Inverse matching with stringr - r

I want to remove all characters which doesn't match a string pattern using stringr package. So far I've been able to remove those before the pattern using "\\w+(?= (grape|satsuma))" as pattern but remove those after the pattern is still imposible.
> str_remove_all("apples grape banana melon olive persimon grape apples satsuma papaya",
+ "\\w+(?= (grape|satsuma))")
[1] " grape banana melon olive grape satsuma papaya"
The desired result is:
"grape grape satsuma"
(NOTE: I am aware the easiest approach in this case is to extract only "grape" and "satsuma" but for analysis purposes I prefer this way)
Edited providing the entire problem
The entire problem is as follow, given a d data frame which contains a column with a string the function should return the same column only with matches:
> d
# A tibble: 2 x 2
string_column c2
<chr> <dbl>
1 apples grape banana satsuma 3
2 grape banana satsuma melon 4
Using the answer provided by #d.r works:
> d %>%
+ mutate_at(vars(string_column), ~ gsub("(grape|satsuma| )(*SKIP)(*FAIL)|.", "", ., perl = TRUE))
# A tibble: 2 x 2
string_column c2
<chr> <dbl>
1 " grape satsuma" 3
2 "grape satsuma " 4
All answers provided so far using stringr package fail returning the string_column
This the dput for d:
d <- structure(list(string_column = c("apples grape banana satsuma",
"grape banana satsuma melon"), c2 = c(3, 4)), row.names = c(NA,
-2L), class = c("tbl_df", "tbl", "data.frame"))

You may want to look at negative lookaheads and some related regex techniques in the linked thread.
However, since we are extracting words I'd rather use str_extract_all and I'd do it like this:
str_extract_all("apples grape banana melon olive persimon grape apples satsuma papaya",
"grape|satsuma")
"grape" "grape" "satsuma"
I also really like this line that #steveLangsford left in a comment:
paste0(unlist(str_extract_all("apples grape banana melon olive persimon grape apples satsuma papaya", "grape|satsuma")), collapse=" ")
"grape grape satsuma"
Taking it a little bit further based on our discussion/comments:
string_column <- c("apples grape banana satsuma", "grape banana satsuma melon")
c2 <- c(3, 4)
d <- tibble(string_column,c2)
myfun <- function(x) {paste0(unlist(str_extract_all(x, "grape|satsuma")), collapse=" ") }
sapply(d$string_column, myfun)
"grape satsuma" "grape satsuma"

Related

Splitting a string into multiple columns (with specific order)

