Perl regexp substitution - multiple matches - regex

Friends,
need some help with substitution regex.
I have a string
;;;;;;;;;;;;;
and I need to replace it by
;\N;\N;\N;\N;\N;\N;\N;\N;\N;\N;\N;\N;
I tried
s/;;/;\\N/;/g
but it gives me
;\N;;\N;;\N;;\N;;\N;;\N;;
tried to fiddle with lookahead and lookbehind, but can't get it solved.

I wouldn't use a regex for this, and instead make use of split:
#!/usr/bin/env perl
use strict;
use warnings;
my $str = ';;;;;;;;;;;;;';
print join ( '\N', split ( //, $str ) );
Splitting on nulls, to get each character, and making use of the fact that join puts delimiters between characters. (So not before first, and not after last).
This gives:
;\N;\N;\N;\N;\N;\N;\N;\N;\N;\N;\N;\N;
Which I think matches your desired output?
As a oneliner, this would be:
perl -ne 'print join ( q{\N}, split // )'
Note - we need single quotes ' rather than double around the \N so it doesn't get interpolated.
If you need to handle variable content (e.g. not just ; ) you can add grep or map into the mix - I'd need some sample data to give you a useful answer there though.
I use this for infile edit, the regexp suits me better
Following on from that - perl is quite clever. It allows you to do in place editing (if that's what you're referring to) without needing to stick with regular expressions.
Traditionally you might do
perl -i.bak -p -e 's/something/somethingelse/g' somefile
What this is doing is expanding out that out into a loop:
LINE: while (defined($_ = <ARGV>)) {
s/someting/somethingelse/g;
}
continue {
die "-p destination: $!\n" unless print $_;
}
E.g. what it's actually doing is:
opening the file
iterating it by lines
transforming the line
printing the new line
And with -i that print is redirected to the new file name.
You don't have to restrict yourself to -p though - anything that generates output will work in this way - although bear in mind if it doesn't 'pass through' any lines that it doesn't modify (as a regular expression transform does) it'll lose data.
But you can definitely do:
perl -i.bak -ne 'print join ( q{\N}, split // )'
And inplace edit - but it'll trip over on lines that aren't just ;;;;; as your example.
So to avoid those:
perl -i.bak -ne 'if (m/;;;;/) { print join ( q{\N}, split // ) } else { print }'
Or perhaps more succinctly:
perl -i.bak -pe '$_ = join ( q{\N}, split // ) if m/;;;/'

Since you can't match twice the same character you approach doesn't work. To solve the problem you can only check the presence of a following ; with a lookahead (the second ; isn't a part of the match) :
s/;(?=;)/;\\N/g

Related

Regex (or bash), get pipes between quotes (perl)

