6.893 Architecture of Database Systems

Term Project, due May 17, 2002

Obtain a copy of POSTGRES off the web, and load it onto some machine that you have access to. Do one of the following two assignments. There is no write-up required. Instead you will be asked to demo your prototype in person on the 17th. You will need to construct a convincing demo to accomplish this. Lastly, you must work alone. Make sure you read papers 5 and 6 of Chapter 6 of the red book before you begin.

1) abstract data type exercise (easiesr task)

Implement an abstract data type in POSTGRES that supports "Wall Street Trading Date" (WSTD). This data type is represented externally as a pair (year, integer); for example, (2002, 3) represents the third trading day of 2002 (which I believe is Jan 4th). When, year is not specified, the default is the current year, so 3 represents January 4, 2002. Choose some reasonable internal representation for the WSTD data type and implement the following operations:

Same_year (WSTD-1, WSTD-2) returns Boolean; this returns true if the values are in the same year.
Same_day_of_week (WSTD-1, WSTD-2) returns Boolean; this returns true if the two values are the same day of the week.
Distance (WSTD-1, WSTD-2) returns integer; this returns the number of trading days between the two WSTDs.
Mon (WSTD), Tues (WSTD), Wed (WTD), Thu (WSTD), Fri (WSTD) returns Boolean; these have the obvious interpretation.
DJIA (WSTD) returns float; returns the Dow Jones Industrial Average on this date
Date (WSTD) returns Date; returns the calendar date of the given WSTD
Greater_than (WSTD-1, WSTD-2) returns Boolean
Less_than (WSTD-1, WSTD-2) returns Boolean
Greater_Equal (WSTD-1, WSTD-2) returns Boolean
Less_Equal (WSTD-1, WSTD-2) returns Boolean
Equals (WSTD-1, WSTD-2) returns Boolean

In addition, POSTGRES should be taught how to construct B-tree indexes on the WSTD type and the optimizer should be taught how to utilize the resulting indexes.

Note:
Wall Street Trading Date: defined to be those days on which the NYSE trades shares. Best way to figure this out is to go to a financial web site (e.g. schwab.com, finanacial section of yahoo.com, etc.) and build a definition by looking at the trading days over the last several months.

Reason to give you this data type is it represents typical issues faced in data type construction. Semantics must be "discovered".

DJIA (WSTD): can look up on the fly from a web site or download in advance. Your choice.

2) POSTGRES enhancement (hard - don't do this unless you are very confident)

Add bit-map indexes to POSTGRES. Construct syntax and an implementation for a create index facility. Change the optimizer so it can make use of your indexes. Change the executor so it can deal with such indexes.