How to list only the last name.partXXX.rar file?

With zsh:

$ files=( *.part<->.rar(Nn) ) $ typeset -A last $ for f ($files) last[$f:r:r]=$f $ print -roC1 -- $last 1yBWVnZCx8CoPrGIG.part23.rar DaHs0QJnJbt.part4.rar n5oTzoLvG.part6.rar okgbuh8VUxSguDNra9uMTtDlXhiLWQmY.part2.rar RSmWMPb0vWr8LIEFtR7o.part5.rar T7yvBIqHK82qDCNTtz9iuvp2NhQ.part11.rar tBJDjsyJtFpY0d3aQ.part6.rar W1Pn8SHf7pbMSf1u99C4f.part2.rar XYcUpv7b1ZpcczFT5y7Uc9mQTvAf88kl.part2.rar 

Explanations:

# Put a listing of all files that match the "glob" pattern # ANYTHING.partANYNUMBER.rar # into the variable called "files" # without erroring out if nothing matches (N), # and sorting things numerically (n) files=( *.part<->.rar(Nn) ) # make an associative array (-A) called "last" # (in Python, this would be called a dict, # in C++ a std::map<std::string,std::string>) typeset -A last # loop over all entries in "files", and for each strip the last file name suffix # (as separated by a "."), twice; i.e., remove the .rar, and remove the .partANYNUMBER. # Store the file name in last[twiceshortened] for f ($files) last[$f:r:r]=$f # Note that this overwrites the entry for 1yBWVnZCx8CoPrGIG until the highest- # numbered part is reached. # Print the full list: print -roC1 -- $last # -r : no fancy escaping (we don't need that and it makes things strange) # -o : print sorted, in ascending order # -C1 : print as 1 column # -- : the stuff to be printed follows after this # $last : print the content (not the keys!) of the "last" associative array # You get: 1yBWVnZCx8CoPrGIG.part23.rar DaHs0QJnJbt.part4.rar n5oTzoLvG.part6.rar okgbuh8VUxSguDNra9uMTtDlXhiLWQmY.part2.rar RSmWMPb0vWr8LIEFtR7o.part5.rar T7yvBIqHK82qDCNTtz9iuvp2NhQ.part11.rar tBJDjsyJtFpY0d3aQ.part6.rar W1Pn8SHf7pbMSf1u99C4f.part2.rar XYcUpv7b1ZpcczFT5y7Uc9mQTvAf88kl.part2.rar 

Using perl:

$ perl -MFile::Basename -e ' my @files = glob q(*.rar); my %bases; foreach (@files) { my ($b, $p) = split /\.part/, basename $_, q(.rar); $bases{$b} = $p if (!defined $bases{$b} or $p > $bases{$b}); }; foreach my $k (sort keys %bases) { print "$k.part$bases{$k}.rar\n" }' 1yBWVnZCx8CoPrGIG.part23.rar DaHs0QJnJbt.part4.rar RSmWMPb0vWr8LIEFtR7o.part5.rar T7yvBIqHK82qDCNTtz9iuvp2NhQ.part11.rar W1Pn8SHf7pbMSf1u99C4f.part2.rar XYcUpv7b1ZpcczFT5y7Uc9mQTvAf88kl.part2.rar n5oTzoLvG.part6.rar okgbuh8VUxSguDNra9uMTtDlXhiLWQmY.part2.rar tBJDjsyJtFpY0d3aQ.part6.rar 

First this gets all the filenames matching the glob *.rar into an array called @files. Then it converts the array into a hash (aka associative array) called %bases where each key is the base filename and the value is the highest part number seen for that basename.

Then it prints out each basename with the part number in the same format as the original filenames: base.partN.rar

This uses the File::Basename module, which is included with perl as part of its standard library.

There are shorter, more obfuscated ways to do this in perl, but this IMO is a nice balance between brevity and readability. It was written to pass the restrictions of use strict (or -Mstrict on the command-line), so it could be re-used as part of a larger script. That's why there are all the my declarations that are usually skipped for one-liners.

Using any awk:

$ ls -r1 | awk -F'.' '!seen[$1]++' tBJDjsyJtFpY0d3aQ.part6.rar okgbuh8VUxSguDNra9uMTtDlXhiLWQmY.part2.rar n5oTzoLvG.part6.rar XYcUpv7b1ZpcczFT5y7Uc9mQTvAf88kl.part2.rar W1Pn8SHf7pbMSf1u99C4f.part2.rar T7yvBIqHK82qDCNTtz9iuvp2NhQ.part11.rar RSmWMPb0vWr8LIEFtR7o.part5.rar DaHs0QJnJbt.part4.rar 1yBWVnZCx8CoPrGIG.part23.rar 

If you don't want to parse the output of ls then you could alternatively do this with any printf, sort and awk:

printf '%s\n' * | sort -t'.' -r -k1,1 -k2,2 | awk -F'.' '!seen[$1]++' 

That assumes your file names do not contain newlines, that the number of digits at the end of each part is consistent for each substring before part, and there is no . before part in any of the strings.

