"bad flag in substitute command: '{" while replacing a string with sed from one file to another file

Question

I am trying to replace strings found in File1 with strings in File2

File1

<IMG SRC="/Repository/GetImage.dll?baseHref=Orange/2011/03/27&amp;EntityID=Ad12911&amp;imgExtension=" />
<IMG SRC="/Repository/GetImage.dll?baseHref=Orange/2011/03/20&amp;EntityID=Ad13304&amp;imgExtension=" />
<IMG SRC="/Repository/GetImage.dll?baseHref=Orange/2010/08/29&amp;EntityID=Ad13724&amp;imgExtension=" />

File2

/getimage.dll?path=Orange/2011/03/27/129/Img/Ad1291103.gif
/getimage.dll?path=Orange/2011/03/20/133/Img/Ad1330402.gif
/getimage.dll?path=Orange/2010/08/29/137/Img/Ad1372408.gif

When I run this command

$ sed -e 's/.*SRC="\/Repository\([^"]*\)".*/\1/p{r File1' -e 'd}' File2

I get this error

sed: 1: "s/.*SRC="\/Repository\( ...": bad flag in substitute command: '{'

Is there something wrong with my regex?

The result I am trying to achieve would be to have File1 look like:

File1

<IMG SRC="/Repository/getimage.dll?path=Orange/2011/03/27/129/Img/Ad1291103.gif" />
<IMG SRC="/Repository/getimage.dll?path=Orange/2011/03/20/133/Img/Ad1330402.gif" />
<IMG SRC="/Repository/getimage.dll?path=Orange/2010/08/29/137/Img/Ad1372408.gif" />

Answer 1

If you are trying to replace in File1 everything inside double quotes with new image names taken from File2 then I would use awk:

awk -F'"' 'NR==FNR{a[i++]=$1;next}{print $1 FS a[j++] FS $3}' File2 File1

The output is the following:

<IMG SRC="/getimage.dll?path=Orange/2011/03/27/129/Img/Ad1291103.gif" />
<IMG SRC="/getimage.dll?path=Orange/2011/03/20/133/Img/Ad1330402.gif" />
<IMG SRC="/getimage.dll?path=Orange/2010/08/29/137/Img/Ad1372408.gif" />

Answer 2

I have no idea what you're trying to do there but then my sed-fu is not that strong so I guess you're using some arcane syntax I am unaware of. Since I can't tell you what's wrong with your sed (but an educated guess is that the special characters contained in your replacement strings (/,? etc) are causing problems), I will instead offer a perl alternative:

perl -i -pe 'BEGIN{open($f,shift); while(<$f>){chomp; push @F,$_}}
            $k=shift(@F); s/(.*SRC=.)([^"]*)/$1$k/' file2 file1 

Here's the same thing written as a commented script to make it clearer. In the one-liner above, the -i causes the actual input file to be changed, just like sed -i.

#!/usr/bin/env perl

## This is the equivalent of the BEGIN{} block.
## @ARGV is the array of arguments and shift returns
## the first element of it. This is file2 which is
## then opened, each line is read, its trailing \n
## is removed by chomp and it is then added to the @F array.
my $file=shift(@ARGV);
open($f,$file);
while(<$f>){chomp; push @F,$_}

## This is the rest of the oneliner above. The -pe options
## cause the file to be read and each line printed after 
## the script is applied. Since the previous block removed 
## file2 from @ARGV, this is applied to file1 only.
while (<>) {
    ## Remove the 1st item of @F. This is a line of file2.
    $k=shift(@F);

    ## Make the substitution. The \ before the " is not 
    ## needed, I just added it here because otherwise, the 
    ## syntax highlighting is broken. 
    s/(.*SRC=.)([^\"]*)/$1$k/;
    ## This print is implied by the -p flag
    print;
}

Answer 3

The error is telling you that your sed command is wrong, not your regexp. You need a newline or semicolon to separate the s command from the following { command. Equivalently, you can put them in separate -e arguments.

sed -e 's/.SRC="/Repository([^"])".*/\1/p' -e '{' -e 'r File1' -e 'd' -e '}' File2

This won't do what you want, though. It removes the prefix …SRC="Repository/ and the part starting at the next double quote from the input, printing only lines that have been replaced (because of the p flag on the s command, and the following d), and it inserts a copy of File1 for each input line (matched or not).

If you want to match data from the two files, you'll need a more powerful tool than sed¹. Awk or Perl are good choices.

¹ _{Technically, sed is Turing complete, but doing that in sed would be extremely complex and obscure.}