2 Goals and philosophy
This section outlines the main goals and philosophy driving the API design.
2.1 A SQL level API
JDBC is intended as a "call-level" SQL interface for Java. This means the focus is on executing
raw SQL statements and retrieving their results. We expect that higher-level APIs will be defined
as well, and these will probably be implemented on top of this base level. Examples of
higher-level APIs are direct transparent mapping of tables to Java classes, semantic tree representations
of more general queries, and an embedded SQL syntax for Java.
We expect that various application builder tools will emit code that uses our API. However we
also intend that the API be usable by human programmers, especially because there is no other
solution available for Java right now.
2.2 SQL Conformance
Database systems support a wide range of SQL syntax and semantics, and they are not consistent
with each other on more advanced functionality such as outer joins and stored procedures.
Hopefully with time the portion of SQL that is truly standard will expand to include more and
more functionality. In the meantime, we take the following position:
- JDBC allows any query string to be passed through to an underlying DBMS driver, so
an application may use as much SQL functionality as desired at the risk of receiving an
error on some DBMSs. In fact, an application query need not even be SQL, or it may
be a specialized derivative of SQL, e.g. for document or image queries, designed for
specific DBMSs.
- In order to pass JDBC compliance tests and to be called "JDBC COMPLIANT(TM)" we
require that a driver support at least ANSI SQL92 Entry Level. This gives applications
that want wide portability a guaranteed least common denominator. We believe ANSI
SQL-2 Entry Level is reasonably powerful and is reasonably widely supported today.
2.3 JDBC must be implementable on top of common database interfaces
We need to ensure that the JDBC SQL API can be implemented on top of common SQL level
APIs, in particular ODBC. This requirement has colored some parts of the specification, notably
the handling of OUT parameters and large blobs.
2.4 Provide a Java interface that is consistent with the rest of the Java system
There has been a very strong positive response to Java. To a large extent this seems to be because
the language and the standard runtimes are perceived as being consistent, simple, and
powerful.
As far as we can, we would like to provide a Java database interface that builds on and reinforces
the style and virtues of the existing core Java classes.
2.5 Keep it simple
We would prefer to keep this base API as simple as possible, at least initially. In general we
would prefer to provide a single mechanism for performing a particular task, and avoid providing
duplicate mechanisms. We will extend the API later if any important functionality is missing.
2.6 Use strong, static typing wherever possible
We would prefer that the JDBC API be strongly typed, with as much type information as possible
expressed statically. This allows for more error checking at compile time.
Because SQL is intrinsically dynamically typed, we may encounter type mismatch errors at
run-time where for example a programmer expected a SELECT to return an integer result but
the database returned a string "foo". However we would still prefer to allow programmers to
express their type expectations at compile time, so that we can statically check as much as possible.
We will also support dynamically typed interfaces where necessary (see particularly
Chapter 14).
2.7 Keep the common cases simple
We would like to make sure that the common cases are simple, and that the uncommon cases
are doable.
A common case is a programmer executing a simple SQL statement (such as a SELECT, INSERT,
UPDATE or DELETE) without parameters, and then (in the case of SELECT statement)
processing rows of simple result types. A SQL statement with IN parameters is also
common.
Somewhat less common, but still important, is the case where a programmer invokes a SQL
statement using INOUT or OUT parameters. We also need to support SQL statements that read
or write multi-megabyte objects, and less common cases such as multiple result sets returned
by a SQL statement.
We expect that metadata access (e.g. to discover result-set types, or to enumerate the procedures
in a database) is comparatively rare and is mostly used by sophisticated programmers or
by builder tools. Metadata functions are therefore documented at the end of the specification,
along with dynamically-typed data access; the average programmer can skip these sections.
2.8 Use multiple methods to express multiple functionality
One style of interface design is to use a very small number of procedures and offer a large number
of control flags as arguments to these procedures, so that they can be used to effect a wide
range of different behavior.
In general the philosophy of the Java core classes has been to use different methods to express
different functionality. This tends to lead to a larger number of methods, but makes each method
easier to understand. This approach has the major advantage that programmers who are
learning how to use the basic interface aren't confused by having to specify arguments related
to more complex behaviors.
We've tried to adopt the same approach for the JDBC interface, and in general have preferred
to use multiple methods rather than using multi-purpose methods with flag arguments.
Contents | Prev | Next
jdbc@wombat.eng.sun.com
or
jdbc-odbc@wombat.eng.sun.com
Copyright © 1996, 1997 Sun Microsystems, Inc. All rights reserved.