Update: Please keep in mind is that regex is my only option.
Update 2: Actually, I can use a bash based solution as well.
Trying to replace the pipes(can be more than one) that are between double quotes with commas in perl regex
Example
continuer|"First, Name"|123|12412|10/21/2020|"3|7"||Yes|No|No|
Expected output (3 and 7 are separated by a comma)
continuer|"First, Name"|123|12412|10/21/2020|"3,7"||Yes|No|No|
There may be more digits, it may not be just the two d\|d. It could be "3|7|2" and the correct output has to be "3,7,2" for that one. I've tried the following
cat <filename> | perl -pi -e 's/"\d+\|[\|\d]+/\d+,[\|\d]+/g'
but it just puts the actual string of d+ etc...
I'd really appreciate your help. ty
If it must be a regex here is a simpler one
perl -wpe's/("[^"]+")/ $1 =~ s{\|}{,}gr /eg' file
Not bullet-proof but it should work for the shown use case.†
Explanation. With /e modifier the replacement side is evaluated as code. There, a regex runs on $1 under /r so that the original ($1) is unchanged; $N are read-only and so we can't change $1 and thus couldn't run a "normal" s/// on it. With this modifier the changed string is returned, or the original if there were no changes. Just as ordered.
Once it's tested well enough add -i to change the input file "in-place" if wanted.
I must add, I see no reason that at least this part of the job can't be done using a CSV parser...
Thanks to ikegami for an improved version
perl -wpe's/"[^"]+"/ $& =~ tr{|}{,}r /eg' file
It's simpler, with no need to capture, and tr is faster
† Tested with strings like in the question, extended only as far as this
con|"F, N"|12|10/21|"3|7"||Yes|"2||4|12"|"a|b"|No|""|end|
I'd use a CSV parser, not regular expressions:
#!/usr/bin/env perl
use warnings;
use strict;
use Text::CSV_XS;
my $csv = Text::CSV_XS->new({ binary => 1, sep_char => "|"});
while (my $row = $csv->getline(*ARGV)) {
#$row = map { tr/|/,/r } #$row;
$csv->say(*STDOUT, $row);
}
example:
$ perl demo.pl input.txt
continuer|"First, Name"|123|12412|10/21/2020|3,7||Yes|No|No|
More verbose, but also more robust and a lot easier to understand.
If you cannot install modules, Text::ParseWords is a core module you can try. It can split a string and handle quoted delimiters.
use Text::ParseWords;
my $q = q(continuer|"First, Name"|123|12412|10/21/2020|"3|7"||Yes|No|No|);
print join "|", map { tr/|/,/; $_ } quotewords('\|', 1, $q);
As a one-liner, it would be:
perl -MText::ParseWords -pe'$_ = join "|", map { tr/|/,/; $_ } quotewords('\|', 1, $_);' yourfile.txt
You said Update 2: Actually, I can use a bash based solution as well. and while this script isn't bash you could call it from bash (or any other shell) which I assume is what you really mean by "bash based" so - this will work using any awk in any shell in every Unix box:
$ awk 'BEGIN{FS=OFS="\""} {for (i=2; i<=NF; i+=2) gsub(/\|/,",",$i)} 1' file
continuer|"First, Name"|123|12412|10/21/2020|"3,7"||Yes|No|No|
Imagine yourself having to debug or enhance the clear, simple loop above above vs the regexp incantation you posted in your answer:
's/(?:(?<=")|\G(?!^))(\s*[^"|\s]+(?:\s+[^"|\s]+)*)\s*\|\s*(?=[^"]*")/$1,/g'
Remember - Some people, when confronted with a problem, think "I know, I'll use regular expressions." Now they have two problems..
I'm sure you could do what I'm doing with awk above natively in perl instead if you're trying to modify a perl script to add this functionality.
I'd use Text::CSV_XS.
perl -MText::CSV_XS=csv -e'
csv
in => \*ARGV,
sep_char => "|",
on_in => sub { tr/|/,/ for #{ $_[1] } };
'
You can provide the file name as an argument or provide the data via STDIN.
This is working right now
's/(?:(?<=")|\G(?!^))(\s*[^"|\s]+(?:\s+[^"|\s]+)*)\s*\|\s*(?=[^"]*")/$1,/g'
Credit goes to my boss at work
Thanks everyone for looking.
I hope some of you realize that some projects require certain ways and complicating an already very complicated pre existing structure is not always an option at work. I knew there would be a one liner for this, do not hate because you did not like that.

Repeating regex pattern

I have a string such as this
word <gl>aaa</gl> word <gl>aaa-bbb=ccc</gl>
where, if there is one ore more words enclosed in tags. In those instances where there are more than one words (which are usually separated by - or = and potentially other non-word characters), I'd like to make sure that the tags enclose each word individually so that the resulting string would be:
word <gl>aaa</gl> word <gl>aaa</gl>-<gl>bbb</gl>=<gl>ccc</gl>
So I'm trying to come up with a regex that would find any number of iterations of \W*?(\w+) and then enclose each word individually with the tags. And ideally I'd have this as a one-liner that I can execute from the command line with perl, like so:
perl -pe 's///g;' in out
This is how far I've gotten after a lot of trial and error and googling - I'm not a programmer :( ... :
/<gl>\W*?(\w+)\W*?((\w+)\W*?){0,10}<\/gl>/
It finds the first and last word (aaa and ccc). Now, how can I make it repeat the operation and find other words if present? And then how to get the replacement? Any hints on how to do this or where I can find further information would be much appreciated?
EDIT:
This is part of a workflow that does some other transformations within a shell script:
#!/bin/sh
perl -pe '#
s/replace/me/g;
s/replace/me/g;
' $1 > tmp
... some other commands ...
This needs a mini nested-parser and I'd recommend a script, as easier to maintain
use warnings;
use strict;
use feature 'say';
my $str = q(word <gl>aaa</gl> word <gl>aaa-bbb=ccc</gl>);
my $tag_re = qr{(<[^>]+>) (.+?) (</[^>]+>)}x; # / (stop markup highlighter)
$str =~ s{$tag_re}{
my ($o, $t, $c) = ($1, $2, $3); # open (tag), text, close (tag)
$t =~ s/(\w+)/$o$1$c/g;
$t;
}ge;
say $str;
The regex gives us its built-in "parsing," where words that don't match the $tag_re are unchanged. Once the $tag_re is matched, it is processed as required inside the replacement side. The /e modifier makes the replacement side be evaluated as code.
One way to provide input for a script is via command-line arguments, available in #ARGV global array in the script. For the use indicated in the question's "Edit" replace the hardcoded
my $str = q(...);
with
my $str = shift #ARGV; # first argument on the command line
and then use that script in your shell script as
#!/bin/sh
...
script.pl $1 > output_file
where $1 is the shell variable as shown in the "Edit" to the question.
In a one-liner
echo "word <gl>aaa</gl> word <gl>aaa-bbb=ccc</gl>" |
perl -wpe'
s{(<[^>]+>) (.+?) (</[^>]+>)}
{($o,$t,$c)=($1,$2,$3);$t=~s/(\w+)/$o$1$c/g; $t}gex;
'
what in your shell script becomes echo $1 | perl -wpe'...' > output_file. Or you can change the code to read from #ARGV and drop the -n switch, and add a print
#!/bin/sh
...
perl -wE'$_=shift; ...; say' $1 > output_file
where ... in one-liner indicate the same code as above, and say is now needed since we don't have the -p with which the $_ is printed out once it's processed.
The shift takes an element off of an array's front and returns it. Without an argument it does that to #ARGV when outside a subroutine, as here (inside a subroutine its default target is #_).
This will do it:
s/(\w+)([\-=])(?=\w+)/$1<\/gl>$2<gl>/g;
The /g at the end is the repeat and stands for "global". It will pick up matching at the end of the previous match and keep matching until it doesn't match anymore, so we have to be careful about where the match ends. That's what the (?=...) is for. It's a "followed by pattern" that tells the repeat to not include it as part of "where you left off" in the previous match. That way, it picks up where it left off by re-matching the second "word".
The s/ at the beginning is a substitution, so the command would be something like:
cat in | perl -pne 's/(\w+)([\-=])(?=\w+)/$1<\/gl>$2<gl>/g;$_' > out
You need the $_ at the end because the result of the global substitution is the number of substitutions made.
This will only match one line. If your pattern spans multiple lines, you'll need some fancier code. It also assumes the XML is correct and that there are no words surrounding dashes or equals signs outside of tags. To account for this would necessitate an extra pattern match in a loop to pull out the values surrounded by gl tags so that you can do your substitution on just those portions, like:
my $e = $in;
while($in =~ /(.*?<gl>)(.*?)(?=<\/gl>)/g){
my $p = $1;
my $s = $2;
print($p);
$s =~ s/(\w+)([\-=])(?=\w+)/$1<\/gl>$2<gl>/g;
print($s);
$e = $'; # ' (stop markup highlighter)
}
print($e);
You'd have to write your own surrounding loop to read STDIN and put the lines read in into $in. (You would also need to not use -p or -n flags to the perl interpreter since you're reading the input and printing the output manually.) The while loop above however grabs everything inside the gl tags and then performs your substitution on just that content. It prints everything occurring between the last match (or the beginning of the string) and before the current match ($p) and saves everything after in $e which gets printed after the last match outside the loop.

