[MRG+1] Change named_steps to Bunch object

[herilalaina]

Reference Issue

Fix #8481

What does this implement/fix? Explain your changes.

This PR changes pipeline.named_steps into bunch object instead of dictionary.

Any other comments?

Usage example

> clf = svm.SVC(kernel='linear')
> anova_svm = Pipeline([('anova', anova_filter), ('svc', clf)])
> anova_svm.named_steps.anova
SelectKBest(k=5, score_func=<function f_regression at 0x7f1d984d8620>)

To do

• Update source code
• Test
• Documentation

[codecov]

Codecov Report

Merging #8586 into master will increase coverage by <.01%.
The diff coverage is 99.56%.

@@            Coverage Diff             @@
##           master    #8586      +/-   ##
==========================================
+ Coverage   95.49%   95.49%   +<.01%     
==========================================
  Files         342      342              
  Lines       61072    61088      +16     
==========================================
+ Hits        58318    58334      +16     
  Misses       2754     2754

Impacted Files	Coverage Δ
sklearn/feature_selection/univariate_selection.py	99.46% <ø> (ø)	⬆️
sklearn/linear_model/coordinate_descent.py	96.94% <ø> (ø)	⬆️
sklearn/ensemble/forest.py	98.16% <ø> (ø)	⬆️
sklearn/metrics/classification.py	97.77% <ø> (ø)	⬆️
sklearn/mixture/tests/test_gaussian_mixture.py	99.63% <ø> (ø)	⬆️
sklearn/model_selection/_validation.py	98.71% <ø> (ø)	⬆️
sklearn/tree/tree.py	98.41% <ø> (ø)	⬆️
sklearn/calibration.py	98.87% <ø> (ø)	⬆️
sklearn/neighbors/approximate.py	98.95% <ø> (ø)	⬆️
sklearn/mixture/gaussian_mixture.py	100% <ø> (ø)	⬆️
... and 62 more

---

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 134c9b7...00e67c9. Read the comment docs.

[codecov]

Codecov Report

Merging #8586 into master will increase coverage by <.01%.
The diff coverage is 99.56%.

@@            Coverage Diff             @@
##           master    #8586      +/-   ##
==========================================
+ Coverage   95.49%   95.49%   +<.01%     
==========================================
  Files         342      342              
  Lines       61072    61088      +16     
==========================================
+ Hits        58318    58334      +16     
  Misses       2754     2754

Impacted Files	Coverage Δ
sklearn/feature_selection/univariate_selection.py	99.46% <ø> (ø)	⬆️
sklearn/linear_model/coordinate_descent.py	96.94% <ø> (ø)	⬆️
sklearn/ensemble/forest.py	98.16% <ø> (ø)	⬆️
sklearn/metrics/classification.py	97.77% <ø> (ø)	⬆️
sklearn/mixture/tests/test_gaussian_mixture.py	99.63% <ø> (ø)	⬆️
sklearn/model_selection/_validation.py	98.71% <ø> (ø)	⬆️
sklearn/tree/tree.py	98.41% <ø> (ø)	⬆️
sklearn/calibration.py	98.87% <ø> (ø)	⬆️
sklearn/neighbors/approximate.py	98.95% <ø> (ø)	⬆️
sklearn/mixture/gaussian_mixture.py	100% <ø> (ø)	⬆️
... and 62 more

---

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 134c9b7...00e67c9. Read the comment docs.

