feat(contentful): memoizer #BOT-467

[dpinol]

Depends on #1640. ⚠️ Please review it first

Description

Use memoization to remember forever the last successful invocation for each combination of arguments, and use its result whenever Contentful fails.

Context

Contentful failing during 1h on 2021/6/8 and 2021/6/9.

Approach taken / Explain the design

• Create a generic Memoizer class with several strategies.
• Create a decorator (FallbackCachedClientApi) around the Contentful API class which uses the Memoizer fallbackStrategy.
• Compose the new decorator within the existing decorator (CachedClientApi) which avoids recent calls.

FallbackCachedClientApi does not use the memoization library used by CachedClientApi (memoizee) because this library does not allow directly accessing the cache.

To document / Usage example

The contentful plugin by default uses memoization to remember forever the last successful invocation for each combination of arguments, and use its result whenever Contentful fails. To disable this feature, use to true the new plugin option beloew

  /**
   * By default, the result of the last delivery invocation will be cached
   * forever and will only be used when a delivery call with the same arguments
   * fail.
   */
  disableFallbackCache?: boolean

Testing

The pull request...

• has unit tests

[linear]

BOT-467 mitigate Contentful outages with inmemory cache

In 2021/6/8 & 2021/6/9 contentful was down during 1h, which caused all our bots using this CMS to stop working.

We'll start with a simple memoization (lazy in memory cache)

In the future maybe we can:

• Use redis
• Preload all contents when starting the bot instance
• Load the contents into the bot before deploying

[vanbasten17]

do you no need to put again extends any[] in this type? I understand that Args will extend it as you have typed it as MemoizerStrategy, could be?

[dpinol]

I typed fallbackStrategy as MemoizerStrategy for the compiler to verify that the lambda I retry complies with this type, but the lambda is not defined to be of this type (generics arguments must always be explicit, otherwise changing their names in the definition would break all usages)

[vanbasten17]

delete?

[dpinol]

thanks, done in ee9efdb

[vanbasten17]

delete?

[dpinol]

thanks, done in ee9efdb

[asastre]

👏 👏 Just one doubt in the tests

[github-actions]

Unit Test Results

  1 files    7 suites   3m 0s ⏱️
32 tests 32 ✔️ 0 💤 0 ❌

Results for commit 63dca40.

♻️ This comment has been updated with latest results.