I receive data that isn't in a very nice format (and I can't change them upstream). There is one column that needs to be reordered and split into 10+ other columns based on certain keywords.
Here's an example of data I receive - for each person, they have chosen one 3 different foods. Their choices for each food category (food1, food2, food3) come right after the text:
list1 <- c(' food1 pasta food2 apple food3 carrot ')
list2 <- c(' food2 banana food3 cucumber food1 brown rice ')
list3 <- c(' food3 bell pepper food2 plum food1 bread ')
foodListDF <- as.data.frame(matrix(c(1,2,3, list1, list2, list3), nrow = 3), stringsAsFactors = FALSE)
colnames(foodListDF) <- c('Person', 'Choices')
foodListDF
Person Choices
1 1 food1 pasta food2 apple food3 carrot
2 2 food2 banana food3 cucumber food1 brown rice
3 3 food3 bell pepper food2 plum food1 bread
The above is the format I receive my data in. My end goal is to split the Choices column into 3 separate columns labeled food1, food2, and food3 which requires things to be ordered properly:
Person food1 food2 food3
1 1 pasta apple carrot
2 2 brown rice banana cucumber
3 3 bread plum bell pepper
I know that I can split the choices doing something like this:
library(stringr)
as.data.frame(str_split_fixed(foodListDF$Choices, c(' food1 | food2 | food3 '), 4))[,2:4]
V2 V3 V4
1 pasta apple carrot
2 banana cucumber brown rice
3 bell pepper plum bread
But this obviously doesn't split them into their proper groups/order which is very necessary.
I'm really just struggling to think how to extract the correct food from the proper group for each person. Any ideas?
You could extract the food number and food items separately (t1 and t2), join them together, unnest the data and get it into wide format.
library(dplyr)
library(tidyr)
foodListDF %>%
mutate(food = stringr::str_extract_all(Choices, 'food\\d+')) %>%
select(-Choices) -> t1
foodListDF %>%
separate_rows(Choices, sep = 'food\\d+') %>%
filter(Choices != ' ') %>%
mutate(Choices = trimws(Choices)) %>%
group_by(Person) %>%
summarise(col = list(Choices)) -> t2
inner_join(t1, t2, by = 'Person') %>%
unnest(c(food, col)) %>%
pivot_wider(names_from = food, values_from = col)
# Person food1 food2 food3
# <chr> <chr> <chr> <chr>
#1 1 pasta apple carrot
#2 2 brown rice banana cucumber
#3 3 bread plum bell pepper
Base R
Here are two base R approaches, both involving regmatches and gregexpr.
The first makes use of unstack. It results in a data.frame.
splitfun1 <- function(string) {
mat <- gregexpr("food\\d+ ", string)
unstack(
list(l1 = unlist(lapply(regmatches(string, mat), trimws), use.names = FALSE),
l2 = unlist(lapply(regmatches(string, mat, invert = TRUE),
function(x) trimws(x[-1])), use.names = FALSE)),
l2 ~ l1)
}
splitfun1(foodListDF$Choices)
# food1 food2 food3
# 1 pasta apple carrot
# 2 brown rice banana cucumber
# 3 bread plum bell pepper
The second makes use of matrix indexing to fill in an empty matrix. It's probably a bit more efficient than the first alternative. It results in a matrix.
splitfun2 <- function(string) {
mat <- gregexpr("food\\d+ ", string)
l1 <- lapply(regmatches(string, mat), trimws)
l2 <- lapply(regmatches(string, mat, invert = TRUE),
function(x) trimws(x[-1]))
ul <- unlist(l1, use.names = FALSE)
cn <- sort(unique(ul))
out <- matrix(NA_character_, nrow = length(string), ncol = length(cn),
dimnames = list(seq_along(string), cn))
out[cbind(rep(seq_along(string), lengths(l1)), ul)] <- unlist(l2, use.names = FALSE)
out
}
splitfun2(foodListDF$Choices)
# food1 food2 food3
# 1 "pasta" "apple" "carrot"
# 2 "brown rice" "banana" "cucumber"
# 3 "bread" "plum" "bell pepper"
Of course, with either of these, you would then need to cbind the result with the relevant columns from the source data.frame.
cbind(foodListDF[1], splitfun2(foodListDF$Choices))
splitstackshape + data.table
Another option is to use cSplit from my "splitstackshape" package along with some pretty straightforward gsub work, followed by dcast to go into a wide form.
library(splitstackshape)
# library(data.table) # if required
# Basic helper function
fun <- function(string) {
list(gsub("(food\\d+) (.*)", "\\1", string),
gsub("(food\\d+) (.*)", "\\2", string))
}
cSplit(as.data.table(foodListDF)[, Choices := gsub(" food", ",food", trimws(Choices))],
"Choices", ",", "long")[, fun(Choices), Person][, dcast(.SD, Person ~ V1, value.var = "V2")]
# Person food1 food2 food3
# 1: 1 pasta apple carrot
# 2: 2 brown rice banana cucumber
# 3: 3 bread plum bell pepper
dplyr + tidyr
Adapting the above to "dplyr" + "tidyr", you can try:
library(dplyr)
library(tidyr)
foodListDF %>%
mutate(Choices = gsub(" food", ",food", trimws(Choices))) %>%
separate_rows(Choices, sep = ",") %>%
separate(Choices, c("var", "val"), extra = "merge") %>%
pivot_wider(names_from = var, values_from = val)
# # A tibble: 3 x 4
# Person food1 food2 food3
# <chr> <chr> <chr> <chr>
# 1 1 pasta apple carrot
# 2 2 brown rice banana cucumber
# 3 3 bread plum bell pepper
In one convoluted Base R expression:
data.frame(cbind(Person = foodListDF$Person,
do.call("rbind", Map(function(x){y <- setNames(x[[2]], x[[1]]); y[order(x[[1]])]},
lapply(strsplit(foodListDF$Choices, "\\s+"), function(x) {
res <- data.frame(t(grep("food\\d+", x, value = TRUE)), stringsAsFactors = FALSE)
res2 <- unlist(strsplit(gsub("^&&\\s*", "",
paste0(Filter(function(y){y != ""}, Vectorize(gsub)("food\\d+", "&&", x)),
collapse = " ")), "\\s*&&\\s*"))
list(res, res2)
}
)
)
)
), stringsAsFactors = FALSE)
You can use food as a delimiter in strsplit, sort the result, remove the first character with substring and return the result to your dataset.
foodListDF[paste0("food",1:3)] <- t(sapply(strsplit(foodListDF$Choices, "food"),
function (x) trimws(substring(sort(x[-1]), 2))))
foodListDF[-2]
# Person food1 food2 food3
#1 1 pasta apple carrot
#2 2 brown rice banana cucumber
#3 3 bread plum bell pepper
Or in case there are not all the time all levels present:
j <- sort(unique(unlist(regmatches(foodListDF$Choices, gregexpr("food\\d+",
foodListDF$Choices)))))
k <- sub("food", "", j)
foodListDF[j] <- t(sapply(strsplit(foodListDF$Choices, "food"), function(x)
trimws(sub("^\\d+", "", x[charmatch(k, x)]))))
foodListDF[-2]
# Person food1 food2 food3
#1 1 pasta apple carrot
#2 2 brown rice banana cucumber
#3 3 bread plum bell pepper

