Just a quick patch. This makes the JDBC driver thread safe, which is an

important step towards making the driver compliant, and means that for
some Java applications and servlets, only a single database connection
is
needed, so in a sence this is a nice little show stopper for 6.4 (and
should still be backward compatible to 6.3.2).

Peter
This commit is contained in:
Bruce Momjian 1998-10-08 00:38:21 +00:00
parent 9042e9d757
commit 25b5faa7cd
4 changed files with 368 additions and 4 deletions

View File

@ -4,7 +4,7 @@
# Makefile for Java JDBC interface
#
# IDENTIFICATION
# $Header: /cvsroot/pgsql/src/interfaces/jdbc/Attic/Makefile,v 1.9 1998/09/03 02:29:41 momjian Exp $
# $Header: /cvsroot/pgsql/src/interfaces/jdbc/Attic/Makefile,v 1.10 1998/10/08 00:38:18 momjian Exp $
#
#-------------------------------------------------------------------------
@ -137,7 +137,8 @@ EX= example/basic.class \
example/datestyle.class \
example/psql.class \
example/ImageViewer.class \
example/metadata.class
example/metadata.class \
example/threadsafe.class
# example/Objects.class
# This rule builds the examples
@ -160,7 +161,9 @@ examples: postgresql.jar $(EX)
@echo " example.psql Simple java implementation of psql"
@echo " example.Objects Demonstrates Object Serialisation"
@echo " "
@echo These are not really examples, but tests various parts of the driver
@echo " example.metadata Tests various metadata methods"
@echo " example.threadsafe Tests the driver's thread safety"
@echo ------------------------------------------------------------
@echo
@ -170,6 +173,6 @@ example/datestyle.class: example/datestyle.java
example/psql.class: example/psql.java
example/ImageViewer.class: example/ImageViewer.java
#example/Objects.class: example/Objects.java
example/threadsafe.class: example/threadsafe.java
example/metadata.class: example/metadata.java
#######################################################################

View File

@ -0,0 +1,353 @@
package example;
import java.io.*;
import java.sql.*;
import java.text.*;
// rare in user code, but we use the LargeObject API in this test
import postgresql.largeobject.*;
/**
* This example tests the thread safety of the driver.
*
* It does this by performing several queries, in different threads. Each
* thread has it's own Statement object, which is (in my understanding of the
* jdbc specification) the minimum requirement.
*
*/
public class threadsafe
{
Connection db; // The connection to the database
Statement st; // Our statement to run queries with
public threadsafe(String args[]) throws ClassNotFoundException, FileNotFoundException, IOException, SQLException
{
String url = args[0];
String usr = args[1];
String pwd = args[2];
// Load the driver
Class.forName("postgresql.Driver");
// Connect to database
System.out.println("Connecting to Database URL = " + url);
db = DriverManager.getConnection(url, usr, pwd);
System.out.println("Connected...Now creating a statement");
st = db.createStatement();
// Clean up the database (in case we failed earlier) then initialise
cleanup();
// Now run tests using JDBC methods, then LargeObjects
doexample();
// Clean up the database
cleanup();
// Finally close the database
System.out.println("Now closing the connection");
st.close();
db.close();
}
/**
* This drops the table (if it existed). No errors are reported.
*/
public void cleanup()
{
try {
st.executeUpdate("drop table basic");
} catch(Exception ex) {
// We ignore any errors here
}
}
/**
* This performs the example
*/
public void doexample() throws SQLException
{
System.out.println("\nThis test runs three Threads. Two simply insert data into a table, then\nthey perform a query. While they are running, a third thread is running,\nand it load data into, then reads from a Large Object.\n\nIf alls well, this should run without any errors. If so, we are Thread Safe.\nWhy test JDBC & LargeObject's? Because both will run over the network\nconnection, and if locking on the stream isn't done correctly, the backend\nwill get pretty confused!\n");
thread3 thread3=null;
try {
// create the two threads
Thread thread0 = Thread.currentThread();
Thread thread1 = new thread1(db);
Thread thread2 = new thread2(db);
thread3 = new thread3(db);
// now run, and wait for them
thread1.start();
thread2.start();
thread3.start();
// ok, I know this is bad, but it does the trick here as our main thread
// will yield as long as either of the children are still running
System.out.println("Waiting for threads to run");
while(thread1.isAlive() || thread2.isAlive() || thread3.isAlive())
thread0.yield();
} finally {
// clean up after thread3 (the finally ensures this is run even
// if an exception is thrown inside the try { } construct)
if(thread3 != null)
thread3.cleanup();
}
System.out.println("No Exceptions have been thrown. This is a good omen, as it means that we are\npretty much thread safe as we can get.");
}
// This is the first thread. It's the same as the basic test
class thread1 extends Thread
{
Connection c;
Statement st;
public thread1(Connection c) throws SQLException {
this.c = c;
st = c.createStatement();
}
public void run() {
try {
System.out.println("Thread 1 running...");
// First we need a table to store data in
st.executeUpdate("create table basic (a int2, b int2)");
// Now insert some data, using the Statement
st.executeUpdate("insert into basic values (1,1)");
st.executeUpdate("insert into basic values (2,1)");
st.executeUpdate("insert into basic values (3,1)");
// For large inserts, a PreparedStatement is more efficient, because it
// supports the idea of precompiling the SQL statement, and to store
// directly, a Java object into any column. PostgreSQL doesnt support
// precompiling, but does support setting a column to the value of a
// Java object (like Date, String, etc).
//
// Also, this is the only way of writing dates in a datestyle independent
// manner. (DateStyles are PostgreSQL's way of handling different methods
// of representing dates in the Date data type.)
PreparedStatement ps = db.prepareStatement("insert into basic values (?,?)");
for(int i=2;i<200;i++) {
ps.setInt(1,4); // "column a" = 5
ps.setInt(2,i); // "column b" = i
ps.executeUpdate(); // executeUpdate because insert returns no data
if((i%50)==0)
DriverManager.println("Thread 1 done "+i+" inserts");
}
ps.close(); // Always close when we are done with it
// Finally perform a query on the table
DriverManager.println("Thread 1 performing a query");
ResultSet rs = st.executeQuery("select a, b from basic");
int cnt=0;
if(rs!=null) {
// Now we run through the result set, printing out the result.
// Note, we must call .next() before attempting to read any results
while(rs.next()) {
int a = rs.getInt("a"); // This shows how to get the value by name
int b = rs.getInt(2); // This shows how to get the value by column
//System.out.println(" a="+a+" b="+b);
cnt++;
}
rs.close(); // again, you must close the result when done
}
DriverManager.println("Thread 1 read "+cnt+" rows");
// The last thing to do is to drop the table. This is done in the
// cleanup() method.
System.out.println("Thread 1 finished");
} catch(SQLException se) {
System.err.println("Thread 1: "+se.toString());
se.printStackTrace();
System.exit(1);
}
}
}
// This is the second thread. It's the similar to the basic test, and thread1
// except it works on another table.
class thread2 extends Thread
{
Connection c;
Statement st;
public thread2(Connection c) throws SQLException {
this.c = c;
st = c.createStatement();
}
public void run() {
try {
System.out.println("Thread 2 running...");
// For large inserts, a PreparedStatement is more efficient, because it
// supports the idea of precompiling the SQL statement, and to store
// directly, a Java object into any column. PostgreSQL doesnt support
// precompiling, but does support setting a column to the value of a
// Java object (like Date, String, etc).
//
// Also, this is the only way of writing dates in a datestyle independent
// manner. (DateStyles are PostgreSQL's way of handling different methods
// of representing dates in the Date data type.)
PreparedStatement ps = db.prepareStatement("insert into basic values (?,?)");
for(int i=2;i<200;i++) {
ps.setInt(1,4); // "column a" = 5
ps.setInt(2,i); // "column b" = i
ps.executeUpdate(); // executeUpdate because insert returns no data
if((i%50)==0)
DriverManager.println("Thread 2 done "+i+" inserts");
}
ps.close(); // Always close when we are done with it
// Finally perform a query on the table
DriverManager.println("Thread 2 performing a query");
ResultSet rs = st.executeQuery("select * from basic where b>1");
int cnt=0;
if(rs!=null) {
// First find out the column numbers.
//
// It's best to do this here, as calling the methods with the column
// numbers actually performs this call each time they are called. This
// really speeds things up on large queries.
//
int col_a = rs.findColumn("a");
int col_b = rs.findColumn("b");
// Now we run through the result set, printing out the result.
// Again, we must call .next() before attempting to read any results
while(rs.next()) {
int a = rs.getInt(col_a); // This shows how to get the value by name
int b = rs.getInt(col_b); // This shows how to get the value by column
//System.out.println(" a="+a+" b="+b);
cnt++;
}
rs.close(); // again, you must close the result when done
}
DriverManager.println("Thread 2 read "+cnt+" rows");
// The last thing to do is to drop the table. This is done in the
// cleanup() method.
System.out.println("Thread 2 finished");
} catch(SQLException se) {
System.err.println("Thread 2: "+se.toString());
se.printStackTrace();
System.exit(1);
}
}
}
// This is the third thread. It loads, then reads from a LargeObject, using
// our LargeObject api.
//
// The purpose of this is to test that FastPath will work in between normal
// JDBC queries.
class thread3 extends Thread
{
Connection c;
Statement st;
LargeObjectManager lom;
LargeObject lo;
int oid;
public thread3(Connection c) throws SQLException {
this.c = c;
//st = c.createStatement();
// create a blob
lom = ((postgresql.Connection)c).getLargeObjectAPI();
oid = lom.create();
System.out.println("Thread 3 has created a blob of oid "+oid);
}
public void run() {
try {
System.out.println("Thread 3 running...");
DriverManager.println("Thread 3: Loading data into blob "+oid);
lo = lom.open(oid);
FileInputStream fis = new FileInputStream("example/threadsafe.java");
// keep the buffer size small, to allow the other thread a chance
byte buf[] = new byte[128];
int rc,bc=1,bs=0;
while((rc=fis.read(buf))>0) {
DriverManager.println("Thread 3 read block "+bc+" "+bs+" bytes");
lo.write(buf,0,rc);
bc++;
bs+=rc;
}
lo.close();
fis.close();
DriverManager.println("Thread 3: Reading blob "+oid);
lo=lom.open(oid);
bc=0;
while(buf.length>0) {
buf=lo.read(buf.length);
if(buf.length>0) {
String s = new String(buf);
bc++;
DriverManager.println("Thread 3 block "+bc);
DriverManager.println("Block "+bc+" got "+s);
}
}
lo.close();
System.out.println("Thread 3 finished");
} catch(Exception se) {
System.err.println("Thread 3: "+se.toString());
se.printStackTrace();
System.exit(1);
}
}
public void cleanup() throws SQLException {
if(lom!=null && oid!=0) {
System.out.println("Thread 3: Removing blob oid="+oid);
lom.delete(oid);
}
}
}
/**
* Display some instructions on how to run the example
*/
public static void instructions()
{
System.out.println("\nThis tests the thread safety of the driver.\n\nThis is done in two parts, the first with standard JDBC calls, and the\nsecond mixing FastPath and LargeObject calls with queries.\n");
System.out.println("Useage:\n java example.threadsafe jdbc:postgresql:database user password [debug]\n\nThe debug field can be anything. It's presence will enable DriverManager's\ndebug trace. Unless you want to see screens of items, don't put anything in\nhere.");
System.exit(1);
}
/**
* This little lot starts the test
*/
public static void main(String args[])
{
System.out.println("PostgreSQL Thread Safety test v6.4 rev 1\n");
if(args.length<3)
instructions();
// This line outputs debug information to stderr. To enable this, simply
// add an extra parameter to the command line
if(args.length>3)
DriverManager.setLogStream(System.err);
// Now run the tests
try {
threadsafe test = new threadsafe(args);
} catch(Exception ex) {
System.err.println("Exception caught.\n"+ex);
ex.printStackTrace();
}
}
}

View File

@ -635,8 +635,11 @@ public class Connection implements java.sql.Connection
* @return a ResultSet holding the results
* @exception SQLException if a database error occurs
*/
public synchronized ResultSet ExecSQL(String sql) throws SQLException
public ResultSet ExecSQL(String sql) throws SQLException
{
// added Oct 7 1998 to give us thread safety.
synchronized(pg_stream) {
Field[] fields = null;
Vector tuples = new Vector();
byte[] buf = new byte[sql.length()];
@ -737,6 +740,7 @@ public class Connection implements java.sql.Connection
if (final_error != null)
throw final_error;
return new ResultSet(this, fields, tuples, recv_status, 1);
}
}
/**

View File

@ -68,6 +68,9 @@ public class Fastpath
*/
public Object fastpath(int fnid,boolean resulttype,FastpathArg[] args) throws SQLException
{
// added Oct 7 1998 to give us thread safety
synchronized(stream) {
// send the function call
try {
// 70 is 'F' in ASCII. Note: don't use SendChar() here as it adds padding
@ -153,6 +156,7 @@ public class Fastpath
throw new SQLException("Fastpath: protocol error. Got '"+((char)in)+"'");
}
}
}
}
/**