Using bash that has the loadable builtin kv (most probably 5.3+ version)

#!/usr/bin/env bash shopt -s nullglob enable kv || exit kv -A assoc -s . -d '' < <( printf '%s\0' *.rar ) for i in "${!assoc[@]}"; do printf '%s.%s\n' "$i" "${assoc["$i"]}" done 

Output

XYcUpv7b1ZpcczFT5y7Uc9mQTvAf88kl.part2.rar DaHs0QJnJbt.part4.rar n5oTzoLvG.part6.rar W1Pn8SHf7pbMSf1u99C4f.part2.rar okgbuh8VUxSguDNra9uMTtDlXhiLWQmY.part2.rar T7yvBIqHK82qDCNTtz9iuvp2NhQ.part11.rar RSmWMPb0vWr8LIEFtR7o.part5.rar 1yBWVnZCx8CoPrGIG.part23.rar 

According to help kv

 kv: kv [-A ARRAYNAME] [-s SEPARATORS] [-d RS] Read key-value pairs into an associative array. Read delimiter-terminated records composed of a single key-value pair from the standard input and add the key and corresponding value to the associative array ARRAYNAME. The key and value are separated by a sequence of one or more characters in SEPARATORS. Records are terminated by the first character of RS, similar to the read and mapfile builtins. If SEPARATORS is not supplied, $IFS is used to separate the keys and values. If RS is not supplied, newlines terminate records. If ARRAYNAME is not supplied, "KV" is the default array name. Returns success if at least one key-value pair is stored in ARRAYNAME. 

If sorted output is needed (from the OP's output), then one more loadable builtin named asort from: https://cgit.git.savannah.gnu.org/cgit/bash.git/plain/examples/loadables/asort.c

Something like:

#!/usr/bin/env bash shopt -s nullglob enable kv || exit enable asort || exit kv -A assoc -s . -d '' < <( printf '%s\0' *.rar ) keys=("${!assoc[@]}") asort keys for i in "${keys[@]}"; do printf '%s.%s\n' "$i" "${assoc["$i"]}" done 

Output

1yBWVnZCx8CoPrGIG.part23.rar DaHs0QJnJbt.part4.rar n5oTzoLvG.part6.rar okgbuh8VUxSguDNra9uMTtDlXhiLWQmY.part2.rar RSmWMPb0vWr8LIEFtR7o.part5.rar T7yvBIqHK82qDCNTtz9iuvp2NhQ.part11.rar tBJDjsyJtFpY0d3aQ.part6.rar W1Pn8SHf7pbMSf1u99C4f.part2.rar XYcUpv7b1ZpcczFT5y7Uc9mQTvAf88kl.part2.rar 

According to help asort

asort: asort [-nr] array ... or asort [-nr] -i dest source Sort arrays in-place. Options: -n compare according to string numerical value -r reverse the result of comparisons -i sort using indices/keys If -i is supplied, SOURCE is not sorted in-place, but the indices (or keys if associative) of SOURCE, after sorting it by its values, are placed as values in the indexed array DEST Associative arrays may not be sorted in-place. Exit status: Return value is zero unless an error happened (like invalid variable name or readonly array). 

NOTE: As per Stéphane Chazelas the separator must not be part of the filename, in this case a dot/period. Also the script works with the given file names by the OP.

Files like:

(In the future, it can be possible that there are files with the ending name.part001.rar , name.part002.rar , ... , name.part123.rar )

Cannot be parsed properly by the current script.

It can be done in bash:

{ unset files declare -A files for f in *.rar do if [[ "$f" =~ (.*\.part)([[:digit:]]+)\.rar ]] then if [[ -v "files[${BASH_REMATCH[1]}]" ]] then (( 10#${BASH_REMATCH[2]} > 10#${files[${BASH_REMATCH[1]}]} )) && files[${BASH_REMATCH[1]}]="${BASH_REMATCH[2]}" else files[${BASH_REMATCH[1]}]="${BASH_REMATCH[2]}" fi fi done for i in "${!files[@]}" do printf '%s\n' "$i${files[$i]}.rar" done } 

... outputs:

tBJDjsyJtFpY0d3aQ.part6.rar RSmWMPb0vWr8LIEFtR7o.part5.rar DaHs0QJnJbt.part4.rar W1Pn8SHf7pbMSf1u99C4f.part2.rar T7yvBIqHK82qDCNTtz9iuvp2NhQ.part11.rar n5oTzoLvG.part6.rar XYcUpv7b1ZpcczFT5y7Uc9mQTvAf88kl.part2.rar okgbuh8VUxSguDNra9uMTtDlXhiLWQmY.part2.rar 1yBWVnZCx8CoPrGIG.part23.rar 

The idea is to arithmetically compare the part numbers in identical filenames (identical except for the part number) and find which number is the greatest.

FOLDER="myarchivefolder"; ( \ cd "$FOLDER"; \ find . -type f -iname '*.rar' | \ cut -d '/' -f 2 | \ LC_ALL=C sort -t "." --key 1,1 --key 2.5n -r | \ sort --merge --unique -t "." --key 1,1 \ ) 

Explanation:

1. cd into the folder inside a subshell, to avoid unexpected dots in the file paths
2. Use find to list all files in current folder (add -maxdepth 1 to avoid descending into nested folders)
3. cut to remove leading ./ from find output
4. Sort file names
   • -t "." - dot . as field separator
   • --key 1,1 - sort by the first field, the archive name
   • --key 2.5n - and then numerical sort by part number (starting at the 5th character)
   • -r reverse, so the highest part number ends up as the first line
5. Only output one archive file for each sorted entry
   • sort --merge - skip sorting
   • --unique - subsequent archives are omitted as a duplicates
   • --key 1,1 - only consider archive name for deduplication

The good:

• avoids globbing: the for loop used in other answers may error with "Argument list too long" on large folders
• correctly handles 3-digit ".part025" numbers starting with zero
• It looks to be POSIX-compatible, if you replace long options with short ones.

The bad:

• Expects that the archive name itself will not contain dots (nor newlines). It would've been more robust to count dots from the end of a file path, but it is not possible with sort's key definition.
• Many process invocations and piping