powershell 获取特定子串位置的总和

Question

How can I get the sum of a file from a substring and placing the sum on a specific position (different line) using powershell if have the following conditions:

Get the sum of the numbers from position 3 to 13 of a line that is starting with a character D. Place the sum on position 10 to 14 on the line that starts with the S

So for example, if i have this file:

F123trial   text
DA00000038.95==xxx11
DA00000018.95==yyy11
DA00000018.95==zzzyy
S        xxxxx

I want to get the sum of 38.95, 18.95 and 18.95 and then place the sum on position xxxxx under the line that starts with the S.

Answer 1

You could try:

• -match to find the lines using regex-pattern
• The .NET string-method Substring() to extract the values from the "D"-lines
• Measure-Object -Sum to calculate the sum
• -replace to insert the value (searches using regex-pattern).

Ex:

$text = Get-Content -Path file.txt

$total = $text -match '^D' |
#Foreach "D"-line, extract the value and cast to double (to be able to sum it)
ForEach-Object { $_.Substring(2,11) -as [double] } |
#Measure the sum
Measure-Object -Sum | Select-Object -ExpandProperty Sum

$text | ForEach-Object {
    if($_ -match '^S') {
        #Line starts with S -> Insert sum
        $_.SubString(0,(17-$total.Length)) + $total + $_.SubString(17)
    } else {
        #Not "S"-line -> output original content
        $_
    }
} | Set-Content -Path file.txt

Answer 2

PowerShell's switch statement has powerful, but little-known features that allow you to iterate over the lines of a file ( -file ) and match lines by regular expressions ( -regex ) .

Not only is switch -file convenient , it is also much faster than using cmdlets in a pipeline (see bottom section).

[double] $sum = 0

switch -regex -file file.txt {

  # Note: The string to the left of each script block below ({ ... }), 
  #       e.g., '^D', is the regex to match each line against.
  #       Inside the script blocks, $_ refers to the input line at hand.

  # Extract number, add to sum, output the line.
  '^D' { $sum += $_.Substring(2, 11); $_; continue }

  # Summary line: place sum at character position 10, with 0-padding
  # Note: `-replace ',', '.'` is only needed if your culture uses "," as the
  #       decimal mark.
  '^S' { $_.Substring(0, 9) + '{0:000000000000000.00}' -f $sum -replace ',', '.'; continue }
  
  # All other lines: pass them through.
  default { $_ }

}

^Note:

• continue in the script blocks short-circuits further matching for the line at hand; by contrast, if you used break , no further lines would be processed .
• Based on a later comment, I'm assuming you want an 18-character 0 -left-padded number on the S line at character position 10 .

With your sample file, the above yields:

F123trial   text
DA00000038.95==xxx11
DA00000018.95==yyy11
DA00000018.95==zzzyy
S        000000000000076.85

---

Optional reading: Comparing the performance of switch -file ... to Get-Content ... | ForEach-Object ... Get-Content ... | ForEach-Object ...

Running the following test script:

& {
  # Create a sample file with 100K lines.
  1..1e5 > ($tmpFile = [IO.Path]::GetTempFileName())
  (Measure-Command { switch -file ($tmpFile) { default { $_ } } }).TotalSeconds, 
  (Measure-Command { get-content $tmpFile | % { $_ }  }).TotalSeconds
  Remove-Item $tmpFile
}

yields the following timings on my machine, for instance (the absolute numbers aren't important, but their ratio should give you a sense):

0.0578924   # switch -file
6.0417638   # Get-Content | ForEach-Object

That is, the pipeline-based solution is about 100 (!) times slower than the switch -file solution.

---

Digging deeper:

Frode F. points out that Get-Content is slow with large files - though its convenience makes it a popular choice - and mentions using the .NET Framework directly as an alternative:

• Using [System.IO.File]::ReadAllLines() ; however, given that it reads the entire file into memory, that is only an option with smallish files.

• Using [System.IO.StreamReader] 's ReadLine() method in a loop.

However, use of the pipeline in itself , irrespective of the specific cmdlets used, introduces overhead. When performance matters - but only then - you should avoid it.

Here's an updated test that includes commands that use the .NET Framework methods, with and without the pipeline (the use of collection operator .ForEach() requires PSv4+):

& {
  # Create a sample file with 100K lines.
  1..1e5 > ($tmpFile = [IO.Path]::GetTempFileName())
  
  (Measure-Command { switch -file ($tmpFile) { default { $_ } } }).TotalSeconds
  (Measure-Command { 
    $sr = [IO.StreamReader] (Convert-Path $tmpFile)
    while(-not $sr.EndOfStream) { $sr.ReadLine() }
    $sr.Close() 
  }).TotalSeconds
  (Measure-Command { [IO.File]::ReadAllLines((Convert-Path $tmpFile)).ForEach({ $_ }) }).TotalSeconds
  (Measure-Command { [IO.File]::ReadAllLines((Convert-Path $tmpFile)) | % { $_ } }).TotalSeconds
  (Measure-Command { Get-Content $tmpFile | % { $_ }  }).TotalSeconds
  
  Remove-Item $tmpFile
}

Sample results, from fastest to slowest:

0.0571143  # switch -file
0.2035162  # [System.IO.StreamReader] in a loop
0.6756535  # [System.IO.File]::ReadAllText() with .ForEach() collection operator
1.5088355  # (pipeline) [System.IO.File]::ReadAllText() with ForEach-Object
5.9815751  # (pipeline) Get-Content with ForEach-Object

switch -file is the fastest by a factor of around 3, followed by the .NET + loop solution; using .ForEach() adds another factor of 3. Simply introducing the pipeline ( ForEach-Object instead of .ForEach() ) adds another factor of 2; finally, using the pipeline with Get-Content and ForEach-Object adds another factor of 4.