lint-readme: insert semicolons more discerningly

[Avaq]

The script (see below) that rewrites code blocks for linting should be smarter about inserting semicolons, so that the linter will not complain about misplaced semicolons.

sanctuary-scripts/bin/lint-readme

Lines 16 to 41 in ce8733d

	rewrite() {
	local state=
	local line
	while IFS='' read -r line ; do
	local a="${line:0:1}"
	local b="${line:1:1}"
	local z="${line: -1}"
	if [[ $a$b == '> ' ]] ; then
	printf '\n%s' "${line:2}"
	state=input
	elif [[ -n $state && $a$b == '. ' ]] ; then
	printf '\n%s' "${line:2}"
	elif [[ $state == input && $a$z == '{}' ]] ; then
	printf ';\n%s' "($line)"
	state=output
	elif [[ $state == input ]] ; then
	printf ';\n%s' "$line"
	state=output
	elif [[ $state == output ]] ; then
	printf ';\n%s' "$line"
	state=
	else
	printf '\n%s' "$line"
	fi
	done
	}

See Avaq/permissionary#5 for discussion leading up to this.

[gabejohnson]

Why do we have a script to rewrite code blocks? Is this to work around limitations of ESLint?

[Avaq]

@gabejohnson Because codeblocks are written for https://github.com/davidchambers/doctest with additional information prefixed to every line in the form of >, ., and ! characters.

[gabejohnson]

Ahhh...those code blocks. Thanks for the explanation Aldwin 😄