[IMPROVEMENT] Add exponential backoff retries to HTTP client

oliver-gordon commented 7 months ago

Description

When a HTTP request fails, we will now retry it 5 times (arbitrary value, non-configurable currently) where the time between each request grows until we stop.

I've also tried to include for missing typehints and docstrings where I was updating code.

Remaining Tasks

Related PRs in other projects (PASQAL developers only)

Additional merge criteria

Breaking changes

Checklist

[ ] The title of the PR follows the right format: [{Label}] {Short Message}. Label examples: IMPROVEMENT, FIX, REFACTORING... Short message is about what your PR changes.

Versioning (PASQAL developers only)

[ ] Update the version of pasqal-cloud in _version.py following the changes in your PR and by using semantic versioning.

Documentation

[ ] Update CHANGELOG.md with a description explaining briefly the changes to the users.

Tests

[ ] Unit tests have been added or adjusted.
[ ] Tests were run locally.

Internal tests pipeline (PASQAL developers only)

[ ] Update and run the internal tests while targeting the branch of this PR. If your PR hasn't changed any functionality, it still needs to be validated against internal tests.

After updating the version (PASQAL developers only)

[ ] Open a PR on the internal tests that updates the version used for the pasqal-cloud backward compatibility tests.

oliver-gordon commented 7 months ago

I couldn't find a test for the retry mechanism, could you add one?

Any test mocking an exception is testing the retry mechanism. But sure, I'll add another one

MatthieuMoreau0 commented 7 months ago

I couldn't find a test for the retry mechanism, could you add one?

Any test mocking an exception is testing the retry mechanism. But sure, I'll add another one

Either adding a new test or amending the existing ones to check that the request was retried as expected. I could get rid of the retry logic entirely and tests would still pass which is no good.

pasqal-io / pasqal-cloud