If you need to find and replace a string in multiple files there are several tools that could do this. But recently I was in a situation where multiple URL:s needed to be updated in multiple files. So, I decided to do this in PowerShell instead.

The problem

Let says you have a file like this:

Sample file 
         First link with http: http://server.com/old1
        First link with https: https://server.com/old1
       First link with quotes: "http://server.com/old1"
       First link with quotes: 'http://server.com/old1'
                   First link: http://server.com/old1. A new sentence.
  First link with extra slash: http://server.com/old1/
 First link with query string: http://server.com/old1?search=no

        Second link with http: http://server.com/old2
       Second link with https: https://server.com/old2
 Second link with extra slash: http://server.com/old2/
Second link with query string: http://server.com/old2?search=no

Not first link: http://server.com/old10
Not second link: http://server.com/old20

Also imagine you have several files like this. Now you need to change all URL:s that end with old1 and old2. How do you do this efficiently?

A naive solution

My first attempt to solve this was a PowerShell script like this:

PowerShell - naive solution 
$files = Get-ChildItem -Include ("*.txt") -Recurse
$data = @(
    [PSCustomObject]@{
        Old = 'http://server.com/old1';
        New = 'https://server.com/new1';
    },
    [PSCustomObject]@{
        Old = 'http://server.com/old2';
        New = 'https://server.com/new2';
    }
)

$files | ForEach-Object {
    $filecontent = Get-Content $_
    $data | ForEach-Object {
        $filecontent = $filecontent -replace $_.Old, $_.New
    }
    Set-Content -Path $_.FullName -Value $filecontent
}

This mostly works, and in some situations, this might be good enough. But there are some limitations:

  • It only replaces URL:s that begins with http. You could work around this by making the array twice as large.
  • It will also partly replace URL:s that end with old10 - these should not be updated at all.
  • If the URL ends with a slash, it will not be removed. Not a big problem, but we want to fix this too.

The solution

This code is a bit more complex, but it solves all problems mention earlier:

PowerShell - naive solution 
$files = Get-ChildItem -Include ("*.txt") -Recurse
$data = @(
    [pscustomobject]@{
        Old = 'server.com/old1';
        New = 'server.com/new1';
    },
    [pscustomobject]@{
        Old = 'server.com/old2';
        New = 'server.com/new2';
    }
)

$files | ForEach-Object {
    $filecontent = Get-Content $_
    $data | ForEach-Object {
        $filecontent = $filecontent -replace "http(s)?://$($_.Old)(\/)?([^a-zA-Z0-9]|`$)", "https://$($_.New)`$3"
    }
    Set-Content -Path $_.FullName -Value $filecontent
}

The replacement expression might need some clarification. First, this is what we are looking for:

  • http(s)?:// - It should start with http:// or https://
  • $($_.Old) - This is the server and path previously defined.
  • (\/)? - This is the slash at the end of the URL - if there is any.
  • ([^a-zA-Z0-9]|`$) - This is a character after the URL. It should not be a number or a letter - but it might be end of string.

This is then replaced with:

  • https:// - It is what it looks like
  • $($_.New) - The new server and path previously defined.
  • `$3 - The extra character after the URL that we might have catch.

By the way, the accent before the dollar sign (like in `$3) is just an escape sequence, telling PowerShell that the dollar sign should be as it is and not be treated like a variable.

Summary

This was a fun little exercise for me. PowerShell and regular expressions are something that I always need to practice on.