fix: allow response headers and fix headers not working in const requests

[sansyrox]

Description
fix: allow response headers and fix headers not working in const requests
This PR fixes #325 , #323

[netlify]

✅ Deploy Preview for robyn canceled.

Name	Link
🔨 Latest commit	8e5cee7
🔍 Latest deploy log	https://app.netlify.com/sites/robyn/deploys/63987b83dc146c0008ffd1db

[sansyrox]

@AntoineRR @cirospaciari , do you folks want to give it a final green light before we release a new version?

I really appreciate all your help 😄

[cirospaciari]

@AntoineRR @cirospaciari , do you folks want to give it a final green light before we release a new version?

I really appreciate all your help smile

I think will work.
But my issue is with jsonify basically this is serializing and deserializing the headers without any need, in CPython this is really slow, even if Robyn can get all raw power of actix web (3.5 mi req/s in TechEmPower) will be limited to JSON serializer performance, the better JSON serializer performs about 1.4 mi req/s in TechEmPower, also is adding a lot of CPU usage, copying, and dynamic allocations. Rewriting now will be better than adding this to the main branch and causing incompatibility issues in the future.

JSON data:
https://www.techempower.com/benchmarks/#section=test&runid=124fd2af-3030-4315-8876-e1db1fa91193&test=json

Plaintext data:
https://www.techempower.com/benchmarks/#section=test&runid=124fd2af-3030-4315-8876-e1db1fa91193&test=plaintext

Take a look in:
https://docs.rs/pyo3/0.2.7/pyo3/struct.PyDict.html
And in:
https://pyo3.rs/main/doc/pyo3/types/struct.pytuple

[AntoineRR]

I agree with @cirospaciari, we shouldn't use jsonify for the headers.
Although I think it is already handled like this in Robyn currently (see this test route from the main branch for example) so there will already be an incompatibility issue. Can you confirm this @sansyrox ?
Also, refactoring to allow something else than a string may be a big refactor for issue #336, maybe it would be good to merge this and start a new PR to remove jsonify ?

[cirospaciari]

I agree with @cirospaciari, we shouldn't use jsonify for the headers.
Although I think it is already handled like this in Robyn currently (see this test route from the main branch for example) so there will already be an incompatibility issue. Can you confirm this @sansyrox ?
Also, refactoring to allow something else than a string may be a big refactor for issue #336, maybe it would be good to merge this and start a new PR to remove jsonify ?

If already will be an incompatible issue so, merging and fix in a new PR is acceptable, but refactoring should be priority, I need to benchmark it first but the performance penalty will be huge without refactoring.

[sansyrox]

but refactoring should be priority, I need to benchmark it first but the performance penalty will be huge without refactoring.

Agreed.

Although I think it is already handled like this in Robyn currently (see this test route from the main branch for example) so there will already be an incompatibility issue.

Yep, yep.

Also, refactoring to allow something else than a string may be a big refactor for issue #336, maybe it would be good to merge this and start a new PR to remove jsonify ?

I have been working this for a few hours now. It is indeed a big refactor. :(

I think it makes sense to merge this one and refactor based on a priority.