Talking About Numbers

The programmers of a TTS (text-to-speech) program found that their product reads numbers poorly. It announces each digit in turn, so the listener loses track after a few digits.

They want a more natural reading. One of them suggested converting numeric strings into English text, for example turning "$1023$" into "one thousand and twenty three", so the speech engine can read the result with no further changes.

Write a program that converts non-negative integers into conventional English wording.

• The numbers $0$ to $19$ have unique names: "zero", "one", "two", "three", "four", "five", "six", "seven", "eight", "nine", "ten", "eleven", "twelve", "thirteen", "fourteen", "fifteen", "sixteen", "seventeen", "eighteen", "nineteen".
• The following multiples of $10$ are "twenty", "thirty", "forty", "fifty", "sixty", "seventy", "eighty", "ninety".
• A multiple of ten combined with a digit $1$ to $9$ is hyphenated, for example $31 \Rightarrow$ "thirty-one" and $77 \Rightarrow$ "seventy-seven".
• Multiples of $100$ use the digit name followed by "hundred", and any non-zero remainder after them is joined with "and", for example $200 \Rightarrow$ "two hundred", $412 \Rightarrow$ "four hundred and twelve", and $777 \Rightarrow$ "seven hundred and seventy-seven".
• Thousands and millions are named with the rules above for numbers $1$ to $999$. A non-zero remainder after them is separated by a comma, for example $1\,253\,101 \Rightarrow$ "one million, two hundred and fifty-three thousand, one hundred and one".
• If a number with a non-empty thousands or millions part is followed by a remainder of $1$ to $99$, the parts are joined with "and" instead of a comma, for example $1\,000\,011 \Rightarrow$ "one million and eleven" and $20\,222\,043 \Rightarrow$ "twenty million, two hundred and twenty-two thousand and forty-three".

The input has one or more datasets. Each dataset is one line containing a non-negative integer from $0$ to $999\,999\,999$. The input has no commas, and positive numbers have no leading zeros.

A line with a negative value ends the input.

For each dataset, print the spelled-out number on its own line.

The output must be left-justified and all letters are lowercase. Exactly one blank separates adjacent words, except around a hyphen or comma. A comma is followed by exactly one blank. A hyphen has no blank on either side.