[jnothman]

Thanks. please add tests and documentation, and prefix the issue title with [WIP] until those are done

…

On 15 Mar 2017 9:29 am, "codecov[bot]" ***@***.***> wrote: Codecov <https://codecov.io/gh/scikit-learn/scikit-learn/pull/8586?src=pr&el=h1> Report Merging #8586 <https://codecov.io/gh/scikit-learn/scikit-learn/pull/8586?src=pr&el=desc> into master <https://codecov.io/gh/scikit-learn/scikit-learn/commit/4ab99c75b4486f5e1a67be730b7afa6d9551ed7e?src=pr&el=desc> will *increase* coverage by <.01%. The diff coverage is 100%. @@ Coverage Diff @@## master #8586 +/- ## ==========================================+ Coverage 95.48% 95.48% +<.01% ========================================== Files 342 342 Lines 61034 61035 +1 ==========================================+ Hits 58280 58281 +1 Misses 2754 2754 Impacted Files <https://codecov.io/gh/scikit-learn/scikit-learn/pull/8586?src=pr&el=tree> Coverage Δ sklearn/pipeline.py <https://codecov.io/gh/scikit-learn/scikit-learn/compare/4ab99c75b4486f5e1a67be730b7afa6d9551ed7e...60a87763ec1ac8c61edb03516ebc16113c073895?src=pr&el=tree#diff-c2tsZWFybi9waXBlbGluZS5weQ==> 99.26% <100%> (ø) ✅ ------------------------------ Continue to review full report at Codecov <https://codecov.io/gh/scikit-learn/scikit-learn/pull/8586?src=pr&el=continue> . *Legend* - Click here to learn more <https://docs.codecov.io/docs/codecov-delta> Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov <https://codecov.io/gh/scikit-learn/scikit-learn/pull/8586?src=pr&el=footer>. Last update 4ab99c7...60a8776 <https://codecov.io/gh/scikit-learn/scikit-learn/compare/4ab99c75b4486f5e1a67be730b7afa6d9551ed7e...60a87763ec1ac8c61edb03516ebc16113c073895?src=pr&el=footer&el=lastupdated>. Read the comment docs <https://docs.codecov.io/docs/pull-request-comments> . — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#8586 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAEz698gCKrmpJUVUrb6q4cdyjAPRt68ks5rlxS6gaJpZM4MdNFz> .

[jnothman]

Thanks. please add tests and documentation, and prefix the issue title with [WIP] until those are done

…

On 15 Mar 2017 9:29 am, "codecov[bot]" ***@***.***> wrote: Codecov <https://codecov.io/gh/scikit-learn/scikit-learn/pull/8586?src=pr&el=h1> Report Merging #8586 <https://codecov.io/gh/scikit-learn/scikit-learn/pull/8586?src=pr&el=desc> into master <https://codecov.io/gh/scikit-learn/scikit-learn/commit/4ab99c75b4486f5e1a67be730b7afa6d9551ed7e?src=pr&el=desc> will *increase* coverage by <.01%. The diff coverage is 100%. @@ Coverage Diff @@## master #8586 +/- ## ==========================================+ Coverage 95.48% 95.48% +<.01% ========================================== Files 342 342 Lines 61034 61035 +1 ==========================================+ Hits 58280 58281 +1 Misses 2754 2754 Impacted Files <https://codecov.io/gh/scikit-learn/scikit-learn/pull/8586?src=pr&el=tree> Coverage Δ sklearn/pipeline.py <https://codecov.io/gh/scikit-learn/scikit-learn/compare/4ab99c75b4486f5e1a67be730b7afa6d9551ed7e...60a87763ec1ac8c61edb03516ebc16113c073895?src=pr&el=tree#diff-c2tsZWFybi9waXBlbGluZS5weQ==> 99.26% <100%> (ø) ✅ ------------------------------ Continue to review full report at Codecov <https://codecov.io/gh/scikit-learn/scikit-learn/pull/8586?src=pr&el=continue> . *Legend* - Click here to learn more <https://docs.codecov.io/docs/codecov-delta> Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov <https://codecov.io/gh/scikit-learn/scikit-learn/pull/8586?src=pr&el=footer>. Last update 4ab99c7...60a8776 <https://codecov.io/gh/scikit-learn/scikit-learn/compare/4ab99c75b4486f5e1a67be730b7afa6d9551ed7e...60a87763ec1ac8c61edb03516ebc16113c073895?src=pr&el=footer&el=lastupdated>. Read the comment docs <https://docs.codecov.io/docs/pull-request-comments> . — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#8586 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAEz698gCKrmpJUVUrb6q4cdyjAPRt68ks5rlxS6gaJpZM4MdNFz> .

