R：将数据框列中的空字符串替换为“0”会导致所有列值都替换为“0”

Question

I have a R dataframe called newdata which I have read in using read.csv() :

PROPDMGEXP   EVTYPE

"K"          WIND
"M"          HAIL
"H"          TORNADO
"B"          WIND
"+"          HIGH WIND
"-"          TORNADO
"?"          HURRICANE
             WIND
             TORNADO
"k"          HAIL

The blank values in the PROPDMGEXP column were blank cells in the CSV file from which the data was read. I am assuming that they are empty strings. I want to replace the values in the PROPDMGEXP column with other values, and I have used regex to do so:

newdata$PROPDMGEXP[grepl("K", newdata$PROPDMGEXP, ignore.case = TRUE)] <- "10^3"

newdata$PROPDMGEXP[grepl("H", newdata$PROPDMGEXP, ignore.case = TRUE)] <- "10^2"

newdata$PROPDMGEXP[grepl("M", newdata$PROPDMGEXP, ignore.case = TRUE)] <- "10^6"

newdata$PROPDMGEXP[grepl("B", newdata$PROPDMGEXP, ignore.case = TRUE)] <- "10^9"

newdata$PROPDMGEXP[grepl("+", newdata$PROPDMGEXP, fixed = TRUE)] <- "1"

newdata$PROPDMGEXP[grepl("-", newdata$PROPDMGEXP, fixed = TRUE)] <- "0"

newdata$PROPDMGEXP[grepl("?", newdata$PROPDMGEXP, fixed = TRUE)] <- "0"

newdata$PROPDMGEXP[grepl("", newdata$PROPDMGEXP)] <- "0"

I have checked whether the values are being replaced through subsetting the data and printing it out:

mydata <- subset(newdata, PROPDMGEXP == "10^6", select=c(EVTYPE, PROPDMGEXP))

mydata1 <- subset(newdata, PROPDMGEXP == "10^3", select=c(EVTYPE, PROPDMGEXP))

mydata2 <- subset(newdata, PROPDMGEXP == "10^2", select=c(EVTYPE, PROPDMGEXP))

mydata3 <- subset(newdata, PROPDMGEXP == "10^9", select=c(EVTYPE, PROPDMGEXP))

mydata4 <- subset(newdata, PROPDMGEXP == "1", select=c(EVTYPE, PROPDMGEXP))

mydata5 <- subset(newdata, PROPDMGEXP == "0", select=c(EVTYPE, PROPDMGEXP))

mydata7 <- subset(newdata, PROPDMGEXP == "-", select=c(EVTYPE, PROPDMGEXP))

mydata8 <- subset(newdata, PROPDMGEXP == "?", select=c(EVTYPE, PROPDMGEXP))

mydata9 <- subset(newdata, PROPDMGEXP == "", select=c(EVTYPE, PROPDMGEXP))

print(head(mydata))

print(head(mydata1))

print(head(mydata2))

print(head(mydata3))

print(head(mydata4))

print("Printing 0...")
print(head(mydata5))
   
print("Printing -")
print(head(mydata7))

print("Printing ?")
print(head(mydata8))

print("Printing blank")
print(head(mydata9))

I have found that when I replace the empty string values in the PROPDMGEXP column with "0", this results in all the other replaced values that are not "0" (eg "10^3", "10^2", "10^6", etc) in the PROPDMGEXP column being replaced with "0".

I have verified this through the following:

print(dim(newdata))
> 902297      7

print(length(which(newdata$PROPDMGEXP == "0")))
> 902297

I am not sure why this is occurring. Any insights are appreciated.

Answer 1

The problem is grepl("", newdata$PROPDMGEXP) is all TRUE . If you ever need to use this kind of approach again, better use

newdata$PROPDMGEXP[newdata$PROPDMGEXP==""] <- "0"

especially if you are doing complete matching and/or you know exactly the string you're going to replace; or if it's empty. Because it will match only empty strings instead of match with everything.