"""What the generators of scripts/ (gen-options.sh, gen-limits.sh, gen-reference.sh, gen-messages.sh, gen-refusals.sh) share: reading C source, writing markdown cells, collecting errors.
They import it with `sys.path.insert(0, 'scripts')` from the repository root (they run as `python3 -I -`, which has no script directory)."""
import functools
import re
import sys

errors = []   # what a generator found wrong; exit_on_errors prints it


@functools.lru_cache(maxsize=None)   # the selftests of the generators parse the same sources many times
def strip_comments(text):
    """the text with every comment replaced by blanks (same offsets and lines); strings and character constants stay"""
    def blank(m):
        t = m.group(0)
        return t if t[0] in '"\'' else re.sub(r'[^\n]', ' ', t)
    return re.sub(r'"(?:\\.|[^"\\\n])*"|\'(?:\\.|[^\'\\\n])*\'|/\*.*?\*/|//[^\n]*', blank, text, flags=re.S)


def split_args(s):
    """the top level arguments of a C call's argument text: commas inside parentheses, braces and strings do not split"""
    out, depth, cur, i = [], 0, '', 0
    while i < len(s):
        c = s[i]
        if c == '"':
            j = i + 1
            while s[j] != '"':
                j += 2 if s[j] == '\\' else 1
            cur += s[i:j + 1]
            i = j + 1
            continue
        depth += (c in '({') - (c in ')}')
        if c == ',' and depth == 0:
            out.append(cur.strip())
            cur = ''
        else:
            cur += c
        i += 1
    return out + [cur.strip()]


def norm(s):
    return re.sub(r'\s+', ' ', s).strip()


def esc(s):
    return s.replace('|', '\\|')


def call_end(text, i):
    depth = 0
    while True:
        c = text[i]
        if c == '"':
            i += 1
            while text[i] != '"':
                i += 2 if text[i] == '\\' else 1
        elif c == '(':
            depth += 1
        elif c == ')':
            depth -= 1
            if depth == 0:
                return i
        i += 1


@functools.lru_cache(maxsize=None)
def testing_lines(text):
    """-> the set of line numbers (1-based) inside #ifdef SIFT_TESTING (its #else branch is outside)"""
    stack, out = [], set()
    for n, line in enumerate(text.split('\n'), 1):
        m = re.match(r'\s*#\s*(ifdef|ifndef|if|elif|else|endif)\b\s*(.*)', line)
        if m:
            d, rest = m.groups()
            if d in ('ifdef', 'ifndef', 'if'):
                stack.append((d == 'ifdef' and rest.strip() == 'SIFT_TESTING') or (d == 'ifndef' and rest.strip() == 'SIFT_TESTING' and 'neg'))
            elif d == 'else' and stack:
                stack[-1] = {True: False, False: False, 'neg': True}[stack[-1]]
            elif d == 'endif' and stack:
                stack.pop()
        if any(stack):
            out.add(n)
    return out


def err(msg):
    errors.append(msg)


def read(path):
    with open(path, encoding='utf-8') as f:
        return f.read()


def exit_on_errors(prefix=''):
    """print every error to stderr, one line each, and exit 1"""
    if errors:
        for e in errors:
            print(prefix + e, file=sys.stderr)
        sys.exit(1)