Suggested improvements for `dataurl`

[Rudxain]

macOS and Linux distros have base64 preinstalled, why not use it directly?

[mathiasbynens]

IIRC macOS didn't have base64 preinstalled when this was written.

[LucasLarson]

Why stop at base64? This can be done using only POSIX utilities!

# Create a data URL from a file
dataurl() {
  # https://github.com/mathiasbynens/dotfiles/commit/5d76fc286f

  # POSIX does not define `-b`/`--brief` or `--mime-type`
  mimeType="$(LC_ALL='C' file -b --mime-type -- "${1-}")"

  if printf -- '%s\n' "${mimeType-}" | grep -q -e '^text/'; then
    mimeType="${mimeType-}"';charset=utf-8'
  fi

  # `uuencode` – part of POSIX since the 1900s – instead of `openssl` or `base64`
  uuencode -m -- "${1-}" /dev/stdout |
    # remove the first line, then the last line, then remove all newlines
    sed -e '1d' -e '$d' - |
    sed -e ':a' -e 'N' -e '$! b a' -e 's/\n//g' - |
    awk -v mimeType="${mimeType-}" -- '{printf "data:%s;base64,%s\n", mimeType, $0}' -

  # `unset` instead of non-standard `local`, but send its errors to `/dev/null` just in case
  unset -v -- mimeType 2>/dev/null
}

[Rudxain]

IIRC macOS didn't have base64 preinstalled when this was written.

Thanks for explaining. I know little about macOS history

Why stop at base64? This can be done using only POSIX utilities! [...]

I like how you used -- to escape args, just-in-case a filename starts with "-". We should probably merge that change (--) into main

BTW, I tried using uuencode in my system, and it wasn't even installed, lol

[Rudxain]

BTW, speaking of special chars, I just noticed that both openssl base64 and base64, are not URL-safe. This completely defeats the purpose of dataurl.

@mathiasbynens Should I rename this issue to "dataurl suggestions", or should I open a new issue about this?

I'm willing to open a PR for this, but I don't know which commands are more appropriate. I'll do some research

[mathiasbynens]

@Rudxain What do you mean with URL-safe?

[Rudxain]

What do you mean with URL-safe?

I haven't tested the output of dataurl on a browser (neither address bar, nor HTML files), but I assume it can sometimes (depends on input data) cause issues because of the "/" chars

AFAIK, most base64 implementations on Unix-like OSes aren't concerned with being URL-safe, so most systems are susceptible to this problem

[Rudxain]

According to this SO answer, basenc --base64url is the best alternative (if we aren't concerned with retro-compatibility)

There were other answers that don't require basenc, but instead require cut, tr, awk, sed, etc... (some have more dependencies than others)

[Rudxain]

I just read Wikipedia: data URI scheme allows standard B64 (not "URL-safe"). I'm so confused lol. I'm sorry for the false alarm