I would like to find all the .html files in a folder and append [file](./file.html) to another file called index.md. I tried the following command:

ls | awk "/\.html$/" | xargs -0 -I @@ -L 1 sh -c 'echo "[${@@%.*}](./@@)" >> index.md'

But it can't substitute @@ inside the command? What am I doing wrong?

Note: Filename can contain valid characters like space


index.md would have each line with [file](./file.html) where file is the actual file name in the folder

  • xargs -0 implies null-terminated strings on the xargs stdin, but awk does not print them. ${} needs a variable name. Both points are addressed in @RoVo's answer
    – weirdan
    Sep 3 '18 at 11:24
  • 1
    Would you please clarify how the content of "index.md" will look like?
    – user88036
    Sep 3 '18 at 11:32
  • @Goro I had appended the clarification at the end of question, but unfortunately, it has been edited out!
    – Porcupine
    Sep 3 '18 at 11:39
  • @Nikhil. Would you please include it again. Thanks!
    – user88036
    Sep 3 '18 at 11:40
  • @Goro Isn't it appropriate to justify accepted answer?
    – Porcupine
    Sep 3 '18 at 14:37

Just do:

for f in *.html; do printf '%s\n' "[${f%.*}](./$f)"; done > index.md

Use set -o nullglob (zsh, yash) or shopt -s nullglob (bash) for *.html to expand to nothing instead of *.html (or report an error in zsh) when there's no html file. With zsh, you can also use *.html(N) or in ksh93 ~(N)*.html.

Or with one printf call with zsh:

printf '[%s](./%s)\n' ${basenames:^files} > index.md

Note that, depending on which markdown syntax you're using, you may have to HTML-encode the title part and URI-encode the URI part if the file names contain some problematic characters. Not doing so could even end up introducing a form of XSS vulnerability depending on context. With ksh93, you can do it with:

for f in *.html; do
  title=${ printf %H "${file%.*}"; }
  uri=${ printf '%#H' "$file"; }
  printf '%s\n' "[$title]($uri)"
done > index.md

Where %H¹ does the HTML encoding and %#H the URI encoding, but we still need to address newline characters separately.

Or with perl:

perl -MURI::Encode=uri_encode -MHTML::Entities -CLSA -le '
  for (<*.html>) {
     $uri = uri_encode("./$_");
     $_ = encode_entities $_;
     print "[$_]($uri)"

Using <br/> for newline characters. You may want to use ␤ instead or more generally decide on some form of alternative representation for non-printable characters.

There are a few things wrong in your code:

  • parsing the output of ls
  • use a $ meant to be literal inside double quotes
  • Using awk for something that grep can do (not wrong per se, but overkill)
  • use xargs -0 when the input is not NUL-delimited
  • -I conflicts with -L 1. -L 1 is to run one command per line of input but with each word in the line passed as separate arguments, while -I @@ runs one command for each line of input with the full line (minus the trailing blanks, and quoting still processed) used to replace @@.
  • using {} inside the code argument of sh (command injection vulnerability)
  • In sh, the var in ${var%.*} is a variable name, it won't work with arbitrary text.
  • use echo for arbitrary data.

If you wanted to use xargs -0, you'd need something like:

printf '%s\0' * | grep -z '\.html$' | xargs -r0 sh -c '
  for file do
    printf "%s\n" "[${file%.*}](./$file)"
  done' sh > file.md
  • Replacing ls with printf '%s\0' * to get a NUL-delimited output
  • awk with grep -z (GNU extension) to process that NUL-delimited output
  • xargs -r0 (GNU extensions) without any -n/-L/-I, because while we're at spawning a sh, we might as well have it process as many files as possible
  • have xargs pass the words as extra arguments to sh (which become the positional parameters inside the inline code), not inside the code argument.
  • which means we can more easily store them in variables (here with for file do which loops over the positional parameters by default) so we can use the ${param%pattern} parameter expansion operator.
  • use printf instead of echo.

It goes without saying that it makes little sense to use that instead of doing that for loop directly over the *.html files like in the top example.

¹ It doesn't seem to work properly for multibyte characters in my version of ksh93 though (ksh93u+ on a GNU system)

  • That overwrites index.md though, which OP's code did not.
    – weirdan
    Sep 3 '18 at 11:27
  • 2
    I think this is still what OP wants. OP uses >> because he uses it inside the loop, while this answer after the loop and a second run of the same script doesn't make too much sense to me.
    – pLumo
    Sep 3 '18 at 11:28
  • @StéphaneChazelas Thanks for the answer. But for f in *.html; do printf '%s\n' "[${f%.*}](./$f)"; done >> index.md appends [*](./*.html) when no html file exists.
    – Porcupine
    Sep 3 '18 at 12:48
  • 1
    @Nikhil, see edit. Sep 3 '18 at 13:02

Do not parse ls.
You don't need xargs for this, you can use find -exec.

try this,

find . -maxdepth 1 -type f -name "*.html" -exec \
    sh -c 'f=$(basename "$1"); echo "[${f%.*}]($1)" >> index.md' sh {} \;

If you want to use xargs, use this very similar version:

find . -maxdepth 1 -type f -name "*.html" -print0 | \
    xargs -0 -I{} sh -c 'f=$(basename "$1"); echo "[${f%.*}]($1)" >> index.md' sh {} \;

Another way without running xargs or -exec:

find . -maxdepth 1 -type f -name "*.html" -printf '[%f](./%f)\n' \
    | sed 's/\.html\]/]/' \
    > index.md
  • Is that an extra sh argument in the first command, or is that intentional? Sep 3 '18 at 15:26
  • 2
    This is taken from this answer. See comments there and man sh -> -c for a documentation why this is needed.
    – pLumo
    Sep 3 '18 at 15:27
  • 1
    Ah, thanks - I had missed that If there are arguments after the command_string, the first argument is assigned to $0 and any remaining arguments are assigned to the positional parameters. Sep 3 '18 at 15:40
  • 1
    Add '-type f' to avoid strangeness with directories matching "*.html"
    – abligh
    Sep 3 '18 at 17:29

Do you really need xargs?

ls *.html | perl -pe 's/.html\n//;$_="[$_](./$_.html)\n"'

(If you have more than 100000 files):

printf "%s\n" *.html | perl -pe 's/.html\n//;$_="[$_](./$_.html)\n"'

or (slower, but shorter):

for f in *.html; do echo "[${f%.*}](./$f)"; done
  • Note that with ls *.html, if any of those html files are of type directory, ls will list their content. More generally, when you use ls with a shell wildcard, you want to use ls -d -- *.html (which also addresses the issues with file names starting with -). Sep 4 '18 at 7:18
  • The first two approaches assume file names don't contain newline characters (anyway, I suppose those would have to be encoded somehow in the markdown syntax). The third one assumes file names don't contain backslash characters. More generally, echo can't be used for arbitrary data. Sep 4 '18 at 7:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.