Convert only parts of a filename to uppercase

If the shell is bash, using just bash parameters expansion:

file="aaa.bbb.dat" name=${file%.*} # delete everything after last dot ext=${file##*.} # delete everything up to last dot upcase=${name^^*}.$ext # uppercase everything echo "$upcase" 

AAA.BBB.dat 

Trying with a more difficult case:

file="déjà vu . dat " name=${file%.*} # delete everything after last dot ext=${file##*.} # delete everything up to last dot upcase=${name^^*}.$ext # uppercase everything echo ":$upcase:" 

Gives:

:DÉJÀ VU . dat : 

So:

• double quotes aren't necessary, until using the result
• Uppercase seems OK even for non-ASCII characters

When there's an ambiguity in how far a group extends, regex engines favor the longest match first. For any file name, \(.*\) matches the whole name and \.*.* matches the empty string.

You'll need two cases: with or without extension. Note also that if a file name starts with a ., that's not the start of an extension.

I don't understand why you want to use expr. Shell parameter manipulation is easier.

On converting to uppercase, note that the tr implementation on Linux does not support non-ASCII locales. It only does byte manipulation. For example echo accentué | tr a-z A-Z results in ACCENTUé, not ACCENTUÉ. Use a locale-aware tool such as awk instead. In bash, you can use ${filename^^?}, but that's not available in sh. Make sure that your script is running in the correct locale for the file names' encoding.

I assume that the filename doesn't contain a directory part. If it does, separate it first.

case $filename in ?*.*) # There is an extension base="${filename%.*}"; ext=".${filename##*.}";; *) # No extension base="$filename"; ext="";; esac upcased_base="$(printf %s. %base | awk '$0 = toupper($0)')" upcased="${upcased_base%.}$ext" 

The trailing . in %s. that then gets stripped from $upcased_base ensures that the script correctly handles file names with a newline immediately before the extension. Without this, the command substitution would strip off trailing newlines. You don't need this if you've already ensured that your file names don't contain newline characters.

Here is an entirely awk-based solution, where you would put the following line in your shell script:

uppercasename="$(echo "$filename" | awk 'BEGIN{FS=OFS="."} NF==1{$1=toupper($1)} {for (i=1;i<NF;i++) $i=toupper($i)} 1')" 

This will use the . as field separator for input and output and, if only one field is found, convert that to uppercase, and in all other cases convert all but the last fields to uppercase. It then prints the result (this is the meaning of the 1, which is a shorthand notation for {print}).

If you are using bash, you could get rid of the pipe and state it as

uppercasename="$(awk 'BEGIN{FS=OFS="."} NF==1{$1=toupper($1)} {for (i=1;i<NF;i++) $i=toupper($i)} 1' <<< "$filename")" 

using a here-string.

Note that this is designed so that in the borderline case of a filename ending in a ., as in myfile.this.txt., it will treat that like an "empty but present suffix" and convert it to MYFILE.THIS.TXT.. Also, if the filename starts with a . and has no other extension (as in .myfile), it will keep that as lowercase.