Optimizing array difference search speed (or other solutions welcome!)

[phatzilla 0]

Hi gang,

I'm in a bit of a pickle here, the gist of it is that i have one "main" file with ~200k lines, and literally hundreds of other files each ranging from 1k - 100k lines.

I need to go through each of the "other" files and (separately) compare them to the "main" file, and save the differences between the two (no duplicates)

 

The issue is that each comparison (especially if the "other" file has 50+ k lines) takes over a minute each... 

Is there anyway to cut this time down? As far as i know im using the most optimized array difference script


Here's a rough mockup of the script im currently using

note: the main file has all unique lines, and the "other" files wont ever have any lines that *DO NOT* appear in the main file, if that helps

expandcollapse popup

#include <array.au3>
#include <file.au3>



global $info_file
global $compare_file
global $Differece

$info_file = FileReadToArray("info-file(200k lines).txt")




$compare_file = FileReadToArray("compare-file(100k lines).txt")




$Differece = _Diff($info_file, $compare_file, 0) ; get the difference between 2 arrays, NO duplicates
_ArrayDisplay($Differece)







;=================================================
; Function Name:   _Diff($Set1, $Set2 [, $GetAll=0 [, $Delim=Default]])
; Description::    Find values in $Set1 that do not occur in $Set2
; Parameter(s):    $Set1    set 1 (1D-array or delimited string)
;                  $Set2    set 2 (1D-array or delimited string)
;      optional:   $GetAll  0 - only one occurence of every difference are shown (Default)
;                           1 - all differences are shown, allowing duplicates
;      optional:   $Delim   Delimiter for strings (Default use the separator character set by Opt("GUIDataSeparatorChar") )
; Return Value(s): Succes   1D-array of values from $Set1 that do not occur in $Set2
;                  Failure  -1  @error  set, that was given as array, isn't 1D-array
; Note:            Comparison is case-sensitive! - i.e. Number 9 is different to string '9'!
; Author(s):       BugFix (bugfix@autoit.de) Modified by ParoXsitiC for Faster _Diff (Formally _GetIntersection)
;=================================================
Func _Diff(ByRef $Set1, ByRef $Set2, $GetAll = 0, $Delim = Default)
    Local $o1 = ObjCreate("System.Collections.ArrayList")
    Local $o2 = ObjCreate("System.Collections.ArrayList")
    Local $oDiff1 = ObjCreate("System.Collections.ArrayList")
    Local $tmp, $i
    If $GetAll <> 1 Then $GetAll = 0
    If $Delim = Default Then $Delim = Opt("GUIDataSeparatorChar")
    If Not IsArray($Set1) Then
        If Not StringInStr($Set1, $Delim) Then
            $o1.Add($Set1)
        Else
            $tmp = StringSplit($Set1, $Delim, 1)
            For $i = 1 To UBound($tmp) - 1
                $o1.Add($tmp[$i])
            Next
        EndIf
    Else
        If UBound($Set1, 0) > 1 Then Return SetError(1, 0, -1)
        For $i = 0 To UBound($Set1) - 1
            $o1.Add($Set1[$i])
        Next
    EndIf

    If Not IsArray($Set2) Then
        If Not StringInStr($Set2, $Delim) Then
            $o2.Add($Set2)
        Else
            $tmp = StringSplit($Set2, $Delim, 1)
            For $i = 1 To UBound($tmp) - 1
                $o2.Add($tmp[$i])
            Next
        EndIf
    Else
        If UBound($Set2, 0) > 1 Then Return SetError(1, 0, -1)
        For $i = 0 To UBound($Set2) - 1
            $o2.Add($Set2[$i])
        Next
    EndIf

    For $tmp In $o1
        If Not $o2.Contains($tmp) And ($GetAll Or Not $oDiff1.Contains($tmp)) Then $oDiff1.Add($tmp)
    Next

    If $oDiff1.Count <= 0 Then Return 0

    Local $aOut[$oDiff1.Count]
    $i = 0
    For $tmp In $oDiff1
        $aOut[$i] = $tmp
        $i += 1
    Next
    Return $aOut
EndFunc   ;==>_Diff

 

Edited October 16, 2015 by phatzilla

[BrewManNH 1,074]

Merge the 2 arrays into one, then run that merged array through _ArrayUnique to get the array with no repeats.

