[Cbc] Message printing with CbcSolver

[Stefan Vigerske]

Hi,

> mipMessageHandler->setLogLevel(0,1);

I additionally set
    msghandler->setLogLevel(1,0);
    msghandler->setLogLevel(2,0);
    msghandler->setLogLevel(3,0);

For an LP, I do
    msghandler->setLogLevel(1,1);
    msghandler->setLogLevel(2,0);
    msghandler->setLogLevel(3,0);
instead.

The problem is that if you pass in your own message handler, then the 
same one is used for all components, Cbc, Clp, and OsiClp.
So at the point where Cbc sets the loglevel for Clp to 0, it actually do 
so everywhere.
Then John implemented something so you can specify loglevel for 
different components, that's what you see above.
setLogLevel(0,x) should set the loglevel for Cbc to x.
I guess 1 is Clp, that's why I turn it on for LP.
I don't remember what 2 and 3 could be.

Stefan


>
> On Sun, 2011-03-13 at 03:43 +0000, Kish Shen wrote:
>> Hi,
>>
>> I have looked at this a little more, and it seems that the messages are
>> not printing because of the passing in of the message handler:
>>
>>   >  DerivedHandler* mipMessageHandler = new DerivedHandler;
>>   >  // model->passInMessageHandler(mipMessageHandler);
>>
>> If I comment out the line that pass the message handler in as above, the
>> messages from Cbc are printed. This happens in both the Trunk branch and
>> stable 2.4 branch I am using.
>>
>> My DerivedHanlder is defined as:
>>
>> class DerivedHandler : public CoinMessageHandler
>> {
>> public:
>>       virtual int print();
>> };
>>
>> int DerivedHandler::print()
>> {
>>     printf("@@@@@HELLO@@@@\n");
>>       int id = currentMessage().externalNumber();
>>       int mtype = (id<3000 ? LogType : (id<6000 ? WrnType : ErrType));
>>       eclipse_out(mtype, messageBuffer());
>>
>>       return 0;
>> }
>>
>> I added the printf line so that I know when the handler is called.
>> With the handler passed in, all I get printed is:
>>
>> @@@@@HELLO@@@@
>> Coin Cbc and Clp Solver version 2.4, build Mar 12 2011
>> @@@@@HELLO@@@@
>> command line - eplexcbcclpsolver -preprocess on -solve -quit (default
>> strategy 1)
>> @@@@@HELLO@@@@
>> Option for preprocess changed from sos to on
>> @@@@@HELLO@@@@
>> 0 fixed, 0 tightened bounds, 4 strengthened rows, 0 substitutions
>> @@@@@HELLO@@@@
>> 0 fixed, 0 tightened bounds, 3 strengthened rows, 0 substitutions
>> @@@@@HELLO@@@@
>> 0 fixed, 0 tightened bounds, 2 strengthened rows, 0 substitutions
>> @@@@@HELLO@@@@
>> 0 fixed, 0 tightened bounds, 1 strengthened rows, 0 substitutions
>> @@@@@HELLO@@@@
>> 0 fixed, 0 tightened bounds, 1 strengthened rows, 0 substitutions
>> @@@@@HELLO@@@@
>> 0 fixed, 0 tightened bounds, 1 strengthened rows, 0 substitutions
>> @@@@@HELLO@@@@
>> 0 fixed, 0 tightened bounds, 1 strengthened rows, 0 substitutions
>> @@@@@HELLO@@@@
>> processed model has 290 rows, 1055 columns (1055 integer) and 2465 elements
>>
>> When I *don't* pass the handler in, I get a lot more output:
>>
>> Coin Cbc and Clp Solver version 2.4, build Mar 12 2011
>> command line - eplexcbcclpsolver -preprocess on -solve -quit (default
>> strategy 1)
>> Option for preprocess changed from sos to on
>> Continuous objective value is 156083 - 0.00 seconds
>> Cgl0003I 0 fixed, 0 tightened bounds, 4 strengthened rows, 0 substitutions
>> Cgl0003I 0 fixed, 0 tightened bounds, 3 strengthened rows, 0 substitutions
>> Cgl0003I 0 fixed, 0 tightened bounds, 2 strengthened rows, 0 substitutions
>> Cgl0003I 0 fixed, 0 tightened bounds, 1 strengthened rows, 0 substitutions
>> Cgl0003I 0 fixed, 0 tightened bounds, 1 strengthened rows, 0 substitutions
>> Cgl0003I 0 fixed, 0 tightened bounds, 1 strengthened rows, 0 substitutions
>> Cgl0003I 0 fixed, 0 tightened bounds, 1 strengthened rows, 0 substitutions
>> Cgl0004I processed model has 290 rows, 1055 columns (1055 integer) and
>> 2465 elements
>> Cbc0038I Pass   1: suminf.    0.66204 (6) obj. 228065 iterations 86
>> Cbc0038I Pass   2: suminf.    0.66204 (6) obj. 228065 iterations 0
>> ....
>> Cbc0035I Maximum depth 12, 3374 variables fixed on reduced cost