Group columns and store variables in list within dataset in R [duplicate]

This question already has an answer here:
Group by columns and summarize a column into a list
(1 answer)
Closed 2 years ago.
So basically, I have a dataframe like this
fruit vendor size
banana Walmart M
banana Sears L
apple Popeye's XL
orange Footlocker S
apple Popeye's W
banana Walmart L
And I need it to look like this (I tried a group_by but idk how to group sizes as a list in a row.
fruit vendor size
banana Walmart c("M", "L")
banana Sears L
apple Popeye's c("XL","W")
orange Footlocker S
#Tried this
df %>%
group_by(fruit, vendor) %>%
##now what?
Then later on I would like to choose from the list on an ifelse.
inc_list <- c("XS","S")
minc_list <- c("M","L", "W", "XL")
df$counter <- ifelse(unlist(df$size) %in% inc_list , 1, 0)
Doesnt work but I want it to look like. So :
if appears in inc_list then count in counter1 how many there are. Ditto for minc_list and counter2, counts how many of the ones in there are in that list.
fruit vendor size counter1 counter2
banana Walmart c("M", "L") 1 1
banana Sears L 0 1
apple Popeye's c("XL","W") 0 2
orange Footlocker S 1 0
EDIT: Last bit, c("S","S") would only be 1, duplicates from the same list shouldn't count.
You can combine size into a list in summarise :
library(dplyr)
df %>% group_by(fruit, vendor) %>%summarise(size = list(size))
# fruit vendor size
# <chr> <chr> <list>
#1 apple Popeyes <chr [2]>
#2 banana Sears <chr [1]>
#3 banana Walmart <chr [2]>
#4 orange Footlocker <chr [1]>
You can also do this in base R :
aggregate(size~fruit+vendor, df, list)
Or data.table :
library(data.table)
setDT(df)[, .(size = list(size)), .(fruit, vendor)]
data
df <- structure(list(fruit = c("banana", "banana", "apple", "orange",
"apple", "banana"), vendor = c("Walmart", "Sears", "Popeyes",
"Footlocker", "Popeyes", "Walmart"), size = c("M", "L", "XL",
"S", "W", "L")), class = "data.frame", row.names = c(NA, -6L))
We can also convert to string with toString
aggregate(size~fruit+vendor, df, toString)
# fruit vendor size
#1 orange Footlocker S
#2 apple Popeyes XL, W
#3 banana Sears L
#4 banana Walmart M, L

