next up previous contents
Next: Make your own Good Up: Useful recipes for the Previous: Useful recipes for the

Rerunning the Analysis

Read this if you would like to redo part of your analysis, e.g. if your run has crashed, or if you want to change some parameters.

In case you want to re-run the analysis with different parameters, run og_create but this time with a different "ogid" parameter. This will create a new tree under obs/ogid where all the new results will be stored. If the pipeline has crashed[*], in general it is safer to restart your analysis from scratch removing the obs/ogid directory and restarting from the og_create step.

In any case, we give below a set of recipes that can be useful.

Because of the group concept you cannot just delete the result you do not like and restart the pipeline. All results that were produced in the course of the analysis are linked to the group, and should be detached before you relaunch the script. To do this you can use the og_clean program, that will clean an Observation Group up to the level specified with parameter endLevel. All data structures with a level equal or prior to endLevel will be kept, while the data structure with a later level will be erased. For example, to run the spectral extraction (SPE level) you should clean from the group whatever comes after the BIN_S level, as this is the level immediately preceeding the spectral one (see Fig.12).

 og_clean ogDOL="og_ibis.fits" endLevel="BIN_S"
If og_clean fails it could be due to the fact that the group was corrupted. You should try to fix it with dal_clean program
 dal_clean inDOL="og_ibis.fits" checkExt="1" backPtrs="1" checkSum="1"
and launch og_clean only afterwards.

Unfortunately the current version of og_clean is very slow. In some cases, if you know exactly which data structure has to be detached it would be much faster to launch dal_detach. For example if you run your analysis till SPE, or LCR level, and would like to produce a mosaic image afterwards, you do not have to clean the group, deleting all your results, but just have to detach ISGRI-SRCL-RES data structure (note that with the option "delete=y" all files with this data structure will be deleted):

 
 dal_detach og_ibis.fits\ pattern=ISGR-SRCL-RES delete=y


next up previous contents
Next: Make your own Good Up: Useful recipes for the Previous: Useful recipes for the

2018-11-20