ryanmklee / elisa_fanclub_part2 Goto Github PK

So currently, we get profiler data we run it through the main block. Would it be possible to do it through command line instead?

When this scales up, ensure that we have an output folder with the graphs so we don't clutter our main directory.

When we start doing code profiling for multiple input size, we want to modify our barchart to construct multiple bars showing the differences in runtime per value of input size.

Possibly another way for us to visualize our profiled data.

https://plot.ly/python/bubble-charts/

Yet again: this does not influence your grade. It's something that I do in both 310 and 410 so we can discuss code design and learn things outside of the scope of what is being evaluated. That being said, let's roll:

Naming conventions:

Why have a prefix viz_ for things which are inside the viz module? It adds redundancy and the usage of the underscore is discouraged.

viz_factory.py

My biggest issue if the lack of a init method. and how the render works. The variable chart_type could be easily moved to an init (I'm not sure about the others) but it seems that the most important variable here is just iteration_data

Try to use more private methods. It will facilitate easy of readiness

if track: 
    self.__get_data_from ....
else:
    self.__get_data_from ....

Another example:

for fn in new_fns:
            for i in range(idx):
                backfill_value = self.get_line(all_lines[i], fn)
                iter_lines[i].append(backfill_value)

could be:

__append_backfill_values(new_fns, idx)

viz_utils.py

Avoid magic numbers. what 3, 1 and the others numbers represent?

builder module: Awesome usage of abstractions/inheritance!

I hope some of these comments make sense. Feel free to ignore any comments that don't.
Awesome project and good luck with your finals!

In this code block, we are only using one preset colour for the bar graph. We should generate a gradient or something to indicate the colour.

for i in range(0, len(x_data[0])): for xd, yd in zip(x_data, y_data): traces.append(go.Bar( x=[xd[i]], y=[yd], orientation='h', marker=dict( # TODO: create colour intensity, not just taking the first color based on x values color=colors[0], line=dict( color='rgb(248, 248, 249)', width=1) ) ))

In the profiler's code, we need to extract the function name not all the other details so our graph doesn't have a large overflow.

For example, using the profiler the function name is

{method 'query' of 'sklearn.neighbors.kd_tree.BinaryTree' objects}

we want instead to be able to clean this data

to return

'sklearn.neighbors.kd_tree.BinaryTree'

Bar chart data filtering is done inside it's class, we should probably move it outside of the class to abstract it for other graphs and extract different data, then pass in the relevant data to each graph we will have. We will need to move the function _filter_data out from the BarGraph class and into the main function or somewhere else.