[Solved] Problem

logar0 · 10-24-2010, 05:39 PM

I have the following one-liner:

echo "one two ΟΟΞ―Ξ± four" | sed -e 's/[ ]*[^ ]*Ξ[^ ]*Ξ[^ ]*/'"$(echo '&' | iconv -f UTF8 -t ISO_8859-7)"'/g'

sed, locates the word with the different encoding (ΟΟΞ―Ξ±). The output should be something like

one two three four

but it returns the same.

one two ΟΟΞ―Ξ± four

It's like the data in & from sed can't be changed.

I also tried passing & to an external script but it did work.

What am I doing something wrong?

Tinkster · 10-24-2010, 06:10 PM

The "&" you're using in your echo is from "outside" sed. The shell
doesn't know anything about sed's "&".

logar0 · 10-25-2010, 06:05 AM

Is there any way to make it work?

GrapefruiTgirl · 10-25-2010, 06:12 AM

I think the OP is trying to use "&" on the RHS the same way that "\&" is used in AWK, to represent the match on the LHS.
For this, I would try: put the whole LHS into a $...$ and then represent it with "\1" on the RHS. However, I'm still not sure whether or not the stuff represented by \1 will be passed into that $(subshell) as you expect-- but `xargs` could solve this, as well as probably other ways.

grail · 10-25-2010, 06:26 AM

Quote:

I think the OP is trying to use "&" on the RHS the same way that "\&" is used in AWK, to represent the match on the LHS.

Sed supports this too

GrapefruiTgirl · 10-25-2010, 06:29 AM

of course - thanks grail!

Well there's more of a problem here, at least on my machine: I tried echoing just the input string, through the `iconv` command, and it doesn't work for me, so if OP's in the same boat, the sed is never going to work.

When I do the echo of the string through `iconv` I get:

Code:

one two     four

I don't know if this is something to do with my locale or what...

logar0 · 10-25-2010, 06:54 AM

It seems that you can't edit the matched string inside sed.

Kenhelm · 10-25-2010, 06:56 AM

In GNU sed the 'e' command can execute commands in the pattern space.
For example

Code:

echo "2.2+3.3" | sed "s/.*/ echo '&'|bc /e"
5.5

echo "one two" | sed -r "s/(.*) (.*)/ echo '\2 \1'|rev /e"
eno owt

The s command creates a line of shell code in the pattern space which is then executed by the e command.
When the code is executed it is replaced in the pattern space by its output.
The 'e' command uses the default shell, /bin/sh, which is often linked to /bin/bash, but in Ubuntu it's now linked to /bin/dash.

grail · 10-25-2010, 07:17 AM

Well I get:

Quote:

��ία

So not much more useful than a blank.

@Kenhelm - thanks for that .. still getting sed-jitsu together

logar0 · 10-25-2010, 10:10 AM

Thanks you very much Guys.

grail · 10-25-2010, 11:41 AM

Please mark as SOLVED if you have your solution.