Replace strings only within a regex match in perl

I have an XML document with text in attribute values. I can't change how the the XML file is generated, but need to extract the attribute values without loosing \r\n. The XML parser of course strips them out.
So I'm trying to replace \r\n in attribute values with entity references
I'm using perl to do this because of it's non-greedy matching. But I need help getting the replace to happen only within the match. Or I need an easier way to do this :)
Here's is what I have so far:
perl -i -pe 'BEGIN{undef $/;} s/m_description="(.*?)"/m_description="$1"/smg' tmp.xml
This matches what I need to work with: (.*?). But I don't know to expand that pattern to match \r\n inside it, and do the replacement in the results. If I knew how many \r\n I have I could do it, but it seems I need a variable number of capture groups or something like that? There's a lot to regex I don't understand and it seems like there should be something do do this.
Example:
preceding lines
stuff m_description="Over
any number
of lines" other stuff
more lines
Should go to:
preceding lines
stuff m_description="Over
any number
of lines" other stuff
more lines
Solution
Thanks to Ikegam and ysth for the solution I used, which for 5.14+ is:
perl -i -0777 -pe's/m_description="\K(.*?)(?=")/ $1 =~ s!\n!
!gr =~ s!\r!
!gr /sge' tmp.xml
. should already match \n (because you specify the /s flag) and \r.
To do the replacement in the results, use /e:
perl -i -0777 -pe's/(?<=m_description=")(.*?)(?=")/ my $replacement=$1; $replacement=~s!\n!
!g; $replacement=~s!\r!
!g; $replacement /sge' tmp.xml
I've also changed it to use lookbehind/lookahead to make the code simpler and to use -0777 to set $/ to slurp mode and to remove the useless /m.
OK, so whilst this looks like an XML problem, it isn't. The XML problem is the person generating it. You should probably give them a prod with a rolled up copy of the spec as your first port of call for "fixing" this.
But failing that - I'd do a two pass approach, where I read the text, find all the 'blobs' that match a description, and then replace them all.
Something like this:
#!/usr/bin/env perl
use strict;
use warnings;
use Data::Dumper;
my $text = do { local $/ ; <DATA> };
#filter text for 'description' text:
my #matches = $text =~ m{m_description=\"([^\"]+)\"}gms;
print Dumper \#matches;
#Generate a search-and-replace hash
my %replace = map { $_ => s/[\r\n]+/
/gr } #matches;
print Dumper \%replace;
#turn the keys of that hash into a search regex
my $search = join ( "|", keys %replace );
$search = qr/\"($search)\"/ms;
print "Using search regex: $search\n";
#search and replace text block
$text =~ s/m_description=$search/m_description="$replace{$1}"/mgs;
print "New text:\n";
print $text;
__DATA__
preceding lines
stuff m_description="Over
any number
of lines" other stuff
more lines

