Skip to content

Latest commit

 

History

History
403 lines (312 loc) · 13.6 KB

README.md

File metadata and controls

403 lines (312 loc) · 13.6 KB
Raw

SimpleJdbc

Simple wrapper library for dealing with SQL-in-Java and JDBC. No dependencies, no reflection.

Features

  • Convenient, fluent syntax
  • Named Parameters (:foo, :bar instead of ?, ?)
  • Always uses PreparedStatement under the hood, helping to avoid SQL injection attacks
  • QueryResult wraps and fixes issues with ResultSet:
    • Boxed types (Integer instead of int) to make dealing with null-valued columns easier
    • Optional use of Optional for null-valued columns if that's your preferred style
    • Favors java.time classes instead of java.sql classes
    • "Escape hatch" lets you use underlying ResultSet for advanced use-cases
  • Basic transaction support
  • Avoids "magic" - no reflection, no code generation, no dynamic proxies

Statement

int rowsAffected =
    SimpleJdbc.using(dataSource)
        .statement("update some_table set foo = :foo, bar = :bar where id = :id")
        .bind("foo", "someValue")
        .bind("bar", 56.25)
        .bind("id", 12345L)
        .execute();

Batch Statement

int[] rowsAffected =
    SimpleJdbc.using(dataSource)
        .batchStatement("update some_table set foo = :foo, bar = :bar where id = :id")
        .bind("foo", "someValue")
        .bind("bar", 56.25)
        .bind("id", 12345L)
        .addBatch()
        .bind("foo", "someOtherValue")
        .bind("bar", 25.56)
        .bind("id", 54321L)
        .executeBatch();

Insert

int rowsAffected =
    SimpleJdbc.using(dataSource)
        .insert()
        .into("some_table")
        .set("column1", 123)
        .set("column2", 456)
        .execute();

Batch Insert

int[] rowsAffected =
    SimpleJdbc.using(dataSource)
        .batchInsert()
        .into("some_table")
        .set("column1", 123)
        .set("column2", "abc")
        .addBatch()
        .set("column1", 456)
        .set("column2", "def")
        .addBatch()
        .executeBatch();

Update

int rowsAffected =
    SimpleJdbc.using(dataSource)
        .update()
        .table("some_table")
        .set("foo", 123)
        .set("bar", 456)
        .where("id = :id and some_other_field = :someOtherField")
        .bind("id", 12345L)
        .bind("someOtherField", "someValue")
        .execute();

It's possible to perform the update without specifying any where() conditions. However, this is relatively rare, and mistakes can be dangerous (updating every row in a table when you meant to update only one would be... unpleasant). For that reason, a special method must be called in order to perform an unconditional update, making it more difficult to do it accidentally:

int rowsAffected =
    SimpleJdbc.using(dataSource)
        .update()
        .table("some_table")
        .set("foo", 123)
        .set("bar", 456)
        .executeUnconditionally();

Query

var result =
    SimpleJdbc.using(dataSource)
        .query("select foo, bar, baz from some_table where id = :id and thing = :thing")
        .bind("id",123)
        .bind("thing",456)
        .select(queryResult -> /* do interesting things with QueryResult here */);

Query Result Extraction

In the above example, the select() method accepts a QueryResultExtractor, which is a functional interface (lambda-compatible) having the following signature:

public interface QueryResultExtractor<T> {
  T extract(QueryResult queryResult) throws SQLException;
}

What this means is you can pass the select() method a lambda which accepts a single parameter (QueryResult) and returns anything you want (T). The result computed by your QueryResultExtractor lambda is what ends up being returned from the select() call.

If you are familiar with Spring's JdbcTemplate class, you will notice that QueryResultExtractor fills the same role as Spring JDBC's ResultSetExtractor. As the names imply, the main difference between these two are that while a ResultSetExtrator operates on a ResultSet, a QueryResultExtractor operates on a QueryResult.

It is important to note that when using the select() method, you need to manually advance the cursor through the query results, just like with a ResultSet. There are a number of convenience methods that save you this trouble (described later).

