone can see a way of speeding up my code using a tidyverse solution (preferred) or other means.

Upvotes: 2

Views: 1096

Answers (2)

Uwe
Uwe

Reputation: 42592

As of version v1.9.8 (on CRAN 25 Nov 2016), data.table has gained the ability to perform non-equi joins.

Here, a non-equi update join is used to append the columns Round, Start.Date, and End.Date from Sample.Dates to Ref.Dates. Ref.Dates is updated by reference, i.e., without copying the whole data object. 

library(data.table)
# coerce to data.table class
setDT(Ref.Dates)[
  # perform update join 
  setDT(Sample.Dates), on = .(ID, Date >= Start.Date, Date <= End.Date), 
  `:=`(Round = Round, Start.Date = Start.Date, End.Date = End.Date)]
Ref.Dates

    ID       Date Player Rows Round Start.Date   End.Date
 1: ID 2018-12-05   John    1     1 2018-12-03 2018-12-09
 2: ID 2018-12-06   John    2     1 2018-12-03 2018-12-09
 3: ID 2018-12-07   John    3     1 2018-12-03 2018-12-09
 4: ID 2018-12-08   John    4     1 2018-12-03 2018-12-09
 5: ID 2018-12-09   John    5     1 2018-12-03 2018-12-09
 6: ID 2018-12-10   John    6     2 2018-12-10 2018-12-16
 7: ID 2018-12-11   John    7     2 2018-12-10 2018-12-16
 8: ID 2018-12-12   John    8     2 2018-12-10 2018-12-16
 9: ID 2018-12-13   John    9     2 2018-12-10 2018-12-16
10: ID 2018-12-14   John   10     2 2018-12-10 2018-12-16
11: ID 2018-12-15   John   11     2 2018-12-10 2018-12-16
12: ID 2018-12-16   John   12     2 2018-12-10 2018-12-16
13: ID 2018-12-17   John   13     3 2018-12-17 2018-12-23
14: ID 2018-12-18   John   14     3 2018-12-17 2018-12-23
15: ID 2018-12-19   John   15     3 2018-12-17 2018-12-23
16: ID 2018-12-20   John   16     3 2018-12-17 2018-12-23
17: ID 2018-12-21   John   17     3 2018-12-17 2018-12-23
18: ID 2018-12-22   John   18     3 2018-12-17 2018-12-23
19: ID 2018-12-23   John   19     3 2018-12-17 2018-12-23
20: ID 2018-12-24   John   20    NA       <NA>       <NA>
21: ID 2018-12-25   John   21    NA       <NA>       <NA>
22: ID 2018-12-26   John   22    NA       <NA>       <NA>
23: ID 2018-12-27   John   23    NA       <NA>       <NA>
24: ID 2018-12-28   John   24    NA       <NA>       <NA>
25: ID 2018-12-29   John   25    NA       <NA>       <NA>
26: ID 2018-12-30   John   26    NA       <NA>       <NA>
27: ID 2018-12-31   John   27    NA       <NA>       <NA>
    ID       Date Player Rows Round Start.Date   End.Date

Upvotes: 5

xiaodai
xiaodai

Reputation: 16074

You can use data.table::foverlaps which is designed for these types of analysis.

library(data.table)
library(dtplyr) # allows you to use dplyr with data.table backend

# make Ref.Dates into a data.table 
setDT(Ref.Dates)

Ref.Dates[,Date_copy := copy(Date)]
# or dplyr syntax if you prefer
# Ref.Dates = Ref.Dates %>% 
#   mutate(Date_copy = copy(Date))

# you must make Sample.Dates into a data.table and index by the join keys
setDT(Sample.Dates)
setkey(Sample.Dates, ID, Start.Date, End.Date)

# fast overlaps
Data.Nest = foverlaps(Ref.Dates, Sample.Dates,
                      by.x = c("ID", "Date", "Date_copy"),
                      by.y = c("ID", "Start.Date", "End.Date"))

# remove the Date_copy column
Data.Nest[,Date_copy := NULL]

Upvotes: 2

Related Questions