One of the most important products of the R1 company is a popular @r1.com mail service. The R1 mailboxes receive and send millions of emails every day.

Today, the online news thundered with terrible information. The R1 database crashed and almost no data could be saved except for one big string. The developers assume that the string contains the letters of some users of the R1 mail. Recovering letters is a tedious mostly manual work. So before you start this process, it was decided to estimate the difficulty of recovering. Namely, we need to calculate the number of different substrings of the saved string that form correct e-mail addresses.

We assume that valid addresses are only the e-mail addresses which meet the following criteria:

• the address should begin with a non-empty sequence of letters, numbers, characters '_', starting with a letter;
• then must go character '@';
• then must go a non-empty sequence of letters or numbers;
• then must go character '.';
• the address must end with a non-empty sequence of letters.

You got lucky again and the job was entrusted to you! Please note that the substring is several consecutive characters in a string. Two substrings, one consisting of the characters of the string with numbers l₁, l₁ + 1, l₁ + 2, ..., r₁ and the other one consisting of the characters of the string with numbers l₂, l₂ + 1, l₂ + 2, ..., r₂, are considered distinct if l₁ ≠ l₂ or r₁ ≠ r₂.

The first and the only line contains the sequence of characters s₁s₂... s_n (1 ≤ n ≤ 10⁶) — the saved string. It is guaranteed that the given string contains only small English letters, digits and characters '.', '_', '@'.

Print in a single line the number of substrings that are valid e-mail addresses.