Adding quotes to a CSV using perl

I've got a CSV that looks as follows:
A,01,ALPHA
00,D,CHARLIE
E,F,02
This is the desired file after transformation:
"A",01,"ALPHA"
00,"D","CHARLIE"
"E","F",02
As you can see, the fields that are entirely numeric are left unquoted, whilst the alpha (or alphanumeric ones) are quoted.
What would be a sensible way to go about this in Perl ?
Already commented below, but I've tried stuff like
perl -pe 's/(\w+)/"$1"/g'
And that doesn't work because \w obviously picks up the numerics.
I recommend not reinventing the wheel, but rather to use an already existing module, as zdim recommends. Here is your example using Text::CSV_XS
test.pl
#!/usr/bin/env perl
use warnings;
use strict;
use Text::CSV_XS;
use Scalar::Util qw( looks_like_number );
my $csv = Text::CSV_XS->new();
while (my $row = $csv->getline(*STDIN)) {
my #quoted_row = map { looks_like_number($_) ? $_ : '"'. $_ .'"' } #$row;
print join(',',#quoted_row) . "\n";
}
Output
cat input | perl test.pl
"A",01,"ALPHA"
00,"D","CHARLIE"
"E","F",02
Another one-liner, input file modified to add a line with alphanumeric fields
$ cat ip.csv
A,01,ALPHA
00,D,CHARLIE
E,F,02
23,AB12,53C
$ perl -F, -lane 's/.*[^0-9].*/"$&"/ foreach(#F); print join ",", #F' ip.csv
"A",01,"ALPHA"
00,"D","CHARLIE"
"E","F",02
23,"AB12","53C"
To modify OP's attempt:
$ perl -pe 's/(^|,)\K\d+(?=,|$)(*SKIP)(*F)|\w+/"$&"/g' ip.csv
"A",01,"ALPHA"
00,"D","CHARLIE"
"E","F",02
23,"AB12","53C"
(^|,)\K\d+(?=,|$)(*SKIP)(*F) this will skip the fields with digits alone and the alternate pattern \w+ will get replaced
It seems that you are after a one-liner. Here is a basic one
perl -lpe '$_ = join ",", map /^\d+$/ ? $_ : "\"$_\"", split ",";' input.csv
Splits each line by , and passes obtained list to map. There each element is tested for digits-only /^\d+$/ and passed untouched, or padded with " otherwise. Then map's return is joined by ,.
The -l removes newline, what is needed since " pad the whole line. The result is assigned back to $_ in order to be able to use -p so that there is no need for explicit print.
The code is very easily used in a script, if you don't insist on an one-liner.
Processing of csv files is far better done by modules, for example Text::CSV

