In this project, you'll be asked to evaluate a few single-table SQL queries. That said, there's a few ways to pull this off, some better than others. In short, this project exists to lay the foundation for the later two projects. In this writeup, we'll be going through a few design decisions, pointing out tradeoffs, and explaining why a strategy that might seem easier in the short term turns out to be significantly harder later.
Specifically, you'll be given a number of queries in one of the following patterns:
This script is basically a form of pattern 2 above
SELECT fields[1] FROM 'data.csv'
WHERE fields[2] != "Ensign" AND CAST(fields[3] AS int) > 25
Or in other words, any query that follows the pattern...
SELECT /*targets*/ FROM /*file*/ WHERE /*condition*/
...becomes a script of the form...
This is nice and simple, but the code is very specific to pattern 2. That's something that will lead us into trouble. To see a simple example of the sort of problems we're going to run into, let's come up with an example of pattern 6:
SELECT SUM(age) FROM 'data.csv' WHERE rank != 'Ensign'
That is, we're asking for the total age of non-ensigns in our example table. An equivalent script would be...
There's a pretty significant difference in the flow of the code in this version of the script. For one, there's a new global variable with the total age. For another, the print statement is now outside of the for loop. Now let's say we wanted to support both patterns 2 and 6. We'd need to make the code quite a bit more complex:
As you can see, this code is already quite a bit more complex and we haven't even looked at patterns 1, 3, 4, or 5 yet or the even more complex queries that will show up in Checkpoint 2 and 3.
There are a number of workflow steps that appear in more than one pattern. For example:
In short, this idea of rows and attributes is pretty fundamental, so let's use it. We're going to work with data expressed in terms of tables: or collections of rows and attributes. This allows us to abstract out each of those workflow steps from before into a set of functions:
But we still have a problem. These table objects are going to be as big as the data they represent... they can get super large. That's a massive drawback compared to our initial script design, which has constant-space usage. So what else can we do?
Let's look at why the original script uses constant-space. We load one record in upfront (that's constant space). We decide whether the record is useful to us (still constant space). Whether or not we print it, by the time we get to the next record, we're done with the current row and can safely discard it. Can we recover the same sort of property?
For this checkpoint, it turns out that we can. If you've used java, you're probably familiar with the Iterator interface. An iterator lets you access elements of a collection without needing to have all of those elements available at once. That is, you define two methods:
In short, iterators give you composability and low memory use. The first property is important for your sanity, while the latter property is important for your performance.
When it comes to figuring out how to represent one row of data, you have two questions to answer: (1) How do I represent a single primitive value, and (2) How do I represent an entire row of primitive values.
For the first question, there are two practical choices: Either as raw strings (taken directly from the CSV file) or parsed into PrimitiveValue objects. PrimitiveValue is an interface implemented by several classes that represent specific types of values, for example longs, dates, and others. Because EvalLib (a library that I'll describe shortly) uses PrimitiveValues internally, most students find that it is easier to write code that performs well if you use PrimitiveValue.
For the second question, I strongly encourage the use of Java arrays. There are a few options, including ArrayLists, Vectors, Maps, and other structures. Java arrays outperform them all pretty drastically.
The JSqlParser Expression type can represent a whole mess of different arithmetic, boolean, and other primitive-valued expressions. For this project, you'll have a library to help you in evaluating these expressions: EvalLib. Before we get into it, you should note a distinction between two types of expression:
Eval eval = new Eval(){ /* we'll get what goes here shortly */ }
// Evaluate "1 + 2.0"
PrimitiveValue result;
result =
eval.eval(
new Addition(
new LongPrimitive(1),
new DoublePrimitive(2.0)
)
);
System.out.println("Result: "+result); // "Result: 3.0"
// Evaluate "1 > (3.0 * 2)"
result =
eval.eval(
new GreaterThan(
new LongPrimitive(1),
new Multiplication(
new DoublePrimitive(3.0),
new LongPrimitive(2)
)
)
);
System.out.println("Result: "+result); // "Result: false"
In short, eval helps you evaluate the Expression objects that JSQLParser gives you. However, there's one thing it can't do: It has no idea how to convert attribute names to values. That is, there's one type of Expression object that Eval has no clue how to evaluate: Column. That is, let's take the following example:
// Evaluate "R.A >= 5"
result =
eval.eval(
new GreaterThanEquals(
new Column(new Table(null, "R"), "A"),
new LongPrimitive(5)
)
);
What value should Eval give to R.A? This depends on the data. Because EvalLib has no way of knowing how you represent your data, you need to tell it:
Eval eval = new Eval(){
public PrimitiveValue eval(Column c){
/* Figure out what value 'c' has */
}
}
For this checkpoint, you'll be running multiple queries in sequence. This means a few changes to your code. First, before calling parser.Statement(), you will need to print a prompt to System.out. Use the string '$> ' (without quotes), and make sure that it's the very first thing on its own. This is so that the testing framework knows when your code is ready for the next query.
Because you are implementing a query evaluator and not a full database engine, there will not be any tables -- at least not in the traditional sense of persistent objects that can be updated and modified. Instead, you will be given a Table Schema and a CSV File with the instance in it. To keep things simple, we will use the CREATE TABLE statement to define a relation's schema. To reiterate, CREATE TABLE statements only appear to give you a schema. You do not need to allocate any resources for the table in reaction to a CREATE TABLE statement -- Simply save the schema that you are given for later use. Sql types (and their corresponding java types) that will be used in this project are as follows:
SQL Type | Java Equivalent |
---|---|
string | StringValue |
varchar | StringValue |
char | StringValue |
int | LongValue |
decimal | DoubleValue |
date | DateValue |
In addition to the schema, you will find a corresponding [tablename].csv file in the data directory (just like in checkpoint 0). The name of the table corresponds to the table names given in the CREATE TABLE statements your code receives. For example, let's say that you see the following statement in your query file:
CREATE TABLE R(A int, B int, C int);
That means that the data directory contains a data file called 'R.dat' that might look like this:
1|1|5 1|2|6 2|3|7
Each line of text (see BufferedReader.readLine()) corresponds to one row of data. Each record is delimited by a vertical pipe '|' character. Integers and floats are stored in a form recognized by Java’s Long.parseLong() and Double.parseDouble() methods. Dates are stored in YYYY-MM-DD form, where YYYY is the 4-digit year, MM is the 2-digit month number, and DD is the 2-digit date. Strings are stored unescaped and unquoted and are guaranteed to contain no vertical pipe characters.
As before, all .java files in the src directory at the root of your repository will be compiled (and linked against JSQLParser). Also as before, the class dubstep.Main will be invoked with no arguments, and a stream of semicolon-delimited queries will be printed to System.in (after you print out a prompt)
For example (red text is entered by the user/grader):
bash> ls data R.dat S.dat T.dat bash> cat data/R.dat 1|1|5 1|2|6 2|3|7 bash> java -cp build:jsqlparser.jar dubstep.Main - $> CREATE TABLE R(A int, B int, C int); $> SELECT B, C FROM R WHERE A = 1; 1|5 2|6 $> SELECT SUM(B), SUM(C) FROM R; 6|18
For this project, your code will not be timed, but you will need to answer some queries with a cap on available memory. You will receive up to 7 points for answering queries successfully, up to 3 additional points for remaining within memory usage bounds, and 5 points awarded as part of a code review after the project deadline.
drThis page last updated 2024-08-27 00:23:19 -0400