ctools

When running cterror with the flag debug=yes, I get the following output.

2016-02-19T16:29:05: +=======================================================+
2016-02-19T16:29:05: | Compute error for source "Crab" parameter "Prefactor" |
2016-02-19T16:29:05: +=======================================================+
2016-02-19T16:29:05:  Confidence level ..........: 68%
2016-02-19T16:29:05:  Log-likelihood difference .: 0.494473
2016-02-19T16:29:05:  Initial factor range ......: [1.8536, 21.8576]
2016-02-19T16:29:11:  Lower parameter factor ....: 10.9497
2016-02-19T16:29:16:  Upper parameter factor ....: 12.8348
2016-02-19T16:29:16:  Error from curvature ......: 1.0002e-16 ph/cm2/s/MeV
2016-02-19T16:29:16:  Error from profile ........: 9.42573e-17 ph/cm2/s/MeV
2016-02-19T16:29:16:  Negative profile error ....: 9.05944e-17 ph/cm2/s/MeV
2016-02-19T16:29:16:  Positive profile error ....: 9.79201e-17 ph/cm2/s/MeV
2016-02-19T16:29:16: 
2016-02-19T16:29:16: +===================================================+
2016-02-19T16:29:16: | Compute error for source "Crab" parameter "Index" |
2016-02-19T16:29:16: +===================================================+
2016-02-19T16:29:16:  Confidence level ..........: 68%
2016-02-19T16:29:16:  Log-likelihood difference .: 0.494473
2016-02-19T16:29:16:  Initial factor range ......: [2.02477, 3.06928]
2016-02-19T16:29:20:  Lower parameter factor ....: 2.49909
2016-02-19T16:29:23:  Upper parameter factor ....: 2.59599
2016-02-19T16:29:23:  Error from curvature ......: 0.0522255 
2016-02-19T16:29:23:  Error from profile ........: 0.0484514 
2016-02-19T16:29:23:  Negative profile error ....: 0.0479414 
2016-02-19T16:29:23:  Positive profile error ....: 0.0489614 
2016-02-19T16:29:23: 
2016-02-19T16:29:23: +==============+
2016-02-19T16:29:23: | Save results |
2016-02-19T16:29:23: +==============+
Output model XML file [result.xml]

1. Providing the output file usually happens at the beginning of the script execution. I therefore suggest to query for the outfile already in cterror::get_parameters (and not in cterror::save).
2. cterror in fact saves the source model. However, the user could expect that asymmetric errors are written in the model XML file, too (which is not the case). It might be more difficult to implement that asymmetric errors get written? We would need to think about names for these attributes “error_neg” and “error_pos” could be an option?