Performance enhancement by caching fastavro schema parsed on every message serialized & deserialized

[bruc3mackenzi3]

The fastavro schema is being re-parsed on each message produced, which is a detriment to performance. This PR addresses it by caching the parsed fastavro schema instead of the JSON dict schema.

We see a 25% performance enhancement when used with large schemas, and a 5% improvement with small schemas.

[ghost]

It looks like @bruc3mackenzi3 hasn't signed our Contributor License Agreement, yet.

The purpose of a CLA is to ensure that the guardian of a project's outputs has the necessary ownership or grants of rights over all contributions to allow them to distribute under the chosen licence.
Wikipedia

You can read and sign our full Contributor License Agreement here.

Once you've signed reply with [clabot:check] to prove it.

Appreciation of efforts,

clabot

[bruc3mackenzi3]

[clabot:check]

[ghost]

@confluentinc It looks like @bruc3mackenzi3 just signed our Contributor License Agreement. 👍

Always at your service,

clabot

[AndreiLieb]

@rnpridgeon any updates on this? This is a pretty significant optimization for us.

[rnpridgeon]

Good catch @bruc3mackenzi3 , all in all the PR looks good. I'll sanity check it against the Avro integration tests this afternoon for both the py2 and py3 interpreters and if everything works as expected I'll merge. Thanks again for your contribution and I apologize for the hold up on this review.

[bruc3mackenzi3]

@rnpridgeon thanks for taking a look and that's great news. We look forward to having our PR merged. If you need anything from us in the meantime let us know.

[rnpridgeon]

Thanks for your contribution @bruc3mackenzi3