Collecting plot data [part of #180]

[mikhailnikolaev]

Plot script now collects plot data and then uses it for creating graphs. Also, it saves plot data to the JSON file.

This PR contains a part of changes from the closed PR #180. It is the second part of the list, described here https://github.com/awf/ADBench/pull/180#issuecomment-569066919

[mikhailnikolaev]

This looks great, but is there any way that the same data structure can be used for both serialisation and generating the graph? Is it possible to use a singled named tuple and write that out as JSON?

Will it be more convenient if the plot_data,json file has the following form:

[
    {
        "figure_info": {
            "build": build,
            "objective": objective,
            "function_type": function_type,
            "test_size": test_size
        },
        "values": [ ... ]
    }
]

Thus the script can use figure_info subdictionary instead of named tuple

[tomjaguarpaw]

Yes, I think that would clear up the biggest differences. It would be good if we could avoid preprocessing sorted_vals_by_tool too.

[mikhailnikolaev]

Yes, I think that would clear up the biggest differences. It would be good if we could avoid preprocessing sorted_vals_by_tool too.

Actually, I have the idea to use only items of the plot_data array everywhere in the script. I mean, we can even avoid such a term as sorted_vals_by_tool and use only dictionaries of the form

{ "figure_info": { ... }, "values": { ... } }

where "values" is a kind of sorted_vals_by_tools and 'figure_info" is figure_info. Thus, we can avoid any additional serialization actions except json.dump. But I don't know is it a good idea or there is a more pythonic way. What do you think?

[tomjaguarpaw]

Ideally the form that we store as JSON would be as close as possible to the form that is used by generate_graph. It seems like it would be nice not to unwrap and rewrap sorted_vals_by_tool. What do you think?

[tomjaguarpaw]

OK, I think your idea sounds good. Let's give it a go!

[mikhailnikolaev]

Ideally the form that we store as JSON would be as close as possible to the form that is used by generate_graph. It seems like it would be nice not to unwrap and rewrap sorted_vals_by_tool. What do you think?

This will be one of the positive corollaries of using dictionary everywhere instead of sorted_vals_by_tools

[mikhailnikolaev]

OK, I think your idea sounds good. Let's give it a go!

Should I do another PR, or I can add commits into this one?

[tomjaguarpaw]

Should I do another PR, or I can add commits into this one?

I think best to add commits into this one and see what it looks like. Then we can decide what to do.

[tomjaguarpaw]

I think best to add commits into this one and see what it looks like. Then we can decide what to do.

In fact, you can do whichever you prefer.

[tomjaguarpaw]

This is improving a lot.