Create variables based on regular expressions with a loop in r

I need help to create variables based on regular expressions.
This is my dataframe:
df <- data.frame(a=c("blue", "red", "yellow", "yellow", "yellow", "yellow", "red"), b=c("apple", "orange", "peach", "lemon", "pineapple", "tomato", NA))
Basically, what I want to do is this, but in one step:
regx_1 <- as.numeric(grep("^[a-z]{5}$", df$b))
regx_2 <- as.numeric(grep("^[a-z]{6,}$", df$b))
df$fruit_1 <- NA
df$fruit_1[regx_1 + 1] <- as.character(df$b[regx_1])
df$fruit_2 <- NA
df$fruit_2[regx_2 + 1] <- as.character(df$b[regx_2])
Here is my try:
regex1 <- "^[a-z]{5}$"
regex2 <- "^[a-z]{6,}$"
regex <- c(regex1, regex1)
make_non_matches_NA <- function(vec, pattern){
df[[newvariable]] <- NA
df[[newvariable]][as.numeric(grep(pattern, vec)) + 1] <- as.character(vec[as.numeric(grep(pattern, vec))])
return(newvariable)
}
df[c("fruit1", "fruit2")] <- lapply(regex, make_non_matches_NA, vec = df$b)
EDIT: Why is my approach wrong? (Please note that the actual problem is bigger, so I have to stick to an approach, where a repetition of a pattern should be avoided)
Any help is much appreciated!
Having numbered items in a your workspace is a good sign that they really belong
to a list, so they are formally linked and we can work with them much more easily. So let's do that first.
regex <- c("^[a-z]{5}$", "^[a-z]{6,}$")
Our core functionality is to copy a source vector, but remove elements that don't match, and leave NA in their place, so we'll make a function for that, and we'll name it explicitly so we understand intuitively what it's doing (and as will our colleagues next reader on SO ;) ) :
make_non_matches_NA <- function(vec, pattern){
# logical indices of matches
matches_lgl <- grepl(pattern, vec)
# the elements which don't match should be NA
vec[!matches_lgl] <- NA
# resulting vector should be returned
vec
}
Let's test this with first pattern
make_non_matches_NA(df$b, regex[[1]])
#> [1] apple <NA> peach lemon <NA> <NA>
#> Levels: apple lemon orange peach pineapple tomato
So far so good! now let's test it with all regex, we avoid for loops when we can generally in R because we have clearer tools like lapply(). Here I want to apply this function to all regex expressions :
lapply(regex, make_non_matches_NA, vec = df$b)
#> [[1]]
#> [1] apple <NA> peach lemon <NA> <NA>
#> Levels: apple lemon orange peach pineapple tomato
#>
#> [[2]]
#> [1] <NA> orange <NA> <NA> pineapple tomato
#> Levels: apple lemon orange peach pineapple tomato
Great, it works!
But I want this in my data.frame, not as a separate list, so I will assign this result to the relevant names in my df directly
df[c("fruit1", "fruit2")] <- lapply(regex, make_non_matches_NA, vec = df$b)
# then print my updated df
df
#> a b fruit1 fruit2
#> 1 1 apple apple <NA>
#> 2 2 orange <NA> orange
#> 3 3 peach peach <NA>
#> 4 4 lemon lemon <NA>
#> 5 5 pineapple <NA> pineapple
#> 6 6 tomato <NA> tomato
tada!
I don't if this qualifies as "at one step" but you could try mutate from the dplyr package:
df <- data.frame(a=c(1:6), b=c("apple", "orange", "peach", "lemon", "pineapple", "tomato"),
stringsAsFactors = FALSE)
Note that I set stringsAsFactors = FALSE inside data.frames.
dplyr::mutate(df, fruit_1 = if_else(grepl("^[a-z]{5}$", b), b, NA_character_),
fruit_2 = if_else(grepl("^[a-z]{6}$", b), b, NA_character_))
a b fruit_1 fruit_2
1 1 apple apple <NA>
2 2 orange <NA> orange
3 3 peach peach <NA>
4 4 lemon lemon <NA>
5 5 pineapple <NA> <NA>
6 6 tomato <NA> tomato

Spreading user reported variables to multiple columns

I am trying to seperate a couple of user filled variables into multiple columns. I have tried to use the spread function, but I am running into some problems. For example, the database looks like this:
SubjID Input1 Input2
1 Banana NA
2 Apple NA
3 NA Banana
4 Apple Banana
And I am trying to get it to look like this:
SubjID Input1 Input2 Banana Apple
1 Banana NA Banana NA
2 Apple NA NA Apple
3 NA Banana Banana Na
4 Apple Banana Banana Apple
I can use the spread function in tidyr to separate input 1, but the problem comes with input 2. I am able to spread it, but I cant put the values into the previously created Banana column, it will instead create two banana columns, which I cannot figure out how to merge correctly. Is there any way to have it sort into the columns correctly? I am new to R and am having a lot of trouble with this aspect of the database. There are too many options for me to discretely state banana and apple and I am really unsure of how to do this.
We may need to gather first before doing the spread
library(dplyr)
library(tidyr)
df1 %>%
gather(key, val, -SubjID, na.rm = TRUE) %>%
mutate(key1 = val) %>%
select(-key) %>% spread(key1, val) %>%
left_join(df1, ., by = 'SubjID')
# SubjID Input1 Input2 Apple Banana
#1 1 Banana <NA> <NA> Banana
#2 2 Apple <NA> Apple <NA>
#3 3 <NA> Banana <NA> Banana
#4 4 Apple Banana Apple Banana
data
df1 <- structure(list(SubjID = 1:4, Input1 = c("Banana", "Apple", NA,
"Apple"), Input2 = c(NA, NA, "Banana", "Banana")), .Names = c("SubjID",
"Input1", "Input2"), class = "data.frame", row.names = c(NA,
-4L))
Try this: Assuming that your data.frame is called dat:
dat$Banana <- ifelse(dat$Input1 == "Banana" | dat$Input2 == "Banana", "Banana", NA)
dat$Apple <- ifelse(dat$Input1 == "Apple" | dat$Input2 == "Apple", "Apple", NA)
For example, the first line checks row-by-row if either df$Input1 or df$Input2 is "Banana"; if so, it puts "Banana" in the Banana column, else it puts NA.