Using string variables containing literal escapes in a Perl substitution

I'm new to Perl and I found behaviour which I don't understand and can't solve.
I'm making a small find and replace program and there are some things I need to do. I have bunch of files that I need to process. Then I have a list of find / replace rules in an external text file. In replacing there I need three special things:
Replacing utf-8 characters (Czech diacritics)
Work with adding/removing lines (so working in a slurp mode)
Use a regular expressions
I want a program that works alone, so I wrote it so that it takes three arguments:
The file to work on
What to find
What to replace.
I'm sending parameters in a loop from a bash script which parse the rules list and loads other files.
My problem is when I have a "\n" string in a rules list and I send it to the Perl script. If it's in the first part of replacement (in the find section) it looks for a newline correctly, but when it's in the second part (the replace section) it just prints \n instead of a newline.
I tried hardcoding "\n" to the string right into the variable instead of passing it from the list and then it works fine.
What's the reason Perl doesn't interpret the "\n" string there, and how can I make it work?
This is my code:
list.txt - One line from the external replacement list
1\. ?\\n?NÁZEV PŘÍPRAVKU;\\n<<K1>> NÁZEV PŘÍPRAVKU;
farkapitoly.sh - The bash script for parsing list.txt and cycling through all of the files and calling the Perl script
...
FILE="/home/tmp.txt"
while read LINE
do
FIND=`echo "$LINE" | awk -F $';' 'BEGIN {OFS = FS} {print $1}'`
REPLACE=`echo "$LINE" | awk -F $';' 'BEGIN {OFS = FS} {print $2}'`
perl -CA ./pathtiny.pl "$FILE" "$FIND" "$REPLACE"
done < list.txt
...
pathtiny.pl - The Perl script for find and replace
#!/usr/bin/perl
use strict;
use warnings;
use Modern::Perl;
use utf8; # Enable typing Unicode in Perl strings
use open qw(:std :utf8); # Enable Unicode to STDIN/OUT/ERR and filehandles
use Path::Tiny;
my $file = path("$ARGV[0]");
my $searchStr = "$ARGV[1]";
my $replaceStr = "$ARGV[2]";
# $replaceStr="\n<<K1>> NÁZEV PRÍPRAVKU"; # if I hardcode it here \n is replaced right away
print("Search String:", "$searchStr", "\n");
print("Replace String:", "$replaceStr", "\n\n");
my $guts = $file->slurp_utf8;
$guts =~ s/$searchStr/$replaceStr/gi;
$file->spew_utf8($guts);
If it's important, I'm using Linux Mint 13 64-bit on VirtualBox (under Win 8.1) and I have Perl v5.14.2. Every file is UTF-8 with Linux endings.
Example files can be found on pastebin. this should end up like this.
But examples varies a lot. I need a universal solution to write down newline in a replacement string so it replaces correctly.
The problem is that the replacement string is read literally from the file, so if your file contains
xx\ny
then you will read exactly those six characters. Also, the replacement part of a substitution is evaluated as if it was in double quotes. So your replacement string is "$replaceStr" which interpolates the variable and goes no further, so you will again have xx\nyy in the new string. (By the way, please avoid using capital letters in local Perl identifiers as in practice they are reserved for globals such as Module::Names.)
The answer lies in using eval, or its equivalent - the /e modifier on the substitution.
If I write
my $str = '<b>';
my $r = 'xx\ny';
$str =~ s/b/$r/;
then the replacement string is interpolated to xx\ny, as you have experienced.
A single /e modifier evaluates the replacement as an expression instead of just a double-quoted string, but of course $r as an expression is xx\ny again.
What you need is a second /e modifier, which does the same evaluation as a single /e and then does an additional eval of the result on top. For this it is cleanest if you use qq{ .. } as you need two levels of quotation.
If you write
$str =~ s/b/qq{"$r"}/ee
then perl will evaluate qq{"$r"} as an expression, giving "xx\nyy", which, when evaluated again will give you the string you need - the same as the expression 'xx' . "\n" . 'yy'.
Here's a full program
use strict;
use warnings;
my $s = '<b>';
my $r = 'xx\nyy';
$s =~ s/b/qq{"$r"}/ee;
print $s;
output
<xx
yy>
But don't forget that, if your replacement string contains any double quotes, like this
my $r = 'xx\n"yy"'
then they must be escaped before putting the through the substitution as the expression itself also uses double quotes.
All of this is quite hard to grasp, so you may prefer the String::Escape module which has an unbackslash function that will change a literal \n (and any other escapes) within a string to its equivalent character "\n". It's not a core module so you probably will need to install it.
The advantage is that you no longer need a double evaluation, as the replacement string can be just unbackslash $r which give the right result if it evaluated as an expression. It also handles double quotes in $r without any problem, as the expression doesn't use double quotes itself.
The code using String::Escape goes like this
use strict;
use warnings;
use String::Escape 'unbackslash';
my $s = '<b>';
my $r = 'xx\nyy';
$s =~ s/b/unbackslash $r/e;
print $s;
and the output is identical to that of the previous code.
Update
Here is a refactoring of your original program that uses String::Escape. I have removed Path::Tiny as I believe it is best to use Perl's built-in inplace-edit extension, which is documented under the General Variables section of perlvar.
#!/usr/bin/perl
use utf8;
use strict;
use warnings;
use 5.010;
use open qw/ :std :utf8 /;
use String::Escape qw/ unbackslash /;
our #ARGV;
my ($file, $search, $replace) = #ARGV;
print "Search String: $search\n";
print "Replace String: $replace\n\n";
#ARGV = ($file);
$^I = '';
while (<>) {
s/$search/unbackslash $replace/eg;
print;
}
You got \n as a content of a string. (as two chacters 1: \ and second n, and not as one newline.
Perl interprets the \n as newline when it is as literal (e.g. it is in your code).
The quick-fix would be:
my $replaceStr=eval qq("$ARGV[2]"); #evaling a string causes interpreting the \n as literal
or, if you don't like eval, you can use the String-Escape cpan module. (the unbackslash function)
You're wanting a literal string to be treated as if it were a double quoted string. To do that you'll have to translate any backslash followed by another character.
The other experts have shown you how to do that over the entire string (which is risky since it uses eval with unvalidated data). Alternatively, you could use a module, String::Escape, which requires an install (not a high bar, but too high for some).
However, the following does a translation of the return value string itself in a safe way, and then it can be used like a normal value in your other search and replace:
use strict;
use warnings;
my $r = 'xx\nyy';
$r =~ s/(\\.)/qq{"$1"}/eeg; # Translate \. as a double quoted string would
print $r;
Outputs:
xx
yy