make subscription data calls more efficient

[matiascognite]

Description

Use a more efficient method to query the subscription list data API

The timeout parameter will enable active wait:
Until there is more data available, we will keep the request waiting in the backend. Only when the timeout has expired, we will return.

Not a visible change to the user; no need to update version or changelog?

Checklist:

• Tests added/updated.
• Documentation updated. Documentation is generated from docstrings - these must be updated according to your change.
  If a new method has been added it should be referenced in cognite.rst in order to generate docs based on its docstring.
• Changelog updated in CHANGELOG.md.
• Version bumped. If triggering a new release is desired, bump the version number in _version.py and pyproject.toml per semantic versioning.

[codecov]

Codecov Report

Merging #1538 (cc8c35e) into master (58a389c) will increase coverage by 0.03%.
Report is 1 commits behind head on master.
The diff coverage is 100.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1538      +/-   ##
==========================================
+ Coverage   91.78%   91.82%   +0.03%     
==========================================
  Files         120      120              
  Lines       15623    15623              
==========================================
+ Hits        14340    14346       +6     
+ Misses       1283     1277       -6     

Files	Coverage Δ
cognite/client/_api/datapoints_subscriptions.py	100.00% <ø> (ø)
cognite/client/_version.py	100.00% <100.00%> (ø)

... and 3 files with indirect coverage changes

[matiascognite]

Thanks for review, even if I didn't ask for it @haakonvt !
Now that we actually implemented it in the backend, the tests pass, but I have some questions

• Given that we change the method signature, we should probably do a (minor) version bump?
• Is poll_timeout a good name? We have pollTimeoutSeconds in the API, but I have the impression that python defaults to seconds (correct me if I'm wrong)
• The behavior changes. If I read it correctly, it will now block while waiting for new data. Before it would just respond with an empty change immediately. It should mostly be an improvement though, and also protect against writing very heavy applications with constant polling, by accident.

[haakonvt]

@matiascognite

• Given that we changed the method signature, we should probably do a (minor) version bump?

Yup, I agree!

• Is poll_timeout a good name? We have pollTimeoutSeconds in the API, but I have the impression that python defaults to seconds (correct me if I'm wrong)

Sounds great 👌 Not sure what you mean with "python defaults to seconds" though? Your docstring description is very clear.

• The behavior changes. If I read it correctly, it will now block while waiting for new data. Before it would just respond with an empty change immediately. It should mostly be an improvement though, and also protect against writing very heavy applications with constant polling, by accident.

Yeah, for sure an improvement! Particularly for our pyodide users who might not know time.sleep doesn't work and thus unknowingly are sending requests in a "DDOS pattern"... Anyway, the datapoints subscriptions API has been issuing warnings that the SDK implementation is still in alpha, so we are free to change it! (and even if it wasn't in alpha, I would merge this feature without hesitation)

[matiascognite]

Sounds great 👌 Not sure what you mean with "python defaults to seconds" though? Your docstring description is very clear.

In java and other languages, timestamps are often measured in milliseconds. So a variable named timeout could be thought to be given in milliseconds, unless it is named timeoutSeconds. Witch is also the name in the API itself.

However, in python, I have the impression that time is measured in seconds, unless specified otherwise.

Anyway, if you are happy, I am happy :)

[haakonvt]

In java and other languages, timestamps are often measured in milliseconds. So a variable named timeout could be thought to be given in milliseconds, unless it is named timeoutSeconds. Witch is also the name in the API itself.
However, in python, I have the impression that time is measured in seconds, unless specified otherwise.
Anyway, if you are happy, I am happy :)

Today I learned! At least epoch time is given in float sec (time.time()). Looks very good to me!