Fixing POSIX sh warning in a small Bash program - bash

Question

I wrote the following code in Bash:
#!/bin/sh
host=$1
regex="^(((git|ssh|http(s)?)|(git#[\w\.]+))(:(\/\/)?)([A-Za-z0-9.#:_/-]+)\.com)(.*)"
if [[ "$host" =~ $regex ]]; then
d=${BASH_REMATCH[1]}
if [[ "$d" = *github* ]]; then
return
fi
fi
die "Current repository is not stored in Github."
I want to learn how to write a better Bash code so I use the shellcheck.net.
Line 5:
if [[ "$host" =~ $regex ]]; then
^-- SC2039: In POSIX sh, [[ ]] is undefined.
Line 6:
d=${BASH_REMATCH[1]}
^-- SC2039: In POSIX sh, array references are undefined.
Line 7:
if [[ "$d" = *github* ]]; then
^-- SC2039: In POSIX sh, [[ ]] is undefined.
I'm trying to understand how to fix those warnings. I understand that in order to fix [[ ]] I need it to switch to [ ] but then I get an error due globs. Also how should I replace the =~ operator?

Answer 1

When you write #!/bin/sh then you shouldn't use bash-specific features like [[. But you don't need to change [[ to [ or anything like that; just change the shebang line to #!/bin/bash. Then you can use all the bash features you like.

Answer 2

Use grep and sed in posix.
# use grep -q to match with regex
if printf "%s\n" "$host" | grep -q '\(git\|ssh\|http\(s\)\)etc. etc. etc.'; then
# use sed to extract part of the string matching regex
d=$(printf "%s\n" "$host" | sed 's/\(g\|ssh\|http\(s\)\)etc. etc. etc./\2/')
if printf "%s\n" "$d" | grep -q github; then
return
fi
fi
Finding out proper regexes is left to others.

Answer 3

You could try to parse out the different parts with parameter expansions though it's going to get a bit tedious. (The link is to the Bash manual; only a few of the expansions supported by Bash are POSIX.)
Assuming the input is a valid, well-formed URL (which may or may not be warranted) maybe try
host=$1
tail=${1#*://*/}
case $tail in "$host") tail=${host#*/};; esac
case ${host%/$tail} in
*github.com) return ;;
esac
die "Current repository is not stored in Github."
(where of course we assume that this is in a context where return makes sense, and where die is defined separately, like we have to assume in the original code).
This is quite a lot simpler than the regex you presented, and definitely does not cover all the strings that the regex would be able to handle; but perhaps it doesn't have to be all that complex if we can assume that the URL has gone through some sort of validation (i.e. if it's the output from git remote it's pretty safe to assume that the user has verified it by other means already).