修改 CSV 文件内容的更有效方法

Question

我正在尝试删除 SSMS 2012 在将查询结果导出为 CSV 时生成的一些碎片。

例如，它包含 null 值的单词 'NULL' 并向 datetime 值添加毫秒：

DATE_COLUMN,DATETIME_COLUMN,TEXT_COLUMN,NUMBER_COLUMN
2015-05-01,2015-05-01 23:00:00.000,LOREM IPSUM,10.3456
NULL,NULL,NULL,0

不幸的是，Excel 不会自动将 datetime 值正确格式化为小数秒，这会导致客户混淆（'What happened to the date field that I requested?') 和更多的工作（必须将 CSV 转换为 XLSX 并在分发之前正确格式化列）。

目标是去除 CSV 文件中的 NULL 和 .000 值：

DATE_COLUMN,DATETIME_COLUMN,TEXT_COLUMN,NUMBER_COLUMN
2015-05-01,2015-05-01 23:00:00,LOREM IPSUM,10.3456
,,,0

Excel 将打开此文件并正确格式化，无需进一步的技术帮助。

为此，我写道：

Function Invoke-CsvCleanser {

  [CmdletBinding()]
  Param(
    [parameter(Mandatory=$true)]
    [String]
    $Path,
    [switch]
    $Nulls,
    [switch]
    $Milliseconds
  )

  PROCESS {

    # open the file
    $data = Import-Csv $path

    # process each row
    $data | Foreach-Object { 

        # process each column
        Foreach ($property in $_.PSObject.Properties) {

            # if column contains 'NULL', replace it with ''
            if ($Nulls -and ($property.Value -eq 'NULL')) {
                $property.Value = $property.Value -replace 'NULL', ''
            }

            # if column contains a date/time value, remove milliseconds
            elseif ( $Milliseconds -and (isDate($property.Value)) ) {
                $property.Value = $property.Value -replace '.000', ''    
            }
        } 

    } 

    # save file
    $data | Export-Csv -Path $Path -NoTypeInformation

  }

}

function IsDate($object) {
    [Boolean]($object -as [DateTime])
}

PS> Invoke-CsvCleanser 'C:\Users\Foobar\Desktop[=12=]00.csv' -Nulls -Milliseconds

这在文件较小时效果很好，但对于大文件来说效率很低。理想情况下，Invoke-CsvCleanser 会利用管道。

有更好的方法吗？

Answer 1

Import-CSV 总是在内存中加载整个文件，所以速度很慢。这是我对这个问题的回答修改后的脚本：.

它使用原始文件处理，因此速度应该快得多。 NULLs 和毫秒是 matched\replaced 使用正则表达式。脚本能够批量转换 CSV。

拆分 CSV 的正则表达式来自这个问题：How to split a string by comma ignoring comma in double quotes

将此脚本另存为 Invoke-CsvCleanser.ps1。它接受以下参数：

InPath：从中读取 CSV 的文件夹。如果未指定，则使用当前目录。
OutPath: 用于保存已处理的 CSV 的文件夹。将创建，如果不存在。
编码: 如果未指定，脚本将使用系统当前的 ANSI 代码页来读取文件。您可以在 PowerShell 控制台中为您的系统获取其他有效编码，如下所示：[System.Text.Encoding]::GetEncodings()
DoubleQuotes: 开关，如果指定，周围的双引号将从值中剥离
Nulls: 开关，如果指定，NULL 字符串将从值中剥离
毫秒：开关，如果指定，.000个字符串将从值中剥离
详细：脚本将通过 Write-Verbose 消息告诉您发生了什么。

示例：

处理文件夹C:\CSVs_are_here中的所有CSV，去除NULL和毫秒，将处理后的CSV保存到文件夹C:\Processed_CSVs，详细：

.\Invoke-CsvCleanser.ps1 -InPath 'C:\CSVs_are_here' -OutPath 'C:\Processed_CSVs' -Nulls -Milliseconds -Verbose

