This question is related to this previous one on how to replace accented strings like México with equivalent Latex code M\'{e}xico.
My problem here is slightly different. I am using a third party database with string variables with Spanish accents like above. However, the encoding appears odd since this is the behavior I get:
> grep("México",temp$dest_nom_ent)
integer(0)
> grep("Mexico",temp$dest_nom_ent)
integer(0)
> grep("xico",temp$dest_nom_ent)
[1] 18 19 20
> temp$dest_nom_ent[grep("xico",temp$dest_nom_ent)]
[2] "México" "México" "México"
where temp$dest_nom_ent is a variable with state names of México.
My question, then, is how to convert the string variable from the third party database into an encoding that standard R functions will recognize. Please note:
> Encoding(temp$dest_nom_ent)
[1] "unknown" "unknown" "unknown" "unknown" "unknown" "unknown" "unknown"
[8] "unknown" "unknown" "unknown" "unknown" "unknown" "unknown" "unknown"
[15] "unknown" "unknown" "unknown" "unknown" "unknown" "unknown" "unknown"
[22] "unknown" "unknown" "unknown" "unknown" "unknown" "unknown" "unknown"
[29] "unknown" "unknown" "unknown" "unknown" "unknown" "unknown" "unknown"
[36] "unknown" "unknown" "unknown" "unknown" "unknown" "unknown" "unknown"
[43] "unknown" "unknown"
For further info I am using Windows 7 64. Also note:
> charToRaw(temp$dest_nom_ent[18])
[1] 4d e9 78 69 63 6f
Which from this source coincides with Windows Spanish (Traditional Sort) locale.
M=4d
é=e9
x=78
i=69
c=63
o=6f
And also note:
> charToRaw("México")
[1] 4d c3 a9 78 69 63 6f
> Encoding("México")
[1] "latin1"
I have tried the following unsuccessfully (e.g. meaning grep("é",temp$dest_nom_ent) returns null vector):
Encoding(temp$dest_nom_ent)<-"latin1"
temp$dest_nom_ent <- iconv(temp$dest_nom_ent,"","latin1")
temp$dest_nom_ent <- enc2utf8(temp$dest_nom_ent)
...
I checked supported character sets using iconvlist() and "WINDOWS-1252" is supported. The following, however, did not work:
> temp1 <- temp$dest_nom_ent[grep("xico",temp$dest_nom_ent)]
> temp1
[1] "México" "México" "México"
> Encoding(temp1)<-"WINDOWS-1252"
> temp1 <- iconv(temp1,"WINDOWS-1252","latin1")
> temp1
[1] "México" "México" "México"
> Encoding(temp1)
[1] "latin1" "latin1" "latin1"
> charToRaw(temp1[1])
[1] 4d e9 78 69 63 6f
> grep("é",temp1)
integer(0)
which compares to:
> temp2 <- c("México","México","México")
> temp2
[1] "México" "México" "México"
> Encoding(temp2)
[1] "latin1" "latin1" "latin1"
> charToRaw(temp2[1])
[1] 4d c3 a9 78 69 63 6f
> grep("é",temp2)
[1] 1 2 3)
Tried to find out the encoding by brute force like:
try(for(i in 1:length(iconvlist())){
temp1 <- temp$dest_nom_ent[grep("xico",temp$dest_nom_ent)]
Encoding(temp1)<-iconvlist()[i]
temp1 <- iconv(temp1,iconvlist()[i],"latin1")
print(grep("é",temp1))
print(i)
},silent=FALSE)
I am not familiar with try function but it still scapes at error instead of ignoring it so cannot check whole list:
...
[1] 17
integer(0)
[1] 18
integer(0)
[1] 19
integer(0)
[1] 20
Error in iconv(temp1, iconvlist()[i], "latin1") :
unsupported conversion from 'CP-GR' to 'latin1' in codepage 1252
Finally:
> Sys.getlocale()
[1] "LC_COLLATE=English_United States.1252;LC_CTYPE=English_United States.1252;LC_MONETARY=English_United States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252"
> d<-c("México","México")
> for(i in 1:7){d1 <- str_sub(d[1],i,i); print(d1)}
[1] "M"
[1] "Ã"
[1] "©
[1] "x"
[1] "i"
[1] "c"
[1] "o"
> print(grep("é",d))
[1] 1 2
So it seems I will have to change the computer’s locale as suggested here. Also see here
PS: In case you wonder how with an English_United States.1252 locale I managed to type d<-c("México","México") the way is by setting up a secondary Spanish keyboard (traditional sort) using Control Panel > Clock, Language and Region > Region and Language > Keyboards and Languages > Change Keyboards and under installed services click add and navigate to Spanish traditional sort. Then under advanced key settings you can create a short-cut to switch keyboards. In my case Shit+Alt. So if I want to type ñ in English default locale, I do Shift+Alt followed by ; and then Shift+Alt to go back to English keyboard.
Well, I could not determine the coding of accents but the following accomplishes what I wanted. The trick was to convert to UTF-8, set the
sub()optionuseBytes=TRUEand Joran’s suggestion to usesanitize.text.function=function(x){x}forxtable(). Here is the sample code. Easy to loop over all accented vowels:As actually implemented in a loop:
Still, the answer is incomplete in that I cannot explain what exactly was going on. Found it through trial and error.