UK postcode validation and format correction tool

I've been working on a regex-based UK postcode validation and format correction tool - with the aim of creating a list of postcodes that can be readily geocoded. The variety of postcode formats is explained here. I have based my regex on a discussion here.

My correction tool is designed to cope with a number of commonly made mistakes when a postcode is inputted in a free text format as such:

• Postcode entered in lower case

• No space between first and second half of postcode

• O instead of 0 in two key areas e.g. CH6O 8MG -> CH60 8MG & M8 OFG -> M8 0FG

I have managed this (somewhat clunky) here:

```
pc<-data.frame(postcode=c("GIR 0AA","M2 0AB","M2 OAB","M2 0ab","M1 1AA","M11AA","M60 1NW","M6O 1NW","M601NW","CR2 6XH","CR26XH","DN55 1PT","DN551PT","W1A 1HQ","W1A1HQ","w1a 1hq","EC1A 1BB","EC1A1BB"), true=c(TRUE,TRUE,FALSE,FALSE,TRUE,FALSE,TRUE,FALSE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,FALSE,TRUE,FALSE))

#Identify postcodes not in correct format
pc$original_validate<-grepl('^((GIR 0AA)|((([A-PR-UWYZ][A-HK-Y]?[0-9][0-9]?)|(([A-PR-UWYZ][0-9][A-HJKSTUW])|([A-PR-UWYZ][A-HK-Y][0-9][ABEHMNPRV-Y]))) [0-9][ABD-HJLNP-UW-Z]{2}))$',pc$postcode)
pc$non_pc<-ifelse(pc$original_validate ==FALSE,as.character(pc$postcode),"")

#Convert all to upper case
pc$new_pc<-toupper(pc$non_pc)

#Identify postcodes without a space and include a space
pc$non_pc2<-grepl('^((GIR 0AA)|((([A-PR-UWYZ][A-HK-Y]?[0-9][0-9]?)|(([A-PR-UWYZ][0-9][A-HJKSTUW])|([A-PR-UWYZ][A-HK-Y][0-9][ABEHMNPRV-Y]))) [0-9][ABD-HJLNP-UW-Z]{2}))$',pc$new_pc)
pc$non_pc2<-ifelse(pc$non_pc2==FALSE & pc$new_pc!="",as.character(pc$new_pc),"")
pc$new_pc2<-gsub("^((GIR 0AA)|((([A-PR-UWYZ][A-HK-Y]?[0-9][0-9]?)|(([A-PR-UWYZ][0-9][A-HJKSTUW])|([A-PR-UWYZ][A-HK-Y][0-9][ABEHMNPRV-Y])))))([0-9][ABD-HJLNP-UW-Z]{2})$","\\1 \\9",pc$non_pc2)
pc$non_pc<-NULL
pc$non_pc2<-NULL

#Identify postcodes containing an O instead of 0 in first half
pc$non_pc3<--grepl('^((GIR 0AA)|((([A-PR-UWYZ][A-HK-Y]?[0-9][0-9]?)|(([A-PR-UWYZ][0-9][A-HJKSTUW])|(

My first suggestion is to write a is_valid function since you are using this code in quite a few places:

is_valid <- function(x)
   grepl('^((GIR 0AA)|((([A-PR-UWYZ][A-HK-Y]?[0-9][0-9]?)|(([A-PR-UWYZ][0-9][A-HJKSTUW])|([A-PR-UWYZ][A-HK-Y][0-9][ABEHMNPRV-Y]))) [0-9][ABD-HJLNP-UW-Z]{2}))

Then, it is also a good idea to write a separate function for each of the fixes you have. It makes your code easier to test and reuse. Also, when using appropriate function names, it makes your code self-explanatory and can replace all the comments:

fix_case <- toupper

fix_space <- function(x)
   sub("^((GIR 0AA)|((([A-PR-UWYZ][A-HK-Y]?[0-9][0-9]?)|(([A-PR-UWYZ][0-9][A-HJKSTUW])|([A-PR-UWYZ][A-HK-Y][0-9][ABEHMNPRV-Y])))))([0-9][ABD-HJLNP-UW-Z]{2})$","\\1 \\9", x)

fix_zero_1 <- function(x)
   sub("([A-Z][0-9]|[A-Z][A-Z]|[A-Z][A-Z][0-9])[O]","\\10", x)

fix_zero_2 <- function(x) {
   x <- sub("^((GIR 0AA)|((([A-PR-UWYZ][A-HK-Y]?[0-9][0-9]?)|(([A-PR-UWYZ][0-9][A-HJKSTUW])|([A-PR-UWYZ][A-HK-Y][0-9][ABEHMNPRV-Y]))))) ([O][ABD-HJLNP-UW-Z]{2})$", "\\1 0\\9", x)
   sub("0O", "0", x)
}


With these, your code can simplify to:

postcodes <- c(
  "GIR 0AA",
  "M2 0AB",
  "M2 OAB",
  "M2 0ab",
  "M1 1AA",
  "M11AA",
  "M60 1NW",
  "M6O 1NW",
  "M601NW",
  "CR2 6XH",
  "CR26XH",
  "DN55 1PT",
  "DN551PT",
  "W1A 1HQ",
  "W1A1HQ",
  "w1a 1hq",
  "EC1A 1BB",
  "EC1A1BB"
)

bg <- postcodes
bg <- ifelse(is_valid(bg), bg, fix_case(bg))
bg <- ifelse(is_valid(bg), bg, fix_space(bg))
bg <- ifelse(is_valid(bg), bg, fix_zero_1(bg))
bg <- ifelse(is_valid(bg), bg, fix_zero_2(bg))

data.frame(original   = postcodes,
           best_guess = bg,
           is_valid   = is_valid(bg))


Note how I am using vectors everywhere and only in the end putting the results in a data.frame. Indeed, there is arguably no need for a data.frame until you want to see the results in a nice format, at the very end. Keeping things in vectors avoid the repetitive pc$ and the annoying conversion from character vectors to factors (what forced you to use as.character).

I have slightly modified your algorithm, where instead of replacing valid postcodes by "" so they won't get affected by fixes, I am propagating a vector of valid or "best guess" codes (bg) via the construct:

bg <- ifelse(is_valid(bg), bg, som_fix(bg))


It is shorter this way and removes the need for the error-prone paste() at the end of your code., x)

Then, it is also a good idea to write a separate function for each of the fixes you have. It makes your code easier to test and reuse. Also, when using appropriate function names, it makes your code self-explanatory and can replace all the comments:

%%CODEBLOCK_1%%

With these, your code can simplify to:

%%CODEBLOCK_2%%

Note how I am using vectors everywhere and only in the end putting the results in a data.frame. Indeed, there is arguably no need for a data.frame until you want to see the results in a nice format, at the very end. Keeping things in vectors avoid the repetitive pc$ and the annoying conversion from character vectors to factors (what forced you to use as.character).

I have slightly modified your algorithm, where instead of replacing valid postcodes by "" so they won't get affected by fixes, I am propagating a vector of valid or "best guess" codes (bg) via the construct:

%%CODEBLOCK_3%%

It is shorter this way and removes the need for the error-prone paste() at the end of your code.

StackExchange Code Review Q#117801, answer score: 3