Generating regular expression that would replace a specific character located in a specific place

[Konrad Zdeb]

Hi,

I have a data set corresponding to the extract below.

clear
input str9 var1
"A11OAB"
"B22OBD"
"CC21OAA "
end

I'm looking for a regular expression that would replace O with 0 across the observations that are meeting the following criteria:

• Are 6 characters long
• Start with a letter
• End with two letters
• Counting from left, O is fourth character

[daniel klein]

generate byte flag = ///
    regexm(var1, "^[a-zA-z]..[O][a-zA-Z][a-zA-Z]$")

Best
Daniel

[Konrad Zdeb]

Thank you very much.

[Christophe Kolodziejczyk]

How do you replace the O with 0? Like this?

gen newvar1 = substr(var1,1,3)+"0"+substr(var1,5,2) if flag

In SAS you could write

​substr(var1,4,1) = "0"

Is there a similar solution in Stata? I know that I could use subinstr(), but it could replace other O's coming before and after the fourth position.

Moreover it does not seem possible to request repeating patterns with Stata regex. I would like to write something like

​generate byte flag = ///
    regexm(var1, "^[a-zA-z]..[O][a-zA-Z]{2}$")

It should be possible if Stata is following the Posix.2 standard, shouldn't it?

[daniel klein]

Your first syntax is what I had in mind. I am not aware of an official Stata command or function that matches your SAS code.

Regarding the last point, the apparent lack of proper documentation of Stata's regex machinery has been discussed on the list repeatedly. It seems there is merely no complete answer. With Stata 14's unicode features, what you want seems to be possible - at least there is an example in the respective help file.

Best
Daniel

[Robert Picard]

I don't like the new names of the unicode regex functions. Here's an example that shows a repeating pattern

clear
input str9 var1
"A11OAB"
"B22OBD"
"CC21OAA "
end

clonevar vcopy = var1
replace var1 = ustrregexs(1) + "0" + ustrregexs(2) if ///
    ustrregexm(var1,"^([a-zA-z]..)O([a-zA-Z]{2})$")
    
list if vcopy != var1