Invoke-CsvCleanser.ps1 脚本：

Param
(
    [Parameter(ValueFromPipelineByPropertyName = $true)]
    [ValidateScript({
        if(!(Test-Path -LiteralPath $_ -PathType Container))
        {
            throw "Input folder doesn't exist: $_"
        }
        $true
    })]
    [ValidateNotNullOrEmpty()]
    [string]$InPath = (Get-Location -PSProvider FileSystem).Path,

    [Parameter(Mandatory = $true, ValueFromPipelineByPropertyName = $true)]
    [ValidateScript({
        if(!(Test-Path -LiteralPath $_ -PathType Container))
        {
            try
            {
                New-Item -ItemType Directory -Path $_ -Force
            }
            catch
            {
                throw "Can't create output folder: $_"
            }
        }
        $true
    })]
    [ValidateNotNullOrEmpty()]
    [string]$OutPath,

    [Parameter(ValueFromPipelineByPropertyName = $true)]
    [string]$Encoding = 'Default',

    [switch]$Nulls,

    [switch]$Milliseconds,

    [switch]$DoubleQuotes
)


if($Encoding -eq 'Default')
{
    # Set default encoding
    $FileEncoding = [System.Text.Encoding]::Default
}
else
{
    # Try to set user-specified encoding
    try
    {
        $FileEncoding = [System.Text.Encoding]::GetEncoding($Encoding)
    }
    catch
    {
        throw "Not valid encoding: $Encoding"
    }
}

$DQuotes = '"'
$Separator = ','
# 
$SplitRegex = "$Separator(?=(?:[^$DQuotes]|$DQuotes[^$DQuotes]*$DQuotes)*$)"
# Regef to match NULL
$NullRegex = '^NULL$'
# Regex to match milliseconds: 23:00:00.000
$MillisecondsRegex = '(\d{2}:\d{2}:\d{2})(\.\d{3})'

Write-Verbose "Input folder: $InPath"
Write-Verbose "Output folder: $OutPath"

# Iterate over each CSV file in the $InPath
Get-ChildItem -LiteralPath $InPath -Filter '*.csv' |
    ForEach-Object {
        Write-Verbose "Current file: $($_.FullName)"
        $InFile = New-Object -TypeName System.IO.StreamReader -ArgumentList (
            $_.FullName,
            $FileEncoding
        ) -ErrorAction Stop
        Write-Verbose 'Created new StreamReader'

        $OutFile = New-Object -TypeName System.IO.StreamWriter -ArgumentList (
            (Join-Path -Path $OutPath -ChildPath $_.Name),
            $false,
            $FileEncoding
        ) -ErrorAction Stop
        Write-Verbose 'Created new StreamWriter'

        Write-Verbose 'Processing file...'
        while(($line = $InFile.ReadLine()) -ne $null)
        {
            $tmp = $line -split $SplitRegex |
                        ForEach-Object {

                            # Strip surrounding quotes
                            if($DoubleQuotes)
                            {
                                $_ = $_.Trim($DQuotes)
                            }

                            # Strip NULL strings
                            if($Nulls)
                            {
                                $_ = $_ -replace $NullRegex, ''
                            }

                            # Strip milliseconds
                            if($Milliseconds)
                            {
                                $_ = $_ -replace $MillisecondsRegex, ''
                            }

                            # Output current object to pipeline
                            $_
                        }
            # Write line to the new CSV file
            $OutFile.WriteLine($tmp -join $Separator)
        }

        Write-Verbose "Finished processing file: $($_.FullName)"
        Write-Verbose "Processed file is saved as: $($OutFile.BaseStream.Name)"

        # Close open files and cleanup objects
        $OutFile.Flush()
        $OutFile.Close()
        $OutFile.Dispose()

        $InFile.Close()
        $InFile.Dispose()
    }

结果：

DATE_COLUMN,DATETIME_COLUMN,TEXT_COLUMN,NUMBER_COLUMN
2015-05-01,2015-05-01 23:00:00,LOREM IPSUM,10.3456
,,,0

It would be interesting to see if one could pass lambdas as a way to make the file processing more flexible. Each lambda would perform a specific activity (removing NULLs, upper-casing, normalizing text, etc.)

