Update accessors to store weak reference to data

[thehomebrewnerd]

• Update accessors to store weak reference to data
• Closes Use weak references to refer to the original data in the accessor #880

[codecov]

Codecov Report

Merging #894 (b467f81) into main (637a019) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##              main      #894   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           44        44           
  Lines         6878      6887    +9     
=========================================
+ Hits          6878      6887    +9     

Impacted Files	Coverage Δ
woodwork/column_accessor.py	100.00% <100.00%> (ø)
woodwork/table_accessor.py	100.00% <100.00%> (ø)
woodwork/tests/accessor/test_column_accessor.py	100.00% <100.00%> (ø)
woodwork/tests/accessor/test_table_accessor.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 637a019...b467f81. Read the comment docs.

[freddyaboulton]

@thehomebrewnerd I think this implementation solves the original issue!

I visualized the object graph with this branch compared to the latest release and the accessor no longer has a reference to the original dataframe:

This branch

0.3.0 release

The memory footprint looks a lot better in evalml. Here is the memory consumption of automl search on the NYC taxi dataset (1.5 million rows)

This branch

0.3.0 release

I've also run our full set of benchmark datasets and we see similar improvements in memory without sacrificing accuracy/performance.

So thank you! 👏

[tamargrey]

Looks good! One comment about something that I was a little confused about, but feel free to ignore if you think adding a second example will be redundant!