Translate Part of a Line - linux

Question

I have a bunch of files that I am moving from one wiki (Markdown based) to another (Creole based). I have written a couple of sed scripts to things such as convert link formats and header formats. But the new wiki allows a directory structure and I would rather use that than the pseudo-directory structure I have now. I have already renamed the files, but I need to convert all of the links from _ delimited to / delimited.
Basic info:
Creole link: [[url]] [[url|name]]
I only want to convert the links that do not contain a . or a /.
I would really appreciate if you explained what the command you give means so that I can learn from it.
Sample
this is a line with a [[Link_to_something]] and [[Something_else|something else]]
this site is cool [[http://example.com/this_page]]
to
this is a line with a [[Link/to/something]] and [[Something/else|something else]]
this site is cool [[http://example.com/this_page]]
What I have tried
y/// only works on the whole line.
s//\u\2 only supports case translations.

Answer 1

I think I'd use Perl. It can be done as a one-liner, thus:
perl -pe 's{\[\[([^/.|]+)(|[^]]+)?\]\]}{$x=$1;$y=$2;$x=~s%_%/%g;"[[$x$y]]"}gex;' <<'EOF'
this is a line with a [[Link_to_something]] and [[Something_else|something else]]
this site is cool [[http://example.com/this_page]]
EOF
The output from that is:
this is a line with a [[Link/to/something]] and [[Something/else|something else]]
this site is cool [[http://example.com/this_page]]
Whether that's good style etc is entirely open to debate.
I'll explain this version of the code, which is isomorphic with the code above:
perl -e 'use strict; use warnings;
while (my $line = <>)
{
$line =~ s{ \[\[ ([^/.|]+) (|[^]]+)? \]\] }
{ my($x, $y) = ($1, $2); $x =~ s%_%/%g; "[[$x$y]]" }gex;
print $line;
} '
The while loop is basically what the -p provides in the first version. I've explicitly named the input variable as $line instead of using the implicit $_ as in the first version. I also had to declare $x and $y because of the use strict; use warnings;.
The substitute command takes the form s{pattern}{replace} because there are slashes in the regexes themselves. The x modifier allows (non-significant) spaces in the two parts, which makes it easier to lay out. The g modifier repeats the substitution as often as the pattern matches. The e modifier says 'treat the right-hand part of the substitution as an expression'.
The matching pattern looks for a pair of open square brackets, then remembers a sequence of characters other than /, . or |, optionally followed by a | and a sequence of characters other than ], finishing at a pair of close square brackets. The two captures are $1 and $2.
The replacement expression saves the values of $1 and $2 in variables $x and $y. It then applies a simpler substitution to $x, changing underscores into slashes. Then the result value is the string of [[$x$y]]. You can't modify $1 or $2 directly in the replacement expression. And the inner s%_%/%g; clobbers $1 and $2, which is why I needed $x and $y.
There might be another way to do it - this is Perl, so TMTOWTDI: there's more than one way to do it. But this does at least work.

Answer 2

This might work for you:
awk -vORS='' -vRS='[[][[][^].]*[]][]]' '{gsub(/_/,"/",RT);print $0 RT}' file
this is a line with a [[Link/to/something]] and [[Something/else|something else]]
this site is cool [[http://example.com/this_page]]
Set the output record separator to null
Set the record separator to [[...]] (where the ... does not contain a ..
Replace all _'s in what is placed in the record separator variable RT with /'s
Print the concatenated record and record separator. i.e. $0 RT
This is a sed solution:
sed 's/\[\[[^].]*]]/\a\n&\a\n/g' file |
sed '/^\[\[[^]]*\]\]\a/y/_/\//;H;$!d;g;s/\a\n//g;s/.//'
this is a line with a [[Link/to/something]] and [[Something/else|something else]]
this site is cool [[http://example.com/this_page]]
Surround [[...]]'s by \a\n's N.B. \a is chosen as an unlikely character to appear in the file.
Translate '_'s to /'s in lines beginning with [[
Remove all occurrences of \a\n's
If you have GNU sed, this will do:
sed '/\[\[[^].]*]]/{s||'\''$(sed "y/_/\\//" <<<"&")'\''|g;s/.*/echo '\''&'\''/}' file
this is a line with a [[Link/to/something]] and [[Something/else|something else]]
this site is cool [[http://example.com/this_page]]

Answer 3

You can use python to simplify the regex:
$ python3 -c '
> import re
> import sys
> for line in sys.stdin:
> print(re.sub(r"\[\[(?!http).*?\]\]", lambda m:m.group(0).replace("_", "/"), line), end="")
> ' <input.txt
this is a line with a [[Link/to/something]] and [[Something/else|something else]]
this site is cool [[http://example.com/this_page]]
Note: $ and > at the beginning of lines are command prompt.
You can also do it in vim visually:
/\[\[\(http\)\#!.\{-}\]\]
:%s##\=substitute(submatch(0), '_', '/', '')#g