horrific but working way to summarize file sizes in a directory for evaluating what's tracked by git lfs

gistfile1.sh

$ find . -type f -name '*.*' |
  sed 's/.*\.//' |
  sort -u |
  while read ext; do
    (
      gfind . -name "*$ext" -printf "%s\n" |
        sed 's/$/+/' |
        tr -d '\n'
      echo 0
    ) | bc | tr -d '\n'
    echo -e "\t$ext"
  done

1003744	.exe
911328	.zip
806448	.dc
623008	.pck
564880	.mp4
497016	.d
351848	.efz
323112	.g
...

tr '[ \t\n]' ',' | sed -n -E -e 's!([^,]+),([^,]+),([^,]+),!\2%\1@!gp' | tr '%' '\t' | tr '@' '\n' | sort -rn is my way of working around sed's lack of... everything. The equivalent regex replace in a modern IDE is: s@^(\..*)\n(.*)\s+total$@$2\t$1

You can condense this part of your bash pipeline into a single sed command and avoid all the character translation (tr). Note, some extra complexity was required to avoid embedding a literal tab, which I know you hate. 😉

from

... | tr -d '[ \t]total' | tr '\n' ',' | sed -n -E -e 's!([^,]+),([^,]+),!\2%\1@!gp' | tr '%' '\t' | tr '@' '\n' | ...

to

... | sed -n -E "/^\\..+/ {
N
s/([^[:space:]]*)[[:space:]]*\\n([^[:space:]]*)[[:space:]]*total/\\2$(echo -ne '\t')\\1/
p
}" | ...

Credit for a WORKING solution to @kenthoward

Notes:

• Switch to iname if case should be insensitive for extensions
• But for Git LFS, case matters, so either lowercase all extensions, or make case insensitive LFS matchs (.[eE][xX][eE]), or:

- sort by size, find a cut off where files are to small to bother
- swap the file size with the file name using `s/(.*) (.*)/$2 $1/g`
- sort naturally
- visually find all related line endings and add each case to .gitattributes

One little bug which probably didn't affect your results. When simplifying the extension extraction to just sed 's/.*\.//' the gfind -name test should have been updated to include a literal period.

gfind . -name "*.$ext" -printf "%s\n"