[MRG] EHN Add a FunctionSampler

[glemaitre]

Reference Issue

related to #9360 in scikit-learn bug tracker

What does this implement/fix? Explain your changes.

Any other comments?

[codecov]

Codecov Report

Merging #342 into master will increase coverage by 0.02%.
The diff coverage is 100%.

@@            Coverage Diff            @@
##           master    #342      +/-   ##
=========================================
+ Coverage   98.77%   98.8%   +0.02%     
=========================================
  Files          67      69       +2     
  Lines        3926    4001      +75     
=========================================
+ Hits         3878    3953      +75     
  Misses         48      48

Impacted Files	Coverage Δ
imblearn/tests/test_misc.py	100% <100%> (ø)
imblearn/utils/estimator_checks.py	96% <100%> (+0.14%)	⬆️
imblearn/misc.py	100% <100%> (ø)
imblearn/tests/test_common.py	94.73% <0%> (-0.72%)	⬇️
imblearn/over_sampling/adasyn.py	98.5% <0%> (ø)	⬆️
imblearn/_version.py	100% <0%> (ø)	⬆️
imblearn/ensemble/classifier.py	100% <0%> (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 3877493...792c29d. Read the comment docs.

[chkoar]

What's the status of this? Can we maintain the coverage and merge it?

[glemaitre]

I have to check why the sampler is complaining. I think that I should also add a gallery example and the documentation.

[glemaitre]

@chkoar We need to find a better location for this than misc. We need to find a name related to user usage rather than developer naming.

[glemaitre]

As we mentioned earlier, we can use this for fast prototyping a new balancing method or getting a bit off topic and use it to make any sampling type (resampling) and use it in pipeline.

[glemaitre]

@chkoar @massich This is almost good to be merged. We need to decide a nice name for the submodule.

[chkoar]

We need to decide a nice name for the submodule.

This is the most difficult part of this PR.

[glemaitre]

what about custom?

[glemaitre]

or customized

[glemaitre]

I think that the proposition to make it available at root is the best:

from imblearn import FunctionSampler