[SOLVED] SED: replace all coincidences only from a position upto another

jmvidalvia · 02-05-2013, 08:27 AM

Hi,
I am getting crazy with an apparently simple problem.
As an example, need to change all "x" into "y" from the 5th position up tp the 10th.

Code:

Source: xbcb1xbx4b5x
Output: xbcb1yby4b5x

Thanks!

linosaurusroot · 02-05-2013, 09:11 AM

[/code]
#!/usr/bin/perl -w

$_=<>;
($first,$middle,$last)=/^(.{4})(.{6})(.*)/;
$middle =~ s/x/y/g;
printf("%s%s%s\n", $first,$middle,$last);
[/code]

danielbmartin · 02-05-2013, 09:33 AM

Quote:

Originally Posted by jmvidalvia

... need to change all "x" into "y" from the 5th position up to the 10th.

Code:

Source: xbcb1xbx4b5x
Output: xbcb1yby4b5x

Thanks!

Difficult to do with sed alone.
Easier with awk.

Code:

awk -F "" 'OFS="";{for (j=5;j<=10;j++) if ($j=="x") $j="y"}'1 <<< 'xbcb1xbx4b5x'

Daniel B. Martin

jmvidalvia · 02-05-2013, 10:06 AM

Quote:

Originally Posted by danielbmartin

Difficult to do with sed alone.
Easier with awk.

Code:

awk -F "" 'OFS="";{for (j=5;j<= 10;j++) if ($j=="x") $j="y"}'1

Daniel B. Martin

Thanks! Works great.
I am facing an unexpected need: same file but with variable fields tab separated instead of fixed width.
I can not use the lenght because of the variable size of every field, and need to "count" fields, in order to replace "x" for "y" in fields from #2 to #8.
Thanks!

PS: will mark as solved as the original question in fact is.

danielbmartin · 02-05-2013, 10:12 AM

Quote:

Originally Posted by jmvidalvia

I am facing an unexpected need: same file but with variable fields tab separated instead of fixed width.
I can not use the length because of the variable size of every field, and need to "count" fields, in order to replace "x" for "y" in fields from #2 to #8.

Please provide a sample input file and a corresponding output file. Tell which character separates fields (blank, tab, etc.).

Daniel B.Martin

danielbmartin · 02-05-2013, 10:40 AM

Quote:

Originally Posted by jmvidalvia

I am facing an unexpected need: same file but with variable fields tab separated instead of fixed width.
I can not use the length because of the variable size of every field, and need to "count" fields, in order to replace "x" for "y" in fields from #2 to #8.

To replace all "e" with "E" in blank-delimited fields 2 through 8, try this ...

Code:

awk -F" " '{for (j=2;j<=8;j++) gsub("e","E",$j)}'1 $Raven > $OutFile

This awk produces output like this ...

Code:

'Prophet!' said I, 'thing of Evil! - prophEt still, if bird or devil!
By that HEavEn that bEnds abovE us - by that God we both adore -
Tell this soul with sorrow ladEn if, within the distant Aidenn,
It shall clasp a saintEd maidEn whom thE angels named Lenore -
Clasp a rarE and radiant maidEn, whom thE angels named Lenore?'
Quoth thE ravEn, 'NEvErmorE.'

Daniel B. Martin

jmvidalvia · 02-05-2013, 10:46 AM

Quote:

Originally Posted by danielbmartin

Please provide a sample input file and a corresponding output file. Tell which character separates fields (blank, tab, etc.).

Daniel B.Martin

Sorry.
The sample input and output files would be something like:

Code:

input: abcd▸   exg▸  hxjk▸  lmxop▸  qrst¬
output: abcd▸   exg▸  hyjk▸  lmyop▸  qrst¬

(field separator is \t)

I found the following to work, but just for one field (#3 in the example).

Code:

awk -F'\t' -vOFS='\t' '{ gsub("x", "y", $3) ; print }' file > file1

Need to learn how to iterate the number of field to get "from one to another", 3 to 5 in the example.
Thanks,.

danielbmartin · 02-05-2013, 11:12 AM

Quote:

Originally Posted by jmvidalvia

Need to learn how to iterate the number of field to get "from one to another", 3 to 5 in the example.

Try this ...

Code:

awk 'OFS="\t" {for (j=3;j<=5;j++) gsub("x","Y",$j)}'1 $InFile > $OutFile

Daniel B. Martin

jmvidalvia · 02-05-2013, 01:29 PM

Quote:

Originally Posted by danielbmartin

Try this ...

Code:

awk 'OFS="\t" {for (j=3;j<=5;j++) gsub("x","Y",$j)}'1 $InFile > $OutFile

Daniel B. Martin

Perfect: thank you very much.

EDIT: had to add -F'\t' though

mina86 · 02-05-2013, 04:11 PM

This can be done in sed with a hold space. Here's what you need to do:

Copy pattern space to hold space.
Strip the first five characters.
Do the desired transformation.
Exchange pattern space and hold space.
Strip everything but the first five characters.
Append hold space to pattern space.
Replace new line characters with nothing.
You're done.

EDIT: Ah, if you are interested in only 5..10 characters than it's a bit different but again can easily be done with hold space as so:

Copy pattern to hold space.
Strip everything but characters 5-10.
Perform desired transformation.
Append hold space to pattern space.
Do s/$.*$\n$.....$.....\(.*)/\2\1\3/ (not tested and there may be more magic required if not all lines are guaranteed to be ≥ 10 characters).