Differences between QueryResult and ResultSet

A QueryResult is just a thin wrapper around a ResultSet, exposing a smaller, opinionated interface with some useful differences in behavior.

Only column labels allowed

Unlike ResultSet, which exposes both getX(int columnIndex) and getX(String columnLabel), the only methods exposed by QueryResult use column labels. The author of this library is of the opinion that referencing query results (or parameters) by index is a smell that makes code more prone to bugs in the future.

NOTE to readers of this documentation: if you disagree or otherwise think it would be useful to elaborate on this point, feel free to open an issue to enhance the docs :-)

To illustrate the difference:

String sql = "select foo from some_table";

ResultSet resultSet = getResultSetFrom(sql);
resultSet.getString("foo"); // works
resultSet.getString(1);     // also works
        
QueryResult queryResult = getQueryResultFrom(sql);
queryResult.getString("foo"); // works
// queryResult.getString(1);  // would not compile

Boxed types over primitives

Because ResultSet returns primitive types whenever possible, dealing with null-valued columns can be problematic (if you care about accurately representing the column value):

// Given a ResultSet retrieved from the following query:
// select 5 as foo, null as bar from some_table

Integer foo = resultSet.getInt("foo"); // returns 5
Integer bar = resultSet.getInt("bar"); // returns 0   <--- this is bad! null is not zero!

Of course, you can determine if a column was null, but you have to explicitly check for it after every getX() call, like this:

Integer baz = resultSet.getInt("baz"); // returns 0
if (resultSet.wasNull()) {             // condition is true, because the last value was *actually* null
  baz = null;                          // there I fixed it
}

On the other hand, QueryResult returns boxed types, so when a column is null, you get back null:

// Same query & ResultSet, but accessed via a QueryResult

Integer foo = queryResult.getInteger("foo"); // returns 5
Integer bar = queryResult.getInteger("bar"); // returns null

Optionally, you can use Optional

If you prefer it, QueryResult can return Optional<T> instead of plain boxed types. Just call the opt() method to switch to the "Optional View", where the return types of all the getX() methods get wrapped in an Optional:

// Normal "boxed" mode
QueryResult queryResult = ...;
Integer foo = queryResult.getInteger("foo");
        
// Optional mode
QueryResult.OptionalView optQueryResult = queryResult.opt();
Optional<Integer> foo2 = optQueryResult.getInteger("foo");

If you prefer boxed mode in general, but want an Optional occasionally, it is canonical to reach for opt() inline whenever you need it:

QueryResult queryResult = ...;

MyThing thing = 
    MyThing.builder()
        .setFoo(queryResult.getInteger("foo"))
        .setBar(queryResult.getLong("bar"))
        .setBaz(queryResult.opt().getString("baz").orElse("some default value"))
        .setQux(queryResult.getInstant("qux"))
        .setQuux(queryResult.opt().getInteger("quux").orElseThrow(() -> new MissingQuuxException()))
        .build();

Favors java.time over java.sql types

QueryResult automatically converts date/time-related java.sql types to java.time types so that you don't have to. Nothing too special here - we're just calling a method on the java.sql types themselves to achieve this, but we handle null as you'd expect it to be handled.

Available methods:

QueryResult queryResult = ...;

queryResult.getInstant("foo");       // from java.sql.Timestamp
queryResult.getLocalDateTime("bar"); // from java.sql.Timestamp
queryResult.getLocalDate("baz");     // from java.sql.Date
queryResult.getLocalTime("qux");     // from java.sql.Time

Escape Hatch

While QueryResult is more convenient to use, it is not as feature-packed as ResultSet. Some glaring omissions include support for CLOB or BLOB types, and many "advanced" use-cases. Thankfully, when adopting SimpleJdbc and using QueryResult, you don't have to throw the baby out with the bathwater.

It is possible to easily access the underlying ResultSet instance from any QueryResult:

ResultSet resultSet = queryResult.toResultSet();

Query - convenience methods

Using query()...select() as described above is great and all, but one annoying chore that QueryResult does not solve for you is the need to call next() to advance the cursor through the underlying ResultSet.

