parser.py

def tokenize(string):

    # Regular expression to extract blocks
    _recipient = re.compile(r'^(.+?)\s(in|to|about)\b',
                            re.IGNORECASE)
    _time = re.compile(r'\bin\s(\d+\s\b.+?\b)(?:\sto\b|\sabout\b|$)',
                            re.IGNORECASE)
    _message = re.compile(r'\b(?:to|about)\s(.+?)(?:\sin\b|$)',
                            re.IGNORECASE)

    recipientmatch = _recipient.search(string)
    timematch = _time.search(string)
    messagematch = _message.search(string)

    if recipientmatch is None:
        raise ReminderSyntaxError("Remind who?")

    if timematch is None:
        raise ReminderSyntaxError("Remind when?")

    if messagematch is None:
        raise ReminderSyntaxError("Remind what..?")

    # Fetch groups, process further and pack as tokens

    # Construct rough, unvalidated list of usernames, using
    # "and" and "," as delimiters to split on. The further
    # tokenization simplifies the parser's operation.
    recipientdelims = re.compile(r'\b,\sand\b\s|\b,\s|\s\band\b\s',
                                    re.IGNORECASE)
    recipientgroups = recipientmatch.group(1)
    recipientblock = recipientdelims.split(recipientgroups)

    # Time is just split in unavlidated elements. Should contain two,
    # the numerical value and the human readable time unit.
    timegroup = timematch.group(1)
    timeblock = timegroup.split()

    # message block is just kept as-is, as a string.
    # It never really gets processed anyways.
    messageblock = messagematch.group(1)

    return {"recipient": recipientblock,
        "time": timeblock, "message": messageblock }