topic Re: sed, ed, (hurts my head) in Operating System - HP-UX

sed, ed, (hurts my head)

Ron Bombard — Tue, 06 Aug 2002 16:25:34 GMT

Greetings!

I have a large text file and was wondering what the easiest way would be to replace "whitespace" between fields (words) with "tabs". is this possible? Each field in the file starts at a specific character position, but was hoping there was an easy way of just replacing whitespace with a tab.

Ideas? Suggestions? Flames?
Thanks.

Re: sed, ed, (hurts my head)

Jean-Luc Oudart — Tue, 06 Aug 2002 16:31:43 GMT

Well not sed but tr may do the trick :
cat | tr " " "\t" >

ET voila

Jean-Luc

Re: sed, ed, (hurts my head)

Pete Randall — Tue, 06 Aug 2002 16:37:42 GMT

Ron,

Here's a couple of handy references to sed and awk commands, originally (to me, at least) courtesy of Princess Paula.

Pete

Re: sed, ed, (hurts my head)

Pete Randall — Tue, 06 Aug 2002 16:38:14 GMT

And here's the awk one.

Pete

Re: sed, ed, (hurts my head)

Thierry Poels_1 — Tue, 06 Aug 2002 17:04:10 GMT

Hi,

how about "unexpand" ???
see "man unexpand"

good luck,
Thierry.

Re: sed, ed, (hurts my head)

Ron Bombard — Tue, 06 Aug 2002 17:19:01 GMT

I think I should have thought this through before posting... The "unexpand" program works good, and the "tr" suggestions also works. But.... all it takes is a blank field to screw it up. If it runs into a blank field, instead of 2 tabs, there will be one.

I'm trying to load a text file into MySQL, and it has to be tab delimited. And the original text file's field delimination is by ASCII Character Position.

Me thinks this is going to be complicated ;)

Re: sed, ed, (hurts my head)

Hai Nguyen_1 — Tue, 06 Aug 2002 17:23:32 GMT

Ron,

Give us one line before and after the format. Then we may help you better.

Hai

Re: sed, ed, (hurts my head)

Tom Maloy — Tue, 06 Aug 2002 17:34:39 GMT

If I understand you correctly, you want to replace each single blank with one tab character.

cat file | sed -e "s/ /\t/g"

Note that \t should be replaced by the tab character.

The "g" does the global replace.

Tom

Re: sed, ed, (hurts my head)

Ron Bombard — Tue, 06 Aug 2002 17:42:25 GMT

Ok, here's the before file. I have 2 lines in it to show what happens when a field is missing. The first line is a complete line.

Re: sed, ed, (hurts my head)

Ron Bombard — Tue, 06 Aug 2002 17:42:56 GMT

Ok, here's the after unexpand file. I have 2 lines in it to show what happens when a field is missing. The first line is a complete line.

Re: sed, ed, (hurts my head)

Ron Bombard — Tue, 06 Aug 2002 17:44:24 GMT

Problem with "unexpand" is... there has to be more than 2 spaces together to convert to tab. Some of my fields are only 2 space apart. So, I ran the "unexpand-ed" file thru a "tr" filter and here's what I got

Re: sed, ed, (hurts my head)

Vincent Fleming — Tue, 06 Aug 2002 18:17:34 GMT

I think it might be easier for you with "cut", which is good at getting things at character positions...

while read LINE
do
firstfield=`echo $LINE | cut -c 2-10`
secondfield=`echo $LINE | cur -c 11-20`

printf( "%s\t%s\n", $firstfield, $secondfield)

done < inputfile

Make a "firstfield" definition for each field in the record.

Personally, I think this would probably be easier to write in C (certainly much faster to execute), but I'm guessing you don't know C.

Good luck!

Re: sed, ed, (hurts my head)

Ron Bombard — Tue, 06 Aug 2002 18:40:34 GMT

Hey thanks for the "cut" idea. I figured I have to do something like that. Now... the "printf", I've never played with.

Here's my script:
=======================

while read LINE
do
time=`echo $LINE | cut -c 3-6`
duration=`echo $LINE | cut -c 8-11`
conncode=`echo $LINE | cut -c 12-13`
acccode=`echo $LINE | cut -c 15-17`
acccode2=`echo $LINE | cut -c 19-21`
dialed=`echo $INE | cut -c 23-37`
originate=`echo $LINE | cut -c 39-43`

printf( "%s\t%s\n", $time, $duration, $conncode, $acccode, $acccode2, $dialed, $originate)

done < inputfile
=========================

I ran piped the text file through this:

cat txtfile | converterprg

I get this error:

./converter: line 11: syntax error near unexpected token `"%s\t%s\n",'
./converter: line 11: `printf( "%s\t%s\n", $time, $duration, $conncode, $acccode, $acccode2, $dialed, $originate)'
I'm not sure what this means...
Thanks!

Re: sed, ed, (hurts my head)

Chris Lonergan — Tue, 06 Aug 2002 18:48:14 GMT

Use sed or ed it doesn't matter. They both have tagged regular expressions and can slpit a line thus:-

s/$...$ *$....$ *$.....$/\1\t\2\t\3\t/

This will take three fields of 3,4 and 5 characters respectively, delimited by any number of spaces (including none) and format the same three fields but seperated by a 'tab' character.

i.e. $...$ takes the first 3 characters and \1 places these in the output.

Change the ... to denote the characters that you require (for many characters use the notation .{m} where m is the number of characters. For whitespace use ['space'tab']* instead of space* or spacespace* which must match at least one space

Re: sed, ed, (hurts my head)

Vincent Fleming — Wed, 07 Aug 2002 12:03:17 GMT

Sorry - my syntax for the printf was incorrect. It's been a while. I gave the syntax for doing that in awk.

Anyway, with these variables you defined:

time=`echo $LINE | cut -c 3-6`
duration=`echo $LINE | cut -c 8-11`
conncode=`echo $LINE | cut -c 12-13`
acccode=`echo $LINE | cut -c 15-17`
acccode2=`echo $LINE | cut -c 19-21`
dialed=`echo $INE | cut -c 23-37`
originate=`echo $LINE | cut -c 39-43`

You should use this syntax (in ksh):

printf "%s\t%s\t%s\t%s\t%s\t%s\t%s\n" "$time" "$duration" "$conncode" "$acccode" "$acccode2" "$dialed" "$originate"

Good luck!