Change imports to avoid pylint namespace errors

[c24t]

This is 9d3ef06d split off from #282 at @mauriciovasquezbernal's request.

See #282 (comment).

Edit: changed to use pylint:ignore=no-member in the modules failing lint instead. These errors are almost never meaningful, and this prevents us from having to play type inference whack-a-mole in the future if these imports change.

[codecov-io]

Codecov Report

Merging #376 into master will increase coverage by 0.14%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #376      +/-   ##
==========================================
+ Coverage   85.08%   85.22%   +0.14%     
==========================================
  Files          38       38              
  Lines        1897     1916      +19     
  Branches      225      226       +1     
==========================================
+ Hits         1614     1633      +19     
  Misses        218      218              
  Partials       65       65

Impacted Files	Coverage Δ
...src/opentelemetry/ext/opentracing_shim/__init__.py	95.93% <100%> (+0.17%)	⬆️
...app/src/opentelemetry_example_app/flask_example.py	100% <0%> (ø)	⬆️
...emetry-sdk/src/opentelemetry/sdk/trace/__init__.py	90.94% <0%> (+0.06%)	⬆️
...xt-jaeger/src/opentelemetry/ext/jaeger/__init__.py	85.18% <0%> (+0.18%)	⬆️
...ts/src/opentelemetry/ext/http_requests/__init__.py	89.74% <0%> (+0.26%)	⬆️
opentelemetry-sdk/src/opentelemetry/sdk/util.py	85.88% <0%> (+0.34%)	⬆️
...opentelemetry/sdk/context/propagation/b3_format.py	87.27% <0%> (+0.48%)	⬆️
...ry-ext-wsgi/src/opentelemetry/ext/wsgi/__init__.py	68.18% <0%> (+0.58%)	⬆️
...ry-sdk/src/opentelemetry/sdk/resources/__init__.py	70.83% <0%> (+2.65%)	⬆️

---

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 ccb97e5...01619bc. Read the comment docs.

[mauriciovasquezbernal]

I tried it on top of #282 (after reverting the previous changes to avoid these errors) and it works good.

[Oberon00]

no-member is IMHO often a very useful warning, so it's a pity that it has false positives here. Do we know the reason? I think we can merge this as a quick & dirty workaround to unblock us, but this should really be resolved. Maybe we need to check the tests in separate pylint invocations each, to not get name clashes that confuse pylint (similar to the http_requests module name debacle #94).

[c24t]

Maybe we need to check the tests in separate pylint invocations each, to not get name clashes that confuse pylint (similar to the http_requests module name debacle #94).

I think that's exactly what's happening. On 5a224ad pylint ext/opentelemetry-ext-opentracing-shim/**/*.py passes as expected, but the big omnibus pylint invocation fails because (evidently) it's picking up the opentracing module in examples/ instead.

Worse, it looks like this happens on class name collisions too. AFAICT lint fails on fec0fbe because test_shim.py does from opentracing import Span, but an earlier import imports Span from opentelemetry.trace. One more argument for importing modules instead of classes.

[lzchen]

LGTM