Transpose Columns to Rows - JREPL?

Discussion forum for all Windows batch related topics.

Moderator: DosItHelp

Post Reply
Message
Author
SIMMS7400
Posts: 544
Joined: 07 Jan 2016 07:47

Transpose Columns to Rows - JREPL?

#1 Post by SIMMS7400 » 04 Oct 2017 11:17

Hi Folks & Dave,

Late last year, everyone, particularity Dave was very kind and helped me meet a requirement using the famed JREPL batch utility to transpose columns to rows - it's still working flawlessly!

Recently, I have another requirement to perform a similar task of transposing columns to rows.

Here is my original data set :

Code: Select all

Cost Center,CONS Project,18-May,18-Jun,18-Jul,18-Aug
US_4465,PFP-E0000001,1.5,1.5,2.3,2.5


However, I need it to look like the following once transposed:

Code: Select all

US_4465,PFP-E0000001,18-May,1.5
US_4465,PFP-E0000001,18-Jun,1.5
US_4465,PFP-E0000001,18-Jul,2.3
US_4465,PFP-E0000001,18-Aug,2.5



Would this still be possible with JREPL? If it helps, below is the current JREPL code that's currently in use:

Code: Select all

@echo off
setlocal


CALL C:\Hyperion_Batch\Scripts\NTScripts\_env.cmd

set "input=%LOCALEXPORTPATH%ASCEND_Portfolio_Attr_Input.csv"
set "output=%LOCALEXPORTPATH%ASCEND_Portfolio_Attr_Output.csv"

set "beg=var id,c,col=new Array()"
set "begln=c=0"
set "find=(?:^|,)(\q(\q\q|[^\q])*\q|[^,]*)"
set "repl=if(++c==1)id=$0;if(ln==1)col[c]=$0;ln>2&&c>1?id+col[c]+$0:false"

CALL "%UTILPATH%JREPL\JREPL.bat" find repl /x /v /jmatch /jbeg beg /jbegln begln /f "%input%" /o "%output%"


Please let me know your thoughts, thank you!
Last edited by dbenham on 05 Oct 2017 04:30, edited 1 time in total.
Reason: Add links to referenced posts

SIMMS7400
Posts: 544
Joined: 07 Jan 2016 07:47

Re: Transpose Columns to Rows - JREPL?

#2 Post by SIMMS7400 » 05 Oct 2017 01:52

One last thing - there could be 15k+ lines of data, not just (2) as shown above as an example.

Thanks!

dbenham
Expert
Posts: 2461
Joined: 12 Feb 2011 21:02
Location: United States (east coast)

Re: Transpose Columns to Rows - JREPL?

#3 Post by dbenham » 05 Oct 2017 05:22

First things first, you can dramatically improve performance of the original script if you update to the newest version of JREPL.BAT, and use /JMATCHQ (which requires addition of "$txt=" - see the /JMATCHQ documentation), instead of /JMATCH .

Code: Select all

@echo off
setlocal


CALL C:\Hyperion_Batch\Scripts\NTScripts\_env.cmd

set "input=%LOCALEXPORTPATH%ASCEND_Portfolio_Attr_Input.csv"
set "output=%LOCALEXPORTPATH%ASCEND_Portfolio_Attr_Output.csv"

set "beg=var id,c,col=new Array()"
set "begln=c=0"
set "find=(?:^|,)(\q(\q\q|[^\q])*\q|[^,]*)"
set "repl=if(++c==1)id=$0;if(ln==1)col[c]=$0;$txt=ln>2&&c>1?id+col[c]+$0:false"

CALL "%UTILPATH%JREPL\JREPL.bat" find repl /x /v /jmatchq /jbeg beg /jbegln begln /f "%input%" /o "%output%"

I tested performance with a 4000 line input file, all with the most recent version. Using the old /JMATCH, it took 18.5 seconds, and with the new /JMATCHQ it only took 1 second :!:

The performance gain might be even more impressive depending on which version of JREPL you are currently using.

Now your new requirement is only slightly changed. You no longer are skipping a line, and you want to preserve the first two columns (what I call the id) with each replication, not just one. I'm assuming there is still a possibility for quoted values, even though you do not show any in your sample. So I will preserve the complexity of allowing quoted values (possibly with embedded commas and or escaped quotes).

So why not create a single batch script that can handle both jobs (and many others) :idea: :!:
You just need to pass in a few configuration parameters:
1 - Input File
2 - Output File
3 - How many extra columns you want to preserve as part of the "ID"
4 - How many lines to skip before you get actual data

To make construction of the FIND regex easier, I first define a FIELD variable that holds a regex that matches a single field value. The variable makes it easy to replicate that regex snippet as needed.

UNPIVOT.BAT

Code: Select all

:: UNPIVOT.BAT  InputFile  OutputFile  ExtraIDColCount  SkipCount
@echo off
setlocal

set "input=%~1"
set "output=%~2"
set "idCnt=%~3"
set "skip=%~4"

set "beg=var id,c,col=new Array()"
set "begln=c=0"
set "field=(?:\q(?:\q\q|[^\q])*\q|[^,]*)"
set "find=^%field%(?:,%field%){%idCnt%}|,%field%"
set "repl=if(++c==1)id=$0;if(ln==1)col[c]=$0;$txt=ln>%skip%&&c>1?id+col[c]+$0:false"

call jrepl find repl /x /v /jmatchq /jbeg beg /jbegln begln /f "%input%"  /o "%output%"


So to process your original files:

Code: Select all

UNPIVOT "originalFormatInput" "originalFormatOutput" 0 2


To process your new format:

Code: Select all

UNPIVOT "newFormatInput" "newFormatOutput" 1 1


Dave Benham

SIMMS7400
Posts: 544
Joined: 07 Jan 2016 07:47

Re: Transpose Columns to Rows - JREPL?

#4 Post by SIMMS7400 » 05 Oct 2017 08:16

Dave -

WOW! This is incredible - thank you so much for taking the time to walk me through this.

Both processes work like a charm!!

In regards to my original request (months ago), the new version of JREPL is exponentially faster with /JMATCHQ- wowza!

Thank you again, Dave! I really appreciate it!

Post Reply