fix: remove reductions import in top-level __init__.py

[akshayka]

Description

Reductions are stated to not be in the public API
(https://www.cvxpy.org/api_reference/cvxpy.reductions.html#disclaimer, #2104 (comment)), but they are imported in the top-level __init__.py.

This change removes that import.

Type of change

• New feature (backwards compatible)
• New feature (breaking API changes)
• Bug fix
• Other (Documentation, CI, ...)

[github-actions]

Benchmarks that have stayed the same:

   before           after         ratio
 [84398785]       [00daf0ef]
      31.1±0s          31.6±0s     1.02  cvar_benchmark.CVaRBenchmark.time_compile_problem
      9.57±0s          9.59±0s     1.00  huber_regression.HuberRegression.time_compile_problem
      4.11±0s          4.12±0s     1.00  simple_LP_benchmarks.SimpleFullyParametrizedLPBenchmark.time_compile_problem
      46.7±0s          46.6±0s     1.00  sdp_segfault_1132_benchmark.SDPSegfault1132Benchmark.time_compile_problem
      8.68±0s          8.65±0s     1.00  optimal_advertising.OptimalAdvertising.time_compile_problem
      11.3±0s          11.3±0s     0.99  svm_l1_regularization.SVMWithL1Regularization.time_compile_problem
      9.17±0s          9.10±0s     0.99  semidefinite_programming.SemidefiniteProgramming.time_compile_problem
      1.89±0s          1.87±0s     0.99  simple_LP_benchmarks.SimpleScalarParametrizedLPBenchmark.time_compile_problem
      24.7±0s          24.4±0s     0.99  simple_LP_benchmarks.SimpleLPBenchmark.time_compile_problem

[SteveDiamond]

LGTM

[SteveDiamond]

Is it possible we put the reductions import there because it helped build the sphinx API docs? Not a reason to keep it of course.

[rileyjmurray]

Honestly I have no idea how long reductions have been in there or for what reasons. Your guess is at least as good as mine.

…

On Sat, Apr 15, 2023 at 11:43 AM Steven Diamond ***@***.***> wrote: Is it possible we put the reductions import there because it helped build the sphinx API docs? Not a reason to keep it of course. — Reply to this email directly, view it on GitHub <#2105 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACRLIFELX2CYXWOP522TN5DXBLT3JANCNFSM6AAAAAAW7QL3QM> . You are receiving this because your review was requested.Message ID: ***@***.***>

[SteveDiamond]

I checked with git blame and it looks like it was related to the API docs:
19c5177
Anyway we can make the docs work easily enough.

[rileyjmurray]

When you said it was added in the API docs I started to sweat since I was the last person to revise those. I was relieved when I saw it was way back in the 0.4.x release series, lol.

…

On Sat, Apr 15, 2023 at 3:00 PM Steven Diamond ***@***.***> wrote: I checked with git blame and it looks like it was related to the API docs: 19c5177 <19c5177> Anyway we can make the docs work easily enough. — Reply to this email directly, view it on GitHub <#2105 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACRLIFA34F37IZJM5ESL4I3XBMK7RANCNFSM6AAAAAAW7QL3QM> . You are receiving this because your review was requested.Message ID: ***@***.***>

[akshayka]

I checked with git blame and it looks like it was related to the API docs: 19c5177 Anyway we can make the docs work easily enough.

We might as well remove them from the API docs, since they're not public.

[rileyjmurray]

@akshayka I think they should be documented on the website. Sometimes advanced users benefit from knowing what goes on under the hood. We can put them under a "private API" heading if that feels more appropriate.

[akshayka]

@akshayka I think they should be documented on the website. Sometimes advanced users benefit from knowing what goes on under the hood. We can put them under a "private API" heading if that feels more appropriate.

I think users who are that advanced should just read the source code. The risk in documenting private symbols on our website is that regular users will find these symbols and become dependent on them; people will just treat the private API as public. Similar to black.format_str: https://github.com/search?q=black.format_str&type=code, psf/black#779

That said, I don't feel strongly. I just think removing these symbols will make maintainers' lives easier.