fix(diff+file): avoid filling defaults in config for kong

[samugi]

Summary

the previous logic was filling defaults in the configuration that was passed to Kong. This was problematic, especially where nils were populated as defaults, e.g. if a shorthand_field was passed with some value and the corresponding new field was auto-populated as nil by decK , the auto-populated nil value would take precedence in Kong thus causing the shorthand_field to be ignored (see linked issues).

This change applies the default values only to configurations used for diff, the original configuration from the file is passed to Kong as is.

Configuration

_format_version: "3.0"
plugins:
- config:
    endpoint: http://example.test
  enabled: true
  name: opentelemetry
  protocols:
  - grpc
  - grpcs
  - http
  - https

Before the change

deck gateway sync kong.yml --verbose 1

PUT /plugins/83c19ebe-4cd5-4ac2-8f6b-416af32bdfaf HTTP/1.1
Host: localhost:8001
User-Agent: Go-http-client/1.1
Content-Length: 743
Content-Type: application/json
Accept-Encoding: gzip

{
  "created_at":1724513513,
  "id":"83c19ebe-4cd5-4ac2-8f6b-416af32bdfaf",
  "name":"opentelemetry",
  "config":{"endpoint":"http://example.test","traces_endpoint":null,"batch_flush_delay":null,"batch_span_count":null,"connect_timeout":1000,"header_type":"preserve","headers":null,"http_response_header_for_traceid":null,"logs_endpoint":null,"propagation":{"clear":null,"default_format":"w3c","extract":null,"inject":null},"queue":{"concurrency_limit":1,"initial_retry_delay":0.01,"max_batch_size":200,"max_bytes":null,"max_coalescing_delay":1,"max_entries":10000,"max_retry_delay":60,"max_retry_time":60},"read_timeout":5000,"resource_attributes":null,"sampling_rate":null,"send_timeout":5000},
  "enabled":true,
  "protocols":["grpc","grpcs","http","https"]}

After the change

deck gateway sync kong.yml --verbose 1

PUT /plugins/83c19ebe-4cd5-4ac2-8f6b-416af32bdfaf HTTP/1.1
Host: localhost:8001
User-Agent: Go-http-client/1.1
Content-Length: 195
Content-Type: application/json
Accept-Encoding: gzip

{
  "created_at":1724513513,
  "id":"83c19ebe-4cd5-4ac2-8f6b-416af32bdfaf",
  "name":"opentelemetry",
  "config":{"endpoint":"http://example.test"},
  "enabled":true,
  "protocols":["grpc","grpcs","http","https"]
}

Issues resolved

KAG-5157
KAG-5210

Documentation

• Link to the website documentation PR

Testing

• Unit tests
• E2E tests
• Manual testing on Universal
• Manual testing on Kubernetes

[CLAassistant]

All committers have signed the CLA.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 0% with 15 lines in your changes missing coverage. Please review.

Project coverage is 42.78%. Comparing base (64dd801) to head (decf4d3).

Files	Patch %	Lines
pkg/diff/diff.go	0.00%	15 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #133      +/-   ##
==========================================
+ Coverage   42.76%   42.78%   +0.01%     
==========================================
  Files          75       75              
  Lines        8820     8809      -11     
==========================================
- Hits         3772     3769       -3     
+ Misses       4586     4579       -7     
+ Partials      462      461       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[randmonkey]

Does this change make the filling defaults only run for diff printing differences but not to sync really updating configuration? Would this approach still require changes in Kong/go-kong?

[samugi]

Does this change make the filling defaults only run for diff printing differences but not to sync really updating configuration?

Correct @randmonkey , I added examples of how the change affects deck in the PR description.

Would this approach still require changes in Kong/go-kong?

No, this PR replaces the other one I opened, I will close the old PR if/once this is approved

[bungle]

Does decK or these tools ever use PATCH or POST?

[samugi]

Does decK or these tools ever use PATCH or POST?

not that I'm aware of, but that's a good question for @Kong/team-apiops maybe

[samugi]

failing CI because httpbin.org is down ... I'll try to rerun / merge later