[jnothman]

You should also test what happens when step names conflict with attributes of dict such as values.

[lesteve]

Travis is failing because you have some PEP8 violations, please invest some time to configure on-the-fly flake8 checks inside your editor of choice. This allows you to spot these kind of problems while editing.

[lesteve]

Travis is failing because you have some PEP8 violations, please invest some time to configure on-the-fly flake8 checks inside your editor of choice. This allows you to spot these kind of problems while editing.

[herilalaina]

Hi @jnothman @lesteve , thank you for your comment.
Actually, the Bunch object didn't work when step names conflict with attribute of dict like values, items ..
So I decide to override __getattribute__ on Bunch class.

[herilalaina]

Hi @jnothman @lesteve , thank you for your comment.
Actually, the Bunch object didn't work when step names conflict with attribute of dict like values, items ..
So I decide to override __getattribute__ on Bunch class.

[jnothman]

To be sure, correct behaviour would be to provide dict's keys method, even if a step is named 'keys'. We can't just break the existing dict behaviour. Nor can we accept an error if a user chooses keys for a step name; they just don't get the new functionality.

…

On 23 Mar 2017 7:20 am, "RAKOTOARISON Herilalaina" ***@***.***> wrote: Hi @jnothman <https://github.com/jnothman> @lesteve <https://github.com/lesteve> , thank you for your comment. Actually, the Bunch object didn't work when step names conflict with attribute of dict like values, items .. So I decide to override __getattribute__ on Bunch class. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#8586 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAEz6xqn2JCmA-Mv9RYXAp9dKM8wxrCRks5roYKCgaJpZM4MdNFz> .

[jnothman]

To be sure, correct behaviour would be to provide dict's keys method, even if a step is named 'keys'. We can't just break the existing dict behaviour. Nor can we accept an error if a user chooses keys for a step name; they just don't get the new functionality.

…

On 23 Mar 2017 7:20 am, "RAKOTOARISON Herilalaina" ***@***.***> wrote: Hi @jnothman <https://github.com/jnothman> @lesteve <https://github.com/lesteve> , thank you for your comment. Actually, the Bunch object didn't work when step names conflict with attribute of dict like values, items .. So I decide to override __getattribute__ on Bunch class. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#8586 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAEz6xqn2JCmA-Mv9RYXAp9dKM8wxrCRks5roYKCgaJpZM4MdNFz> .

[herilalaina]

To confirm what you said. I will remove the __getattribute__ in the Bunch class then change test_pipeline.py (line 524) with

pipeline = Pipeline([('values', transf), ("items", mult2)])
assert_true(pipeline.named_steps.values is not transf)
assert_true(pipeline.named_steps.items is not mult2)

[herilalaina]

To confirm what you said. I will remove the __getattribute__ in the Bunch class then change test_pipeline.py (line 524) with

pipeline = Pipeline([('values', transf), ("items", mult2)])
assert_true(pipeline.named_steps.values is not transf)
assert_true(pipeline.named_steps.items is not mult2)

[jnothman]

I still think this deserves a brief mention in doc/modules/pipeline.rst, where it says "and as a dict in named_steps", with something like:

Attributes of named_steps map to keys, enabling tab completion in interactive environments:
    >>> pipe.named_steps.reduce_dim is pipe.named_steps['reduce_dim']
    True

Otherwise LGTM. Please add an entry to what's new.

[jnothman]

I think this should mention pipeline

[herilalaina]

Hi, I tried to organize into section each entry in API change summary. This resulted the conflict. I resolved it but appveyor test still failed (without merge). Should I remove my last commit ?

[jnothman]

Don't worry about sections. We'll do that at release

[herilalaina]

Ok, thank you. I cleaned it.

[amueller]

Awesome, thanks!

[amueller]

Awesome, thanks!

[amueller]

I'm not sure I understand this sentence.