>> Cuts at root node changed objective from 196138 to 394807
>> Probing was tried 184 times and created 309 cuts of which 44 were active
>> after adding rounds of cuts (0.220 seconds)
>> Gomory was tried 182 times and created 993 cuts of which 48 were active
>> after adding rounds of cuts (0.315 seconds)
>> Knapsack was tried 182 times and created 372 cuts of which 51 were
>> active after adding rounds of cuts (0.320 seconds)
>> Clique was tried 69 times and created 0 cuts of which 0 were active
>> after adding rounds of cuts (0.004 seconds)
>> MixedIntegerRounding2 was tried 182 times and created 193 cuts of which
>> 6 were active after adding rounds of cuts (0.022 seconds)
>> FlowCover was tried 69 times and created 0 cuts of which 0 were active
>> after adding rounds of cuts (0.005 seconds)
>> TwoMirCuts was tried 69 times and created 354 cuts of which 0 were
>> active after adding rounds of cuts (0.089 seconds)
>> Result - Finished objective 405935.1799999999 after 78 nodes and 5281
>> iterations - took 2.11 seconds (total time 2.11)
>> Total time 2.12
>> Segmentation fault
>>
>> I have cut out most of the printing in the middle, but it looks like
>> none of the output from Cbc are sent to the handler, only the messages
>> from Cgl at the start are printed (though without the "Cgl*" identifier
>> at the start, I do get this identifier when the handler prints messages
>> from Clp when a linear problem is solved.
>>
>> There is also a segmentation fault at the end when I don't pass in the
>> handler. Besides commenting out the line as I show, the rest of the code
>> should be the same.
>>
>> Any idea why this is happening, and if there is anyway to work around
>> this? I need to use the print handler to redirect the messages to my own
>> output stream, so not using a print handler is not really an option.
>>
>> Thanks and cheers,
>>
>> Kish
>>
>> On 11/03/2011 17:55, Kish Shen wrote:
>>> Hi,
>>>
>>> Currently, during the solving of a MIP problem in my code, where CBC is
>>> called using the CbcMain0/CbcMain1 interface (which invokes CbcSolver, I
>>> believe), I get no printing of the messages from Cbc.
>>>
>>> This is the behaviour I want normally, but recently some of our users
>>> have asked if they can see solver messages to get some idea of what the
>>> solver is doing. I have been unable to see any of Cbc's messages.
>>>
>>> As far as I can tell, Cbc is not calling my messageHandler -- I don't
>>> know if this is because I have not passed the handler to the right place.
>>>
>>> Here is an outline of the code I use, and if anyone can see what I need
>>> to do to get the Cbc messages, I would be very grateful:
>>>
>>> CbcModel* model = new(...);
>>>
>>> CbcMain0(*model);
>>>
>>> DerivedHandler* mipMessageHandler = new DerivedHandler;
>>> model->passInMessageHandler(mipMessageHandler);
>>>
>>> model->solver()->setHintParam(OsiDoReducePrint, true, OsiHintTry);
>>>
>>> const char* cbc_args[5];
>>>
>>>
>>> cbc_args[0] = "eplexcbcclpsolver";
>>> cbc_args[1] ="-preprocess"
>>> cbc_args[2] = "on";
>>> cbc_args[3] = "-solve";
>>> cbc_args[4] = "-quit";
>>> CbcMain1(5, cbc_args,*model,callBack);
>>>
>>> [I hope there are no typos in the above; I was not able to cut and paste
>>> from my VNC window with the code listing to my Windows mailer]
>>>
>>> DerivcedHandler is a derived class of CoinMessageHandler that sends the
>>> message to different streams depending on the externalNumber() of the
>>> message, and does seem to work (I can see Clp messages when solving a
>>> linear problem).
>>>
>>> An aside: I remember John (Forrest) suggesting the use of
>>> CbcMain0/CbcMain1 to use the CbcSolver. I just tried to look up the
>>> documentation for them in Cbc's Doxygen documentation, but was not able
>>> to find them at all. Does anyone know where I can find the doucmentation?
>>>
>>> Thanks in advance for any help!
>>>
>>> --Kish Shen
>>>
>>
>>
>
>
> _______________________________________________
> Cbc mailing list
> Cbc at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/cbc
>