output MR

[slosar]

Another fairly trivial fix whereby I also save output in a machinereadable format so that I don't have to write script after scipt that parses log files. Really used for example when fitting e.g. correlation function.

[dkirkby]

This functionality is now handled by the latest updates to likely and baofit, using two new FunctionMinimum methods that parallel the existing saveData and saveInverseCovariance for BinnedData:

    // Saves our parameter values and diagonal errors in plain text to the specified stream,
    // using full double precision. The format is a list of "index value error" lines, where
    // index values start at zero and include floating and fixed parameters, and error = 0
    // for fixed parameters.
    void saveParameters(std::ostream &os, bool onlyFloating = false) const;
    // Saves our floating-parameter covariance matrix in plain text to the specified stream,
    // use full double precision. The format is a list of "index1 index2 value" lines,
    // where value is the scaled covariance matrix element for (index1,index2), with indexing
    // that matches saveParameterValues(). Lines with value==0 or index2 < index1 are not
    // written to the file. Throws a RuntimeError if the covariance is not positive-definite.
    void saveFloatingParameterCovariance(std::ostream &os, double scale = 1) const;

baofit now always dumps two files using these two methods for the initial fit:

  outputPrefix + "save.pars"
  outputPrefix + "save.pcov"