# Find and Remove Duplicates

# This is the directory we'll check for duplicates
# TODO: make this a parameter
$check_dir = "C:\Program Files\Microsoft Office\Updates\Download\PackageFiles"
$backup_dir = "C:\Users\richa\tmp\backups"



# Methodolgy
# A hash value is computed for each file, files with the same hash are considered duplicate
# Sort the duplicate files found for a hash by their creation date in ascending order
# Deletion canidate will be the first file in the list


class FoundFile {
    [string]$Name
    [string]$FullName
    [string]$Directory
    [int]$Size
    [double]$fileSizeInMB
    [double]$fileSizeInKB
    [datetime]$Creation
    [string]$Hash

    FoundFile([System.IO.FileInfo] $file, [string]$hash) {
        $this.Name = $file.Name
        $this.FullName = $file.FullName
        $this.Directory = $file.DirectoryName
        $this.Size = $file.Length
        $this.fileSizeInMB = [math]::Round($this.Size / 1MB, 4)
        $this.fileSizeInKB = [math]::Round($this.Size / 1KB, 2)
        $this.Creation = $file.CreationTimeUtc        
        $this.Hash = $hash
    }

    # Method to display information
    [void]DisplayInfo() {
        Write-Output "hello"
    }
    
    [string]ToString() {
        return "$($this.Directory) $($this.Name) $(Format-NumberWithCommas $this.fileSizeInKB) KB $($this.Creation)"     
    }
}

function ShowSizes() {
    param (
        [Parameter(Mandatory = $true, Position = 0)]
        [string]$description,

        [Parameter(Mandatory = $true, Position = 1)]
        [int]$sizeInBytes
    )

    # Convert the size to kilobytes (KB)
    $sizeInKB = $sizeInBytes / 1KB

    # Convert the size to megabytes (MB)
    $sizeInMB = $sizeInBytes / 1MB

    # Convert the size to gigabytes (GB)
    $sizeInGB = $sizeInBytes / 1GB

    # Display the size in different units
    Write-Output ""
    Write-Output "$description $(Format-NumberWithCommas -Number $sizeInBytes) bytes."
    Write-Output "$description $(Format-NumberWithCommas -Number $sizeInKB, 2) KB."
    Write-Output "$description $(Format-NumberWithCommas -Number $sizeInMB, 2) MB."
    Write-Output "$description $(Format-NumberWithCommas -Number $sizeInGB, 4) GB."
}


function Format-NumberWithCommas {
    param (
        [Parameter(Mandatory = $true, Position = 0)]
        # no idea why this has to be System.Object, suggestions don't work 
        # for converting [math]::Round result into double and using a type of [double]
        [System.Object]$Number,

        [Parameter(Position = 1)]
        [int]$NumDecimals = 0
    )

    # TODO: sort out why this isn't working as expected, works fine on the command line!
    return "{0:N$NumDecimals}" -f $Number    
}

# Function to compute file hash
function Get-FileHash ($path) {
    $stream = [System.IO.File]::OpenRead($path)
    $sha256 = New-Object System.Security.Cryptography.SHA256Managed
    $hash = $sha256.ComputeHash($stream)
    $stream.Close()
    return [BitConverter]::ToString($hash) -replace '-', ''
}


function Get-AllFilesByHash {
    param (
        [Parameter(Mandatory = $true)]
        [string]$Dirname
    )

    # Get all files in the directory    
    $files = Get-ChildItem -Path $Dirname -File -Recurse

    # Create a hashtable to store file hashes
    $hashTable = @{}

    # Iterate through each file and compute hash
    foreach ($file in $files) {
        $hash = Get-FileHash $file.FullName

        # initialize array for this hash if needed
        if (!$hashTable.ContainsKey($hash)) {
            $hashTable[$hash] = @()
        }
        $ff = [FoundFile]::new($file, $hash)
        #Write-Output $ff.DisplayInfo()
        $hashTable[$hash] += $ff
    }
    $hashTable
}


function Get-Duplicates {
    param (
        [Parameter(Mandatory = $true)]
        [hashtable] $files
    )
    $dups_hash = @{}
    foreach ($entry in $files.GetEnumerator()) {
        $dup_files = $entry.Value
    
        if ($dup_files.Count -gt 1) {
            $dups_hash[$entry.key] = $dup_files | Sort-Object -Property Creation        
            #$dup_files = $dups_hash[$entry.key]
        }
    }
    $dups_hash
}


function Show-Duplicates {
    param (
        [Parameter(Mandatory = $true)]
        [hashtable] $files
    )