Thankfully, there are a number of additional convenience methods that do save you this trouble:

  • selectList(queryRowResult -> ...)
  • selectFirst(queryRowResult -> ...)
  • selectMaybeOne(queryRowResult -> ...)
  • selectExactlyOne(queryRowResult -> ...)
  • selectExists()

Each of these are described in more detail below, but first it should be noted that the convenience methods that accept a lambda are provided with a QueryRowResult and are called once per result row. This, as opposed to select(queryResult -> ...), which is provided a QueryResult and called only one time.

The only material difference between a QueryResult and a QueryRowResult is that the latter _ lacks_ a next() method. This is to help programmers avoid shooting themselves in the foot, in case they forgot they were using a convenience method.

Below you will find additional details about each convenience method:

Returning a List of Results

Most often, our queries return multiple rows, and we want to treat each row as an individual result. The selectList() method makes this more convenient by invoking the provided callback once per result row, and collecting the results as a list.

List<Foo> results = simpleJdbc.query(...).selectList(queryRowResult -> /* return Foo */);

Just this once, we'll compare the difference between the select() method and the convenience method:

// Inconvenient
List<String> foos =
    simpleJdbc.query("select foo from some_table where filter_column = :filterValue")
        .bind("filterValue", "SOME_VALUE")
        .select(
            queryResult -> {
                List<String> returnVal = new ArrayList<>();
                while (queryResult.next()) {
                  returnVal.add(queryResult.getString("foo"));
                }
                return returnVal;
            });

// More Convenient
List<String> foos =
    simpleJdbc.query("select foo from some_table where filter_column = :filterValue")
        .bind("filterValue", "SOME_VALUE")
        .selectList(queryRowResult -> queryRowResult.getString("foo"));

Getting the First Result (if it exists)

The selectFirst() method invokes the provided callback for the first result row only, and returns the result in an Optional. If the query yields no results, the returned Optional is empty. If the query yields multiple results, everything beyond the first result is simply ignored.

Optional<Foo> result = simpleJdbc.query(...).selectFirst(queryRowResult -> ...);

Expecting At Most One Result

The selectMaybeOne() method behaves exactly like the previous method, with the exception (pun intended) that if the query yields more than one result, an exception will be thrown rather than returning a result.

Optional<Foo> result = simpleJdbc.query(...).selectMaybeOne(queryRowResult -> ...);

Expecting Exactly One Result

The selectExactlyOne() method behaves very similarly to the previous method, but tightens the contract even further. If the query yields one row, the callback is invoked and result returned directly (not wrapped in Optional<>). If the query yields anything other than a single row, an exception is thrown.

Foo result = simpleJdbc.query(...).selectExactlyOne(queryRowResult -> ...);

Existence Check

Sometimes, you just want to know if there where any results, and you don't even need to extract anything. The selectExists() method simply returns true if the query returned any results, false otherwise.

boolean isResultExists = simpleJdbc.query(...).selectExists();

Transactions

SimpleJdbc.using(dataSource)
    .transactionally(simpleJdbc -> {
        // use lambda parameter (a SimpleJdbc instance) to do stuff to the database 
        // automatically calls rollback() on underlying Connection if an exception is raised
        // or commit() at end of function if no exceptions raised
    });

It is possible to control the transaction isolation level by passing an additional parameter:

SimpleJdbc.using(dataSource)
    .transactionally(
        TransactionIsolationLevel.REPEATABLE_READ,
        simpleJdbc -> {
            // use lambda parameter (a SimpleJdbc instance) to do stuff to the database 
            // automatically calls rollback() on underlying Connection if an exception is raised
            // or commit() at end of function if no exceptions raised
        });

TODO

Definitely

  1. More integration tests with more DB vendors (at minimum: MySQL & Oracle)
  2. Clob and Blob support
  3. More Javadoc
  4. SimpleJdbcException subtypes (different errors, different exception types)

Maybe

  1. SQL script support
  2. Actively prevent creation of nested transactions (because they wouldn't work correctly anyway)
  3. Plugins / hooks? Something that could conveniently implement an in-clause builder