diagnose-large-types.py

import subprocess
import glob
import re

def type_lines(ll_files):
    for ll_file in ll_files:
        for line in open(ll_file):
            if line.startswith('%'):
                yield line

PATTERN = re.compile(r'->|([^<>, ]+)|<|>|,| ')
def parse_type(type_name):
    stack = [[]]
    for match in PATTERN.finditer(type_name):
        part = match.group(0)
        if part == '<':
            stack.append([stack[-1].pop()])
        elif part == '>':
            top = stack.pop()
            if not stack:
                print(match)
                print(len(type_name))
            stack[-1].append(top)
        else:
            stack[-1].append(part)
    assert len(stack) == 1
    return stack[0]

def fl(t):
    if type(t) is str:
        return len(t)
    else:
        return sum(fl(x) for x in t)

def process_type(cache, t):
    if type(t) is str:
        if t in '() ->,!':
            return t
        processed = t
    else:
        if len(t) == 1:
            return process_type(cache, t[0])
        else:
            processed = t[0] + '<' + ''.join(process_type(cache, x) for x in t[1:]) + '>'

    try:
        return cache[processed]
    except KeyError:
        ident = str(len(cache) + 1)
        cache[processed] = ident
        print(ident, processed, '(', fl(t), ')')
        return ident





def main():
    subprocess.run(['cargo', 'rustc', '--', '--emit', 'llvm-ir'], check=True)

    ll_files= glob.glob('target/debug/deps/*.ll')

    longest_type = max(type_lines(ll_files), key=len)
    actual_type = longest_type[longest_type.find('"')+1:longest_type.rfind('" = type')]
    parsed = parse_type(actual_type)

    cache = {}
    print(process_type(cache, parsed))

if __name__ == '__main__':
    main()