Written output data is too large

[nikohansen]

even after the current modification of the logger as reported here. This is on the bbob-constrained testbed. Maybe the first thing to check is which logger is used and whether it is different from the one used with bbob (which should have been successfully fixed).

[brockho]

In my code, I just use the default code with the suite_name exchanged:

suite_name = "bbob-constrained"
[...]
observer = cocoex.Observer(suite_name,
            # see https://numbbo.github.io/coco-doc/C/#observer-parameters
            'result_folder: {0}  algorithm_name: {1}'.format(
                output_folder, fmin.__module__ + '.' + fmin.__name__))
[...]
    problem.observe_with(observer)  # generate data for cocopp

[nikohansen]

Right, so we need to check next in the (C) code which logger is invoked with this suite by default.

[brockho]

Looks like it is the bbob one:

} else if (0 == strncmp(observer_name, "bbob-constrained", 16)) {
    observer_bbob(observer, observer_options, &additional_option_keys);

see

coco-experiment/src/coco_observer.c

Line 760 in b80e048

} else if (0 == strncmp(observer_name, "bbob-constrained", 16)) {

[nikohansen]

Agreed, looks like it's the default logger which is also used for 'bbob'.

[nikohansen]

x-values threshold 9c44ba2 and target triggers e5d068f are amended to reduce data size.