Apply grepl across columns by saving columns in lists?

I have a dataframe in the format mentioned below:
String Keyword
1 Apples bananas mangoes mangoes
2 Apples bananas mangoes bananas
3 Apples bananas mangoes peach
.....
Its a dataframe (50000+ rows). I'm currently manually using the ifelse statement in batches.
data$Result<- ifelse(grepl("apples",data$String,ignore.case = TRUE)==TRUE,"apples",
ifelse(grepl("bananas",data$String,ignore.case = TRUE)==TRUE,"bananas",
ifelse(grepl("mangoes",data$String,ignore.case = TRUE)==TRUE,"mangoes","unavailable")))
String Keyword Result
Apples bananas mangoes mangoes mangoes
Apples bananas mangoes bananas bananas
Apples bananas mangoes peach unavailable
Is there a way, where I could store String and Keyword in a list and then apply grepl on the entire list?
Here's a simple and efficient solution with a combination of data.table and the stringi package:
library(data.table)
library(stringi)
setDT(df)[stri_detect_fixed(String, Keyword, case_insensitive = TRUE), result := Keyword]
# String Keyword result
# 1: Apples bananas mangoes mangoes mangoes
# 2: Apples bananas mangoes bananas bananas
# 3: Apples bananas mangoes peach NA
Alternatively, a data.table-only version:
library(data.table)
setDT(df)[, result := Keyword[grep(Keyword, String, ignore.case = TRUE)], by = .(Keyword, String)]
Benchmark
Here's a benchmark on a 5e5 data set against the mapply answer. (The for loop answer haven't finished running yet):
set.seed(123)
df1 <- data.frame(String = rep('Apples bananas mangoes', 5e5),
Keyword = sample(c("mangoes", "bananas", "peach"), 5e5, replace = TRUE))
system.time(df1$result2 <- ifelse(mapply(grepl,df1$Keyword, df1$String, ignore.case = TRUE), as.character(df1$Keyword), "Unavailable"))
# user system elapsed
# 40.78 0.02 41.12
system.time(setDT(df1)[stri_detect_fixed(String, Keyword, case_insensitive = TRUE), result3 := Keyword])
# user system elapsed
# 0.52 0.01 0.53
I'm assuming this is what you want:
df <- data.frame(string=rep("Apples bananas mangoes",3), keyword=c("mangoes", "bananas", "peach"))
df$result <- ifelse(mapply(grepl,df$keyword, df$string), as.character(df$keyword), "Unavailable")
string keyword result
1 Apples bananas mangoes mangoes mangoes
2 Apples bananas mangoes bananas bananas
3 Apples bananas mangoes peach Unavailable
Update
Based on the comment, it sounds like you have a list of words that you want to check against the keyword. If that is the case, something like this might work:
#Set up toy dataset
set.seed(123)
df <- data.frame(Keyword = sample(c("mangoes", "bananas", "apples","lemons" , "peach"), 10, replace = TRUE))
df
#Choose your searchwords globally
searchwords <- c("apples", "bananas", "mangoes")
library(data.table)
library(stringi)
setDT(df)
for (x in searchwords) df[Keyword == x, result := Keyword]
df[is.na(result), result := "Unavailable"]
df
Keyword result
1: bananas bananas
2: lemons Unavailable
3: apples apples
4: peach Unavailable
5: peach Unavailable
6: mangoes mangoes
7: apples apples
8: peach Unavailable
9: apples apples
10: apples apples
Here is a version using 'dplyr' and 'stringr':
library(dplyr)
library(stringr)
df <- mutate(df, result = ifelse(str_detect(string, keyword)==TRUE,
keyword, "Unavailable"))
Here is the line I used to create the play data:
df <- data.frame(string = rep("Apples bananas mangoes", 3), keyword = c("mangoes", "bananas", "peaches"), stringsAsFactors=FALSE)
And here is the output I get:
string keyword result
1 Apples bananas mangoes mangoes mangoes
2 Apples bananas mangoes bananas bananas
3 Apples bananas mangoes peaches Unavailable

Resources