IBM is implementing an online pharmacy application for a major drug store chain. As part of the application, customers can start to type in the name of a drug and the system will allow them to choose from among the list of drugs which begins with the letters that they have typed. For example, if they type the letters "ASP", the system could offer "ASPRIN" as a possible match, along with other drugs that begin with "ASP". If there's no match, the system will show <NONE>. Assumptions: The list may be very long, hundreds of thousands of drugs. The call will be made very often, so search speed is crucial. Assume that in the production version, whatever preprocessing you do on the list of drugs is only done once, and the search is the only part that is called repeatedly. Input: The entire list of drugs in alphabetical order, followed by a blank line, and then a partial sequence to search for. Output: The first 2 matching drugs, one per line, or <NONE> if there's no match.

IBM is implementing a streaming analytics service for an online retailer to provide data on the best time to sell specific products. To accomplish this goal, the service needs a very fast way to calculate a number of basic formulas for a series of dates, quantity and product id. Assumptions: In production, the list of data could be in the hundreds of thousands of dates with quantities and product ids. The service will be called frequently as the user explores different data models. The data may not be normalized. There could be multiple lines with the same date and product id. In these cases you should sum all the quantity on input and work on that. Product ids are case sensitive. Input: Each line will contain a comma separated list of text date (yyyy-mm-dd), quantity, product id. For example: 2017-06-02,5,Apples 2017-06-02,2,pears 2017-06-03,3,pineapples Output: For each date create a line of this format: date, total items sold for date, average quantity of all items (to 2 decimal places), count of unique parts sold. For example: