Pymqi combines the power of Python with the benefits of the messaging model. It can be used to develop test harnesses for MQ based systems, for rapid prototyping of MQ applications, for development of administrative GUI's (when combined with e.g., TkInter), and even for mainstream MQ application development!
Pymqi does not replace MQI, but is layered on top of it, so you must have MQ (either client or server) installed before you can use pymqi.
Pymqi consists of several modules that are used together:
Its easiest to use the pymqi package. Here is some minimal code to put a message on a queue:
And here's some more to get it again:import pymqi qmgr = pymqi.QueueManager() putQ = pymqi.Queue(qmgr, 'TESTQ1') putQ.put('Hello from Python!')
import pymqi qmgr = pymqi.QueueManager() getQ = pymqi.Queue(qmgr, 'TESTQ1') print 'Heres the message:', getQ.get()
Easy, Eh?
For the MQI calls that pymqi supports, the full set of options are implemented. This allows you access to the full functionality of the MQI call from Python. Pymqi defines classes that are synonymous with the MQI structure parameters, such as MQGMO etc. This lets you program MQI 'the Python way'.
setup.py file. Download, unzip & untar the file, then
cd into the pymqi directory. To build a MQI client, run python
setup.py build client, to build a MQI server, python
setup.py build server. Finally, run python setup.py
install
If you are building pymqi on Windows and you don't have MSVC installed, you can use MinGW instead. Jaco Smuts has written notes here explaining how its done.
If you port pymqi to other platforms, please let me know.
In support of these, pymqi implements the following structures. Others will follow to support additional MQI calls as required.
For a client build, pymqi.__mqbuild__ is set to the
string 'client', otherwise it is set to 'server'. The supported
command levels (from 5.0 onwards) for the version of MQI linked with
pymqi are available in the tuple pymqi.__mqlevels__. To
determine if a particular level is supported, do something like:
if '5.3' in pymqi.__mqlevels__: print 'New MQI things to try!'
To use PCF with pymqi, instantiate a PFCExecute object as you would a
QueueManager. Then call a PCF command, passing a dictionary of
attributes and values, as appropriate to the command. Commands and
attributes are as defined by IBM in cmqc.h,
cmqcfc.h, and in their documentation.
You can also use PCF to query a Queue Manager. In this case, PCFExecute returns a list of dictionaries, with the attributes and values you have requested.
An example PCFExecute usage is given below.
import pymqi, CMQC from CMQCFC import * # Connect to default Queue Manager pcf = pymqi.PCFExecute() # Ping the Queue Manager pcf.MQCMD_PING_Q_MGR() # Create a channel "FNURGLE" chanArgs = {MQCACH_CHANNEL_NAME : "FNURGLE", MQIACH_CHANNEL_TYPE : CMQC.MQCHT_RECEIVER} pcf.MQCMD_CREATE_CHANNEL(chanArgs) # Query all queues beginning with "QTEST" queues = pcf.MQCMD_INQUIRE_Q({CMQC.MQCA_Q_NAME : "QTEST*"}) for q in queues: print pcf.stringifyKeys(q)
This depends on the MQ Transaction Manager, which in turn coordinates resources through the XA interface. To make this work from Python, you must use an XA enabled database API. The only one I'm aware of that works is DCOracle2. You will need to apply these (unofficial) patches.
You will also have to configure the Queue Manager with XA information for the resources you want it to coordinate. See the MQSeries System Administration book for more details.
You might also be interested in a Python extension for the XA Switch interface. This lets Python scripts participate directly in the Distributed Transaction two phase commit protocol. See here for more details. Pyxasw & pymqi are related but independent of each other.
This code fragment shows a transactional MQ put & DCO2 database insert, coordinated by MQ.
import pymqi, DCOracle2 ... qmgr = pymqi.QueueManager() q = pymqi.Queue(qmgr, 'TESTQ1') pmo = pymqi.pmo(Options = CMQC.MQPMO_SYNCPOINT) md = pymqi.md() # Begin a global transaction qm.begin() # Connect a XA managed database conn=DCOracle2.connectXA() curs = conn.cursor() # Do a transactional put & db insert q.put('TM comes to Linux!', md, pmo) curs.execute("INSERT INTO TESTTABLE VALUES(42, 'Lala')") # Now commit the transaction qmgr.commit()
The QueueManager.inquire() and Queue.inquire() calls allow you to inquire on a single MQ attribute. An integer or string attribute value is returned, as appropriate.
The Queue.set() call lets you set a single Queue attribute. The value passed must be of the appropriate type.
When inquiring or setting on a queue, the queue must be specifically opened for inquire or set. The operations cannot be mixed on a queue, neither can a get/put queue be used for inquire or set. No such restriction applies to QueueManager objects.
An example inquire is shown below.
import pymqi, CMQC ... qmgr = pymqi.QueueManager() q = pymqi.Queue(qmgr, 'TESTQ1') print 'Queue depth:', q.inquire(CMQC.MQIA_CURRENT_Q_DEPTH) print 'Queue Manager platform:', qmgr.inquire(MQIA_PLATFORM)
And you're certain the Queue Manager is running, then you're probably running in an MQ Client environment without telling pymqi where the server is. If you don't know the difference between an MQ Client and Server, please read the IBM documentation.pymqi.MQMIError: MQI Error. Comp: 2, Reason 2058: FAILED: MQRC_Q_MGR_NAME_ERROR
The answer is to either:
MQSERVER environment variable to something like
"SYSTEM.DEF.SVRCONN/TCP/192.168.1.24(1414)" (or whatever
your MQ Server channel/address is).
connectTCPClient call as in:
qMgr = QueueManager(None) qMgr.connectTCPClient('les.queue.manager', cd(), 'SYSTEM.DEF.SVRCONN', '192.168.1.24(1414)')
If it still doesn't work, then you've probably built pymqi against the MQ server library rather than the client.
MQGMO_ACCEPT_TRUNCATED_MSG, as
the message will be lost for good. The work round is to let pymqi
allocate the buffer for you, or specify a big enough buffer, or stop
accepting truncated messages. The latter options don't work too well,
as you can't find out how big the actual message is!
Pymqi unpacks the the MsgId and CorrelId
fields of the MQMD structure as a '24s', regardless of
what you put in it. This leads to some asymmetry when sending integers
with these field. e.g., On the put side:
And on the get side:md.MsgId = struct.pack('l', 42)
msgId = struct.upack('l', md.MsgId[0:4])[0]
Sami Salkosuo of IBM has written a developerWorks article about pymqi here. It gives a good overview and includes several good example applications.
Pascal Gauthier
John Kittel
Yves Lepage
John OSullivan
Tim Couper
Maas-Maarten Zeeman
Rich LaMarche
Kevin Kalbfleisch
Mauricio Strello
Brian Vicente
Jaco Smuts
Dariusz Suchojad
|
Donate to Python MQI Interface |