Goal
– Oh no! The killer must have followed her on vacation! But to find them we'd have to search through 200MB of emails looking for something formatted like an address!
– It's hopeless!

– Everybody stand back.

You are given a raw email archive. You must count the number of likely email addresses in there.

As opposed to real life, a likely email address in this problem:
• is wrapped in angle brackets <>
• contains a unique at sign @
• left of the at sign, contains any non-empty combination of ASCII letters, digits, and symbols ! " # $ % & ' ( ) * + , - . / : ; = ? [ \ ] ^ _ ` { | } ~
• right of the at sign, contains any non-zero number of words separated by a dot .
 – those words must start with a letter, and may contain ASCII letters, digits, and a hyphen -
 – the last word must be letters only
 – if there's only one word, it must be succeded by a dot .
Input
A raw archive.
It's so raw you do not know in advance how long it is. Read all you can until end of stream.

(except for the site limit of 10k)
Output
An integer N, number of likely email addresses in the archive.
Constraints
The domain name need not be actually registered.
Example
Input
Valid:
- <steve@example.com>
- <john.doe@codingame.com>
- <un{ba[lan)cea'ble@xxx.>
Invalid:
- [(balanced)@example.com]
- <inigo montoya@buttercup.net>
- <bill@example>
- <timey!uucp!gw!ucb>
Output
3