regs-parser/main.py

import re
import sys
import json

ADDRESS_PATTERN = r'0x[0-9A-F]+'

def parse_table_entries(fields):
    """Parse aligned table entries"""
    pattern = r'\s+'.join([f"(?P<{v[:v.find(':')]}>{v[v.find(':') + 1:].strip()})"
               for v in fields])
    pattern = re.compile(pattern)
    fields = [v[:v.find(':')] for v in fields]
    prev = None
    entries = []
    def flush():
        nonlocal prev
        if prev is not None:
            entries.append(prev['data'])
            prev = None

    while True:
        line = yield
        if line == None:
            break
        line = line.rstrip()
        if len(line.strip()) == 0:
            flush()
            continue
        match = pattern.fullmatch(line)
        if match is not None:
            flush()
            starts = [match.span(v)[0] for v in fields]
            prev = {
                'data': {v: [match.group(v)] for v in fields},
                'starts': starts
            }
        elif prev is not None:
            for i, field in enumerate(fields):
                if i + 1 == len(fields):
                    v = line[prev['starts'][i]:].strip()
                else:
                    v = line[prev['starts'][i]:prev['starts'][i + 1]].strip()
                if len(v) == 0:
                    continue
                prev['data'][field].append(v)
    flush()
    return entries

def parse_register_fields():
    fields = [
        r'name: \w+',
        r'bits: \d+(:\d+)?',
        r'type: \w+',
        f'reset: {ADDRESS_PATTERN}',
        r'description: .+'
    ]
    it = parse_table_entries(fields)
    next(it)
    return it

def end_iterator(it):
    try:
        it.send(None)
    except StopIteration as e:
        return e.value

def parse_registers():
    def two_column(width):
        # only require the leftmost column starts with a \w, and the a space
        # separator between the two columns...
        fields = [
            f'key: \\w.{{1,{width-2}}}\\s',
            r'value: \w.+'
        ]
        it = parse_table_entries(fields)
        next(it)
        return it

    def inner():
        def_start = re.compile(r'Name\s+(\w+)\s+')
        field_start = re.compile(r'\s+Field Name\s+Bits\s+Type\s+Reset '
                                 r'Value\s+Description\s+')

        state = 0
        it = None
        results = []
        while True:
            line = yield
            if line == None:
                break
            line = line
            if state == 0:
                m = def_start.fullmatch(line)
                if m is not None:
                    if it is not None:
                        results[-1]['fields']= end_iterator(it)
                    it = two_column(m.span(1)[0])
                    it.send(line)
                    state = 1
                else:
                    if it is not None:
                        it.send(line)
            elif state == 1:
                m = field_start.fullmatch(line)
                if m is not None:
                    if it is not None:
                        results.append({'def': end_iterator(it)})
                    it = parse_register_fields()
                    state = 0
                else:
                    if it is not None:
                        it.send(line)

        if it is not None:
            if state == 0:
                results[-1]['fields']= end_iterator(it)
        return results
    it = inner()
    next(it)
    return it

def interpret_field(field):
    bit_pattern = re.compile(r'(\d+)(:(\d+))?')
    m = bit_pattern.fullmatch(field['bits'][0])
    if m is None:
        raise ValueError(field['bits'])
    high = int(m.group(1))
    low = high
    if m.group(3) is not None:
        low = int(m.group(3))
    result = {}
    result['name'] = field['name'][0]
    result['bits'] = [low, high]
    result['type'] = field['type'][0]
    return result

def interpret(reg):
    pattern = re.compile(r'(.+\w)\d+')
    width_pattern = re.compile(r'(\d+)\s?bits')
    result = {}
    name_pattern = None
    expected = [
        ['Name', 'name', lambda v: ''.join(v)],
        ['Relative Address', 'rel', lambda v: int(v[0], 16)],
        ['Absolute Address', 'abs', lambda v: int(v[0], 16)],
        ['Width', 'width', lambda v: int(width_pattern.fullmatch(v[0]).group(1))],
        ['Access Type', 'access', lambda v: v[0]],
        ['Reset Value', 'reset', lambda v: v[0]],
        ['Description', 'description', lambda v: ' '.join(v)]
    ]
    for v in reg['def']:
        a = v['key']
        b = v['value']
        key = a[0].strip()
        if len(expected) > 0 and key == expected[0][0]:
            result[expected[0][1]] = expected[0][2](b)
            expected = expected[1:]
            if key == 'Name':
                m = pattern.fullmatch(b[0].strip())
                if m is not None:
                    name_pattern = m.group(1)
        elif name_pattern != None and key.startswith(name_pattern):
            if 'similar' not in result:
                result['similar'] = []
            result['similar'].append({'name': key, 'abs': int(b[0], 16)})
    result['fields'] = [interpret_field(f) for f in reg['fields']]
    assert len(expected) == 0
    return result

def snake_to_camel(name: str):
    result = []
    start = True
    for c in name:
        if c == '_':
            start = True
        elif start:
            start = False
            result.append(c.upper())
        else:
            result.append(c.lower())
    return ''.join(result)

def emit_rust(base_addr, ending_addr, registers):
    current_addr = base_addr
    reserved_id = 0
    code = []
    for reg in registers:
        addr = int(reg['address'], 16)
        if addr > ending_addr:
            break
        if addr < base_addr:
            continue
        padding = addr - current_addr
        if padding > 0:
            if padding % 4 == 0:
                code.append(f'unused{reserved_id}: [RO<u32>; {padding // 4}],')
            else:
                code.append(f'unused{reserved_id}: [RO<u8>; {padding}],')
            reserved_id += 1
        access = ''
        unknown = False
        if reg['type'] == 'ro':
            access = 'RO'
        elif reg['type'] == 'wo':
            access = 'WO'
        elif reg['type'] in ['rw', 'mixed']:
            access = 'RW'
        else:
            access = reg['type']
            unknown = True
        size = int(reg['size'])
        if size not in [8, 16, 32]:
            unknown = True
        current_addr += padding + size // 8
        line = f"pub {reg['id'].lower()}: {access}<u{size}>,"
        if unknown:
            line = '// FIXME: ' + line
        code.append(f"/// {reg['description']}")
        code.append(line)
    return code

parser = parse_registers()
for line in sys.stdin:
    parser.send(line)
v = end_iterator(parser)
for reg in v:
    reg = interpret(reg)
    print(reg)