[mikell 635]

You could directly try Scripting.Dictionary , see here
 

[BrewManNH 1,074]

_ArrayUnique uses a scripting dictionary

 

[phatzilla 0]

BrewManNH,

Will _arrayunique remove ALL duplicate entries?   If i recall, it simply keeps one of the entries.

If you run:

1
3
3
2

through _arrayunique, you'll get

1
3
2

I need it to be
1
2
 

 

 

What i need is a new array that's the difference between "main file" and "other file".

So if main file array is

2
1
4
5
3

and other file array is

2
3
1

new array should be

4
5

 

In other words, i dont want to preserve any instances of the duplicate lines.  the _Diff function achieves this, but i need to get it a bit faster, if possible...

Or, failing that, a new solution.

 

Edited October 16, 2015 by phatzilla

[BrewManNH 1,074]

What's in the files that you're comparing between the 2? What do some of the entries look like? I personally would sort the array to be searched and then use _ArrayBinarySearch to quickly search the array for any entries that are duplicates. I'd need a sample of a couple of files you're comparing, even generic ones that the format is similar to what you're working with, I can probably create something that will do this very fast. Using objects on arrays probably isn't very efficient.

[phatzilla 0]

They are text files.

Each line is about 100-120 chars in length alphanumeric

Edited October 16, 2015 by phatzilla

[mikell 635]

_ArrayUnique uses a scripting dictionary

I know, I said directly because the OP asked for "save the differences between the two"  and _ArrayUnique doesn't do that

Edited October 16, 2015 by mikell

[jchd 981]

How do lines differ, typically? Are they 99% the same or much more random? Are the differences located near the head, near the tail or can they differ anywhere in between?

This can dramatically change the speed of comparisons. Have you tried comparing the hash of the strings (MD5 will do here): comparing MD5s is pretty fast.

Depending on actual contents, SQLite might prove useful for your use case.

[phatzilla 0]

Actually, i just adjusted my script based on mikells suggestion....



 

global $a = FileReadToArray("info-file(200k lines).txt")
global $b = FileReadToArray("compare-file(100k lines).txt")

Local $sda = ObjCreate("Scripting.Dictionary")
Local $sdb = ObjCreate("Scripting.Dictionary")
Local $sdc = ObjCreate("Scripting.Dictionary")

For $i In $a
    $sda.Item($i)
Next
For $i In $b
    $sdb.Item($i)
Next

For $i In $a
    If $sdb.Exists($i) Then $sdc.Item($i)
Next
$asd3 = $sdc.Keys()

For $i In $asd3
    If $sda.Exists($i) Then $sda.Remove($i)
    If $sdb.Exists($i) Then $sdb.Remove($i)
Next
$asd1 = $sda.Keys()
$asd2 = $sdb.Keys()

_ArrayDisplay($asd1, "$asd1")
_ArrayDisplay($asd2, "$asd2")
_ArrayDisplay($asd3, "$asd3")

Now this is much faster... $ASD1 is the only array that i need as it contains the lines that are *ONLY* found in "info" file (because the "other" compare file will never contain any lines that aren't found in info file).  Is there any way to speed this script up?  I dont really need the other 2 new arrays asd2,asd3.  asd1 contains all the information i need.

Edited October 16, 2015 by phatzilla

[kcvinu 64]

I have a  small doubt. You people see the above code ?. Is this safe to use the same variable name twice or thrice in the same script ?. I mean "$i". 

Edit - Never mind i got the answer. 

Edited October 17, 2015 by kcvinu

[Melba23 2,683]

kcvinu,

To answer your question for any later readers: Yes, it is quite safe to use the same variable name in several For...Next loops as long as each loop is entirely separate and has terminated before the next one starts. If there is any overlap or nesting then the loop count variables must be differentiated.

