Re: sed or awk under XP in batch file (DOS box) - print $1 lines then delete $1 lines from a file

[John Bartley K7AAY]

On May 1, 5:32 pm, "Kam-Hung Soh" <kamhung....@gmail.com> wrote:
> On Fri, 02 May 2008 09:13:35 +1000, John Bartley K7AAY  
>
>
>
> <john.bart...@gmail.com> wrote:
> > On May 1, 4:02 pm, b...@proulx.com (Bob Proulx) wrote:
> >> John Bartley K7AAY wrote:
> >> > I need to print $1 lines from a file, and then delete that number of
> >> > lines.
> >> > $1 has been derived in the prior line with
> >> > wc -l sourcefile.txt | awk '{$1 /= 4 ; $1 = int($1) ; print $1 }'
>
> >> That all looks okay.  But I would probably personally do it all in the
> >> shell.  Try this:
>
> >>   echo $(( $(wc -l < sourcefile.txt) / 4 ))
>
> >> > I've tried numerous awk and sed statements, a la:
>
> >> > sed -e -n "$1,p" sourcefile.txt > list.1
> >> > sed -i "$1d" sourcefile.txt
> >> > sed $1q list.txt > list.1 & sed -i $1d sourcefile.txt
> >> > awk "{(FNR < $1); print}" sourcefile.txt > list.1
>
> >> Try this:
>
> >>   l=$(( $(wc -l < sourcefile.txt) / 4 ))
> >>   sed --in-place "1,${l}d" sourcefile.txt
>
> >> Bob
>
> > Dangit, those don't work as you expected with XP and GNUwin32
>
> > echo $(( $(wc -l < sourcefile.txt) / 4 ))
> > The system cannot find the file specified.
>
> > l=$(( $(wc -l < sourcefile.txt) / 4 ))
> > The system cannot find the file specified.
>
> > sed --in-place "1,${l}d" sourcefile.txt
> > sed: -e expression #1, char 7: extra characters after command
>
> To set the output of a command to a variable in cmd.exe, use this hack:
>
> for /f %i in ('"<command>"') do set VARIABLE=%i
>
> So for your problem, here's a solution:
>
> for /f %i in ('"wc -l sourcefile.txt"') do set /a LINES=%i/4
> split -l %LINES% sourcefile.txt
>
> "set /a" instructs cmd.exe to evaluate a numerical expression.
>
> PS. John, I replied to your original message in alt.msdos.batch.nt (I was  
> reading newsgroups in alphabetical order).
>
> --
> Kam-Hung Soh <a href="http://kamhungsoh.com/blog";>Software Salariman</a>


Thank you, Software Salariman! You're my hero du jour!
Now, my eight-cpu box can go run two processes each using the four
outfile-n.txt lists....


Here's the final module with a few tweaks, changing to GNUwin32
csplit   (which works)
from    split    (which returned this error)
  c:/progra~1/gnuwin32/bin/split.exe: invalid number
  Try 'c:/progra~1/gnuwin32/bin/split.exe --help' for more information

Hope someone else can use this:


rem SPLITTER.BAT which lives in C:\Program Files\GNUwin32\bin - that
directory is in the PATH
cls
echo y | time | grep current | awk '{print $5,$6,$7}' > start
echo off
c:
cd\
net use V: /delete
net use V: \\server.division.corporation.com\area\directory\subdir /
persistent:NO
REM obtain list of files with fully qualified UNC filenames and use
sed  to strip off leading blanks
attrib \\server.division.corporation.com\area\directory\subdir\*.* /s
| sed "s/^...........//" > rawlist.txt
REM Grep looks for lines with specific extensions and copies them to
LIST.TXT
grep "\.doc$" rawlist.txt  > list.txt
grep "\.docx$" rawlist.txt  >> list.txt
grep "\.wp$" rawlist.txt  >> list.txt
grep "\.rtf$" rawlist.txt  >> list.txt
grep "\.wpd$" rawlist.txt  >> list.txt
grep "\.wks$" rawlist.txt  >> list.txt
grep "\.txt$" rawlist.txt  >> list.txt
grep "\.odb$" rawlist.txt  >> list.txt
REM Count all the lines in list.txt, divide by 4 and put value in
LINES
for /f %%i in ('"wc -l list.txt"') do set /a LINES=%%i/4
REM use GNUwin32 to divide list.txt into four chunks, not splitting
lines
csplit -f outfile- -n 1 list.txt %LINES% {2}
REM rename divided list files because I don't grok CSPLIT's syntax for
how to do that
ren outfile-? outfile-?.txt
REM Show original list length, filtered list length and length of each
divided list file
wc -l rawlist.txt
wc -l list.txt
wc -l outfile-?.txt
REM Show start & end time for the run: 4.5 secs to find 3,000 files,
filter and divide
echo y | time | grep current | awk '{print $5,$6,$7}'
type start