Closed msdogan closed 8 years ago
@msdogan ok, I have to admit some ignorance with the --debug flag. So for debug --ALL you would like to see something like:
SR_ISB.2002-08-31,DBUGSNK.2002-08-31,0,0,1,0,1000000 DBUGSRC.2002-08-31,SR_ISB.2002-08-31,0,0,1,0,1000000
? If so, is 1 for amplitude correct?
@jrmerz yes, this is correct. k=0, cost=0, a=1, lb=0, ub=1000000. And also links
SOURCE DBUGSNK.2002-08-31,0,2000000,1,0,1000000
and
DBUGSNK.2002-08-31 SINK,0,2000000,1,0,1000000
These are source and sink links that we apply very high penalty. But, we apply high penalties only on these links. SR_ISB.2002-08-31,DBUGSNK.2002-08-31,0,0,1,0,1000000
and
DBUGSRC.2002-08-31,SR_ISB.2002-08-31,0,0,1,0,1000000
will have zero cost as you said.
There is also some other issues, for example there is a inflow to D43a (https://cwn.casil.ucdavis.edu/#info/D43a), but it does not show up in the exported matrix. I couldn't find a reason for that. One drawback right now is that, Pyomo shows one error at a time. When it finds an error, it stops. Maybe, I should wait until Debug flag to attack all these issues.
Ya, about to have that flag ready for you. If you don't think the missing D43a is part of this issue, go ahead a create a new one.
Great! No missing D43 inflow is not part of this issue. I just wanted to say :) So, new calvin-network-tools version for it? 1.0.38?
Just published. --debug is available for the matrix @ v1.0.38
This is working perfectly. However, could you please delete time information on SOURCE for SOURCE.2002-08-31 DBUGSRC.2002-08-31 0 2000000 1 0 10000000
and
SOURCE.2002-09-30 DBUGSRC.2002-09-30 0 2000000 1 0 10000000
So, correct representation would be:
SOURCE DBUGSRC.2002-08-31 0 2000000 1 0 10000000
and
SOURCE DBUGSRC.2002-09-30 0 2000000 1 0 10000000
Thanks
Ok, date is removed from SOURCE @ v1.0.39
@jrmerz okay, this is working. Debug mode will save a lot of time. Thanks.
Ooops. I got a FATAL ERROR when I tried to export all 82-year period in debug mode with cnf matrix --format csv --start 1921-10 --stop 2003-09 --ts . --to networklinks --max-ub 10000000 --outnodes networknodes --debug All
ok, ya, this will take a bit. Going to have to request more memory at runtime, but even then going to hit some other walls. I'll let you know when I have fixes.
@msdogan give it a go. I fixed memory issue on linux and mac. Should work on windows, but let me know.
@jrmerz this exports the 82-year matrix to .csv
but it is giant (300MB). And I can't even open it. It says "File not loaded completely" I wonder if there is a way to reduce the size. Maybe we can do that right now, there is DBUGSRC and DBUGSINK nodes for each time step, DBUGSRC.2002-09-30
, DBUGSNK.2002-09-30
. If we create only one DBUGSRC
and DBUGSINK
without time-step information, and connect nodes to those, we can eliminate 2000 links. What do you think?
@jrmerz Update: let's leave time-step information there, otherwise I would need to update post-processors. Besides, 2000 links is nothing compared to the rest.
@jrmerz can I close the issue?
Ya, there are current 5,185,762 so removing those would be a 0.03% file reduction.
I would recommend running with a sub-network if you want to run over 82 year period.
And sure if you think it's go to go.
yeah it is what it is. 82-year run file w/o debug mode is also big. About 200MB
FYI, bash tools like less, more and tail should let you inspect the file. They should be available on windoz via the git-bash command prompt.
@jrmerz Similar to debug mode in HEC-PRM run from HOBBES network with a command
prm build --prefix test --debug All
, I see a great value on adding DEBUG MODE to matrix export. That way, we should be able to identify issues all at ones.