Is there a reason you can't do this yourself?

You'll need to document this in a little more detail

The problem is was really simple, it's just a bit difficult to describe precisely and even more difficult to fix properly.

For every expression that matches, the script will would prefix the match with ANSI codes to change the color, then suffix it with another ANSI code to "reset" to the default color. If this overlaps with a previous match on the same line, the ANSI "reset" code will would remove any other color codes that should still have been in effect.

Example 1:
echo "aabbccddee" | em bbccdd red cc blue
would produce
aa[RED]bb[BLUE]cc[RESET]dd[RESET]ee
which is incorrect because the BLUE/RESET resets "dd" which should have been RED.

Example 2:
echo "aabbccddee" | em cc blue bbccdd red
would produce
aabb[BLUE]cc[RESET]ddee
because "bbBLUEccRESETdd" does not match "bbccdd".

I fear the only real solution would be to replace the simple regex substitute with a state machine that uses a color stack to see what color we should reset to. Unfortunately, this means we go from a quick and simple tool to a complicated piece of code that could fail even harder.

Instead of a simple substitution, I now keep an array of the codes I would insert. For each matching rule, I now scan this array to find the proper "reset color code" instead of just inserting a plain ANSI reset. Only when all the matching has finished I apply the codes using substr().