Fix bashrc .complete completion behavior

This commit updates the completion function for the .functions (`..`,
`.2`, etc.) to behave as much like tab completion for `cd` as possible.
This means it has better listing for possible results when multiple
options are available, properly handles completed and nested directories
(!!) and is much more thoroughly documented so it's not such a dense
cluster of bash magic.

src/.bashrc

@@ -385,7 +385,15 @@ function _bashrc_.complete {
 		return 1
 	fi
 
-	local parent_depth path_array i parent_path word_list matched_word
+	COMPREPLY=()
+
+	# don't try to find matches if the search term starts with '/'
+	if [[ "${word:0:1}" == '/' ]]; then
+		return
+	fi
+
+	local parent_depth path_array i IGNORE_CASE search_path search_term \
+	      search_word matched_words matched_word
 	parent_depth="${cmd//.}"
 	if [ -z "$parent_depth" ]; then
 		parent_depth=1
@@ -394,18 +402,62 @@ function _bashrc_.complete {
 	for (( i=0; i<parent_depth; i++ )); do
 		path_array+=( '..' )
 	done
-	parent_path="$(_bashrc_join_by '/' "${path_array[@]}")"
 
-	COMPREPLY=()
-	word_list="$(
-		printf "%s\\n" "$parent_path"/*/ |
-		sed -e 's|^\(../\)*||' -e 's| |\\ |g'
-	)"
-	while IFS='' read -r matched_word; do
-		if [[ "$matched_word" =~ ^$word ]]; then
-			COMPREPLY+=( "$matched_word" )
+	# don't append a space to completed words
+	compopt -o nospace
+
+	# save configured value of readline's completion-ignore-case value to match
+	# behavior
+	IGNORE_CASE=( "$(
+		bind -V |
+		grep 'completion-ignore-case' |
+		grep -q "\`on'$" &&
+			echo '-i'
+	)" )
+
+	# search_path and search_term store the currently-searched path and
+	# remaining search term from the original search word, after traversing into
+	# directories. Eg. `.. foo/bar/baz`
+	search_path="$(_bashrc_join_by '/' "${path_array[@]}")"
+	search_term="$word"
+	while # do-while loop
+		# search_word refers to the string/directory to search for in the
+		# current directory, eg. `foo/`, `bar/`, or `baz`
+		search_word="$(sed -r -e 's|^([^/]+/?).*$|\1|' <<< "$search_term")"
+		# directories matching the search word
+		matched_words=()
+		while IFS='' read -r matched_word; do
+			# this loop is executed once with an empty string value even if
+			# `find` yields no results; it's not a valid entry, so skip it
+			if [[ -z "$matched_word" ]]; then
+				continue
+			fi
+
+			# add found directory as a match if the search word is present as a
+			# prefix of it
+			if grep -q -E "${IGNORE_CASE[@]}" "/${search_word}[^/]*/?$" <<< "${matched_word}/"; then
+				matched_words+=( "$matched_word/" )
+			fi
+		done <<< "$(find "$search_path" -mindepth 1 -maxdepth 1 -type d | sort -V)"
+
+		# continue searching through the next nested directory if only one
+		# result was found (assumed to be exact result) and the remaining search
+		# term has more directories to search through
+		if [[ "${#matched_words[@]}" == 1 && "$search_term" == */* ]]; then
+			search_path="${matched_words[0]}"
+			search_term="$(cut -d '/' -f 2- <<< "$search_term")"
+			continue
 		fi
-	done <<< "$word_list"
+
+		for matched_word in "${matched_words[@]}"; do
+			# transform each line in the following manner:
+			#  1. remove all leading `../` so the file or directory appears as
+			#     it would from the parent path
+			#  2. escape ` ` as would be needed to prevent argument splitting in
+			#     the shell
+			COMPREPLY+=( "$(sed -e 's|^\(../\)*||' -e 's| |\\ |g' <<< "$matched_word")" )
+		done
+	do break; done
 }
 
 # Prints color sequence according to the token type passed