More descriptive DFS progress bar

[bschreck]

Updated DFS progress bar to display number of features actually constructed, as well features removed at the end due to restrictions on variable types and the drop_contains/drop_exact list.

[kmax12]

what does this look like when it gets written?

I like the main change, but not sure it's that helpful to have this print out as things are removed.

[bschreck]

It actually looks pretty good I think, try running it and check it out.

Basically the progress bar is like:

[45 features] 100 features/s

Eventually the feature building stops and the progress bar stops, then above the progress bar the number removed is written:

Removed 2 features with undesired variable types
Removed 2 features from drop_contains list
[45 features] 100 features/s

[codecov-io]

Codecov Report

Merging #69 into master will increase coverage by 0.02%.
The diff coverage is 33.33%.

@@            Coverage Diff             @@
##           master      #69      +/-   ##
==========================================
+ Coverage    88.3%   88.33%   +0.02%     
==========================================
  Files          73       73              
  Lines        7447     7439       -8     
==========================================
- Hits         6576     6571       -5     
+ Misses        871      868       -3

Impacted Files	Coverage Δ
featuretools/synthesis/deep_feature_synthesis.py	93.06% <0%> (+0.95%)	⬆️
...computational_backends/calculate_feature_matrix.py	96.88% <100%> (+0.01%)	⬆️
featuretools/utils/gen_utils.py	64.44% <0%> (-2.23%)	⬇️

---

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 959069d...d219a0a. Read the comment docs.

[Seth-Rothschild]

Here's a screenshot.

The top three lines are new and give some information about what DFS has done. I think that's probably a step in the right direction.

It might be worth modifying some of the words around the pbar itself through tqdm. The line

1/1 [00:03<00:00,  3.16s/cutoff time]

probably doesn't give enough information to a new user. It presupposes they know that the 1/1 is counting the cutoff times they've given, and that they don't parse < to mean "elapsed time is less than remaining time".

How about something like

Elapsed: 1:30 | Remaining 4:10 |+-bar-here-----| 
Calculated matrix at 3/12 cutoff times (avg. 30s/calc): 

[bschreck]

Yup that all makes sense. I didn't test the actual feature calculation step, just the building step (because I noticed what we show for that step right now doesn't make much sense).

[Seth-Rothschild]

Seems difficult to get multiple lines into the bar_format in tqdm. Here's a single line version

[bschreck]

Updated to include feature exclusion line inside of pbar:

[bschreck]

Now looks like this:

[Seth-Rothschild]

I'm really enjoying using this.

A small note is that the order/phrasing of the drop contains message is a little confusing:

Here there are 18 (12 dropped + 6 excluded) features which aren't calculated. Is it possible to add the 12 into features_excluded or to move the "Removed 12 features..." line to after the calculation?

[bschreck]

I didn't see that line- should be fixed now

[bschreck]

Now with removed pbar from Deep Feature Synthesis.

In progress:

Complete:

[kmax12]

Looks great!