CODEWITHSUNDEEP
rdplyrmergematch
user3570187
user3570187

Reputation: 1773

partial matching of strings in different two datasets to obtain a match with higher frequency

I have strings in two datasets and i would like to do a partial match. Here is the code that I have written

 df1 <- data.frame(A=c(.87,.11,.44,.45), B=c("I have a beard", "I slept for two hours", "I have had two courses","this is not true"))

 df2 <- data.frame(X=c(127,10,433,344,890,4),Y=c("have","beard","syllabus","true","three","maths"))

I want to do a pmatch and I am expecting output as follows 

  A     B                            X      Y
.87   I have a beard               127      have
.11   I slept for two hours        NA       NA
.44   I have had two courses       127      have
.45   this is not true             344      true

I would like to a partial match with a left join on df1. I want to get the higher of the two matches(for example in "I have a beard" string "have" match has 127 and "beard" has 10 and i want to get the higher match. Any suggestions?

Upvotes: 1

Views: 778

Answers (2)

AntoniosK
AntoniosK

Reputation: 16121

This dplyr method doesn't need a join (which is reasonable as you don't have a common column to join on). It combines the 2 datasets and finds the matches. As long as you don't have thousands of rows it will work fast enough. Of course you can make the script smaller, but you can run this step by step to see how it works.

df1<- data.frame(A=c(.87,.11,.44,.45), B=c("I have a beard", "I slept for two hours", "I have had two courses","this is not true"))

df2<- data.frame(X=c(127,10,433,344,890,4),Y=c("have","beard","syllabus","true","three","maths"))

library(dplyr)

df1 %>% 
  rowwise() %>%
  do(data.frame(.,df2)) %>%                    # combine datasets
  do(data.frame(.,flag = grepl(.$Y,.$B))) %>%  # for each row check if there's a match and name it flag
  ungroup %>%
  group_by(A,B) %>%                            # for each A and B
  mutate(N=sum(flag)) %>%                      # count how many matches you have
  filter(flag==TRUE | N == 0) %>%              # keep only A,B where you have some matches or no match at all
  top_n(1,X) %>%                               # pick one row based on max value of X
  ungroup %>%
  mutate(Y = ifelse(flag==FALSE,NA,as.character(Y)),   # if there's no match replace Y with NA
         X = ifelse(flag==FALSE,NA,X)) %>%             # if there's no match replace X with NA
  select(-c(flag,N)) 


#      A                      B   X    Y
# 1 0.87         I have a beard 127 have
# 2 0.11  I slept for two hours  NA   NA
# 3 0.44 I have had two courses 127 have
# 4 0.45       this is not true 344 true

Try to experiment and change various column values to see how it works. You might be able to spot any bugs in advance.

Upvotes: 1

Steven Beaupr&#233;
Steven Beaupr&#233;

Reputation: 21621

Another option: 

library(dplyr)

df1 %>% 
  mutate(X = sapply(strsplit(as.character(B), ' '), 
                    function(x) with(df2, max(X[Y %in% x])))) %>%
  left_join(., df2)

Which gives:

#Joining by: "X"
#     A                      B    X    Y
#1 0.87         I have a beard  127 have
#2 0.11  I slept for two hours -Inf <NA>
#3 0.44 I have had two courses  127 have
#4 0.45       this is not true  344 true

Upvotes: 1

Related Questions