I'm pulling my hair, to RegEx-tract the bare version information from some filenames. e.g. "1.2.3.4"

Let's assume, I have the following Filenames:

VendorSetup-x64-1.23.4.exe
VendorSetup-1-2-3-4.exe
Vendor Setup 1.23.456Update.exe
SoftwareName-1.2.34.5-x64.msi
SoftwareName-1.2.3.4-64bit.msi
SoftwareName-64-Bit-1.2.3.4.msi
VendorName_SoftwareName_64_1.2.3_Setup.exe

(And I know there are still some filenames out there, that have "x32" as well as "x86" in them, so I've added them to the title)

First of all, I replaced the _'s & -'s by .'s which I'd like to avoid in general, but haven't found a cleverer approach and to be honest - only works well if there's no other "digit"-information in the String for example like the 2nd Filename.

I then tried to extract the Version information using Regex like

-replace '^(?:\D )?(\d ((\.\d ){1,4})?)(?:.*)?', '$1'

Which lacks the ability to omit "x64", "64Bit", "64-Bit" or any variation of that generally.

Additionally, I played around with RegExes like

 -replace '^(?:[xX]*\d{2})?(?:\D )?(\d ((\.\d ){1,4})?)(?:.*)?$', '$1'

to try to omit a leading "x64" or "64", but with no success (most probably because of the replacement from -'s to .'s.

And before it gets even worse, I'd like to ask if there's anybody who could help me or lead me in the right direction?

Thanks in advance!

CodePudding user response：

This one works with all of the given samples:

(?<=[\s_-])(?:\d (?:\.\d ){1,3}|\d (?:-\d ){1,3})

You just have to replace dashes by dots in the captured value.

Demo:

# Create an array of sample filenames
$names = @'
VendorSetup-2022-05-x64-1.23.4.exe
VendorSetup-x64-1.23.4-2022-05.exe
VendorSetup-1-2-3-4.exe
VendorSetup_2022-05_1-2-3-4.exe
Vendor Setup 1.23.456Update.exe
SoftwareName-1.2.34.5-x64.msi
SoftwareName-1.2.3.4-64bit.msi
SoftwareName-64-Bit-1.2.3.4.msi
VendorName_SoftwareName_64_1.2.3_Setup.exe
'@ -split '\r?\n'

$versionPattern = '(?<=[\s_-])(?:\d (?:\.\d ){1,3}|\d (?:-\d ){1,3})'

foreach( $name in $names ) {
    
    $version = $null
    if( $matched = [regex]::Matches( $name, $versionPattern ) ) {

        $matchedValues = @($matched.Value)

        # Prefer the last match that contains '.'
        if( $lastValueWithDot = $matchedValues.Where({ $_.Contains('.') }, 'Last') ) {
            $version = $lastValueWithDot[ 0 ]  # Need index because .Where() always outputs an array
        }
        else {
            # Otherwise take the last match
            $version = $matchedValues[ -1 ]
        }
    }

    # Normalize the version number separators
    $version = $version -replace '-', '.'

    # Output custom object for nice table formatting
    [PSCustomObject]@{
        Name    = $name
        Version = $version
    }
}

Output:

Name                                       Version
----                                       -------
VendorSetup-2022-05-x64-1.23.4.exe         1.23.4
VendorSetup-x64-1.23.4-2022-05.exe         1.23.4
VendorSetup-1-2-3-4.exe                    1.2.3.4
VendorSetup_2022-05_1-2-3-4.exe            1.2.3.4
Vendor Setup 1.23.456Update.exe            1.23.456
SoftwareName-1.2.34.5-x64.msi              1.2.34.5
SoftwareName-1.2.3.4-64bit.msi             1.2.3.4
SoftwareName-64-Bit-1.2.3.4.msi            1.2.3.4
VendorName_SoftwareName_64_1.2.3_Setup.exe 1.2.3

Explanation:

The RegEx has three important parts:

• (?<=[\s_-]) … positive lookbehind assertion makes sure that the version is separated by space, underscore or dash on the left side. This prevents sub string 64-1 from the first sample to match as a version.
• (?: … introduces a non-capturing group of two alternatives:
  • \d (?:\.\d ){1,3} … 2 to 4 version numbers separated by .
  • \d (?:-\d ){1,3} … 2 to 4 version numbers separated by -

Note that the pattern also matches 2022-05 in some of the sample filenames which I modified. Instead of further complicating the pattern we can use [regex]::Matches() to capture all matches and resolve ambiguities through PowerShell code:

• Prefer the last matched value that contains ..
• If none of the matched values contain ., just take the last value.

See detailed explanation at RegEx101.