    $num_dups = 0
    $total_dup_size = 0
    
    foreach ($entry in $files.GetEnumerator()) {
        $dup_files = $entry.Value
        $num_dups += 1
    
        if ($dup_files.Count -lt 2) {
            throw "duplicates collection contains non duplicate for entry: $($entry.Key))"
        }

        Write-Output ""
        Write-Output "Found duplicate $($num_dups): $($entry.Key)"
        for ($i = 0; $i -lt $dup_files.Count; $i++) {
            $total_dup_size += $dup_files[$i].Size
            Write-Output "$($hash_num): $($dup_files[$i].ToString())"
        }        
    }
    Write-Output ""
    Write-Output "Found $num_dups duplicate file hashes."
    ShowSizes "Duplicate Files Size" $total_dup_size
}


function BackupDuplicateFile {
    param (
        [Parameter(Mandatory = $true)]
        [FoundFile] $file,

        [Parameter(Mandatory = $true)]
        [string] $backupDirectory
    )

    # Define the source file path
    $sourceFilePath = Join-Path -Path $file.Directory -ChildPath $file.Name

    # Get the FileInfo object for the source file
    $fileInfo = Get-Item -Path $sourceFilePath

    # Define the backup directory
    Write-Output "backup directory is: $backupDirectory"

    # Ensure the backup directory exists
    if (-not (Test-Path -Path $backupDirectory)) {
        New-Item -Path $backupDirectory -ItemType Directory
    }

    # Remove the drive letter using Split-Path
    $pathWithoutDrive = Split-Path -Path $File.Directory -NoQualifier

    # Construct the full backup file path
    #$fullBackupDirectory = Join-Path -Path $backupDirectory -ChildPath $File.Directory
    $fullBackupDirectory = Join-Path -Path $backupDirectory -ChildPath $pathWithoutDrive
    


    if (-not (Test-Path -Path $fullBackupDirectory)) {
        New-Item -Path $fullBackupDirectory -ItemType Directory
    }

    # Copy the file to the backup directory

    $fullDestinationPath = Join-Path -Path $fullBackupDirectory -ChildPath $file.Name
    $fileInfo.CopyTo($fullDestinationPath, $true)

    # Output the result
    # Write-Output "File '$($fileInfo.FullName)'"
    Write-Output "   + backed up to: $fullBackupDirectory"
}

function DeleteDuplicateFile {
    param (
        [Parameter(Mandatory = $true)]
        [FoundFile] $File
    )
    
    Write-Output ""
    Write-Output "Deleting File: $($File.Name)  $(Format-NumberWithCommas $File.fileSizeInKB) KB"
    BackupDuplicateFile -file $File -backupDirectory $backup_dir    
    # Force delete a read-only or protected file
    Write-Output "  ... would deleting: $($File.FullName)"
    #Remove-Item -Path $File.FullName -Force 
    Write-Output "   - removed from: $($File.Directory)"
}




# NOTE: Duplicates are pre-sorted by creation date
function ProcessDuplicates {
    param (
        [Parameter(Mandatory = $true)]
        [hashtable] $files
    )

    $num_deleted = 0
    $total_deleted_size = 0
    
    Write-Output ""
    foreach ($entry in $files.GetEnumerator()) {
        $dup_files = $entry.Value
        if ($dup_files.Count -lt 2) {
            throw "duplicates collection contains non duplicate for entry: $($entry.Key))"
        }

        # delete the first duplicate in the array of dups, they are sorted by creation date ascending
        $oldest_dup = $dup_files[0]
        DeleteDuplicateFile -File $oldest_dup
        $total_deleted_size += $oldest_dup.Size
        $num_deleted += 1
    }
    Write-Output ""
    Write-Output "Deleted $num_deleted duplicate files"
    ShowSizes "Recovered Size" $total_deleted_size
}





$files = Get-AllFilesByHash($check_dir) | Select-Object -First 10
$dups = Get-Duplicates $files
Show-Duplicates $dups
ProcessDuplicates $dups