此版本提供了对 CSV 处理的完全控制。只需按照您希望它们执行的顺序将脚本块传递给 Action 参数。

示例：去掉NULLs，去掉毫秒，然后去掉双引号。

.\Invoke-CsvCleanser.ps1 -InPath 'C:\CSVs_are_here' -OutPath 'C:\Processed_CSVs' -Action {$_ = $_ -replace '^NULL$', '' }, {$_ = $_ -replace '(\d{2}:\d{2}:\d{2})(\.\d{3})', ''}, {$_ = $_.Trim('"')}

Invoke-CsvCleanser.ps1 与 "lambdas":

Param
(
    [Parameter(ValueFromPipelineByPropertyName = $true)]
    [ValidateScript({
        if(!(Test-Path -LiteralPath $_ -PathType Container))
        {
            throw "Input folder doesn't exist: $_"
        }
        $true
    })]
    [ValidateNotNullOrEmpty()]
    [string]$InPath = (Get-Location -PSProvider FileSystem).Path,

    [Parameter(Mandatory = $true, ValueFromPipelineByPropertyName = $true)]
    [ValidateScript({
        if(!(Test-Path -LiteralPath $_ -PathType Container))
        {
            try
            {
                New-Item -ItemType Directory -Path $_ -Force
            }
            catch
            {
                throw "Can't create output folder: $_"
            }
        }
        $true
    })]
    [ValidateNotNullOrEmpty()]
    [string]$OutPath,

    [Parameter(ValueFromPipelineByPropertyName = $true)]
    [string]$Encoding = 'Default',

    [Parameter(Mandatory = $true, ValueFromPipelineByPropertyName = $true)]    
    [scriptblock[]]$Action
)


if($Encoding -eq 'Default')
{
    # Set default encoding
    $FileEncoding = [System.Text.Encoding]::Default
}
else
{
    # Try to set user-specified encoding
    try
    {
        $FileEncoding = [System.Text.Encoding]::GetEncoding($Encoding)
    }
    catch
    {
        throw "Not valid encoding: $Encoding"
    }
}

$DQuotes = '"'
$Separator = ','
# 
$SplitRegex = "$Separator(?=(?:[^$DQuotes]|$DQuotes[^$DQuotes]*$DQuotes)*$)"

Write-Verbose "Input folder: $InPath"
Write-Verbose "Output folder: $OutPath"

# Iterate over each CSV file in the $InPath
Get-ChildItem -LiteralPath $InPath -Filter '*.csv' |
    ForEach-Object {
        Write-Verbose "Current file: $($_.FullName)"
        $InFile = New-Object -TypeName System.IO.StreamReader -ArgumentList (
            $_.FullName,
            $FileEncoding
        ) -ErrorAction Stop
        Write-Verbose 'Created new StreamReader'

        $OutFile = New-Object -TypeName System.IO.StreamWriter -ArgumentList (
            (Join-Path -Path $OutPath -ChildPath $_.Name),
            $false,
            $FileEncoding
        ) -ErrorAction Stop
        Write-Verbose 'Created new StreamWriter'

        Write-Verbose 'Processing file...'
        while(($line = $InFile.ReadLine()) -ne $null)
        {
            $tmp =  $line -split $SplitRegex |
                        ForEach-Object {
                            # Process each item
                            foreach($scriptblock in $Action) {
                                . $scriptblock
                            }
                            # Output current object to pipeline
                            $_
                        }
            # Write line to the new CSV file
            $OutFile.WriteLine($tmp -join $Separator)
        }

        Write-Verbose "Finished processing file: $($_.FullName)"
        Write-Verbose "Processed file is saved as: $($OutFile.BaseStream.Name)"

        # Close open files and cleanup objects
        $OutFile.Flush()
        $OutFile.Close()
        $OutFile.Dispose()

        $InFile.Close()
        $InFile.Dispose()
    }

修改 CSV 文件内容的更有效方法

More-efficient way to modify a CSV file's content

csv

powershell

powershell-3.0

ssms-2012