M23

[jchd 981]

In this case you can spare significant steps:

Local $a = FileReadToArray("info-file(200k lines).txt")
Local $sda = ObjCreate("Scripting.Dictionary")

For $i In $a
    $sda.Item($i)
Next

; for each "compare-file" do
; see _FileListToArray and friends

    ; also $a content is no more needed,
    ; we reuse it for all target files
    $a = FileReadToArray("compare-file(100k lines).txt")

    For $i In $a
        If $sda.Exists($i) Then $sda.Remove($i)
    Next
; next file

Local $asd1 = $sda.Keys()
$sda = 0
_ArrayDisplay($asd1, "$asd1")

 

[kcvinu 64]

@Melba23 Thank you for the clarification. 

[mikell 635]

kcvinu,
Note also that the $i is declared as Local in the For loop (Thus Saith The Help File) but keeps its last value as long as it is not re-used

For $i = 1 to 10
   If $i = 5 Then Exitloop
Next
;...
Msgbox(0,"", $i)

[phatzilla 0]

Problem is solved!(for now).  Thank you for everyones help....  Got the time down to ~10 seconds max for the biggest files

[kylomas 411]

phatzilla,

Just for grins can you run your largest files through the following.  I'm interested in comparing the results to the dictionary method...

#include <FileConstants.au3>
#include <sqlite.au3>
#include <array.au3>

Local $sSQLiteMod = 'C:\Program Files (x86)\AutoIt3\install\Extras\SQLite\sqlite3.exe'

; do NOT use AutoIt macros for the next two file names

Local $sMasterFile = "fully.qualified.filename" ; your file with ALL entries
Local $sDetailFile = "fully.qualified.filename" ; your file with SOME entries

$st = TimerInit()

Local $sIn = 'create table if not exists mstr (c1);' & _
        'create table if not exists det (c1);' & _
        @CRLF & '.import ' & $sMasterFile & ' MSTR' & _
        @CRLF & '.import ' & $sDetailFile & ' DET' & _
        @CRLF & 'SELECT DISTINCT c1 FROM mstr WHERE c1 Not IN (SELECT DISTINCT c1 FROM det);'

Local $Out = _TempFile()
Local $in = _TempFile()

FileWrite($in, $sIn)

Local $sCmd = @ComSpec & ' /c "' & $sSQLiteMod & '" < ' & $in & ' > ' & $Out
Local $pid = RunWait($sCmd, @WorkingDir, @SW_HIDE)
ProcessWaitClose($pid)
ConsoleWrite('Time to find entries unique to mstr = ' & Round(TimerDiff($st) / 1000, 3) & ' seconds' & @CRLF)
ConsoleWrite(StdoutRead($pid) & @CRLF)
ConsoleWrite('! out ' & @CRLF & FileRead($Out) & @CRLF)

FileDelete($Out)
FileDelete($in)

kylomas

[phatzilla 0]

Hmm, i dont seem to have sqlite3.exe

[jguinch 387]

For this kind of handling, which requires to be fast enough, I use an external program from the CoreUtils for Windows (http://gnuwin32.sourceforge.net/packages/coreutils.htm) : uniq.exe. This tool can quickly extract/remove duplicate lines in a text file. In your case, you just have to put the 2 file contents in 1 file and remove the duplicates (it requires the file to be sorted first)

So, using the same idea than BrewManNH with _ArrayUniq, you can do something like this :

#Include <Array.au3>

Local $sMainFile = @ScriptDir & "\mainFile.txt"
Local $sPartialFile = @ScriptDir & "\secondFile.txt"

Local $sTempFile = @ScriptDir & "\tempFile.txt", $sResultFile = @ScriptDir & "\resultFile.txt"

; Concatenate the 2 files in 1
Local $sfile = FileWrite($sTempFile, FileRead($sMainFile) & FileRead($sPartialFile) )

; Sort the file and then remove all duplicate lines
RunWait(@ComSpec & ' /c sort "' & $sTempFile & '" | uniq -u > "' & $sResultFile & '"', "C:\Program Files (x86)\GnuWin32\bin", @SW_HIDE)
FileDelete($sTempFile)

Local $aResult = FileReadToArray($sResultFile)
_ArrayDisplay($aResult)

I know the problem is solved, but maybe someone else could be interested by this method...

Edited October 17, 2015 by jguinch