test: remove cgi and ostruct dependency for ruby 4

[xuan-cao-swi]

Added cgi and ostruct gem since they are removed from ruby 4.
Also fixed some warning that is about constants reinitialization.

[arielvalentin]

@xuan-cao-swi I know this is not ready for review but I think that cgi is a runtime dependency of the API and SDK:

https://github.com/search?q=repo%3Aopen-telemetry%2Fopentelemetry-ruby%20cgi&type=code

If we do not want to add a transitive dependency on the cgi gem then we should find a stdlib alternative to CGI.unescape or CGI.escape.

[arielvalentin]

ostruct on the other hand is a test only dependency:

opentelemetry-ruby/metrics_sdk/test/opentelemetry/sdk/metrics/fork_hooks_test.rb

Line 79 in b8f887a

reader2 = OpenStruct.new

If we can find an alternative for OpenStruct I think that we should go for it.

[xuan-cao-swi]

For replacement of cgi escape and unescape maybe using URI.decode_www_form_component/encode_www_form_component?

I just did test and found out if rake is 13.3, then no OpenStruct is required. I think old rake (13.1.0 and before) use OpenStruct. Currently most gemspec using rake ~> 12.0 (except log_sdk), maybe they can change to newer rake.

[xuan-cao-swi]

Updated all rake to 13.3 for ruby 4.
Using URI to replace CGI for url encoding/decoding.
Avoid use openstruct for test case.

[arielvalentin]

I do not know if they are functionally equivalent, e.g. of en

irb(main):006> CGI.escape("*")
=> "%2A"
irb(main):007> URI.encode_www_form_component("*")
=> "*"
irb(main):008> URI.decode_www_form_component("%2A")
=> "*"
irb(main):010> CGI.unescape("*")
=> "*"

[xuan-cao-swi]

It seems URI encode/decode is for newer URL encoding rules (RFC 3986), and CGI is for legacy rules.

Decode seems aligned for asterisk.

irb(main):011> URI.decode_www_form_component("%2A")
=> "*"
irb(main):014> CGI.unescape("%2A")
=> "*"

If we are ok with adding cgi in dependency then I can revert back the changes.
I don't see any other stdlib have uri encode and decode support.
We can also copy over the function from cgi library and integrate them in sdk/exporter codebase.

[arielvalentin]

I think what we should do is consult the specification.

Does it specify which to use?

In inclined to keep this change but report it as potentially a breaking change.

[xuan-cao-swi]

Based on the spec, baggage propagator and exporter protocol both use w3c baggage specification. Jaeger propagator will be deprecated. RFC 3986 is compatible with W3C Baggage specification. I think the most important consideration is to following Character Encoding guide

Difference about space

# Ruby encode_uri_component/encode_www_form_component
URI.encode_uri_component(" ")       # Space Encoding %20
URI.encode_www_form_component(" ")  # Space Encoding +
CGI.escape(" ")                     # Space Encoding +

# Python is using unquote/quote and unquote_plus/quote_plus (RFC 3986)
urllib.parse.quote(" ")        # Space Encoding	%20
urllib.parse.quote_plus(" ")   # Space Encoding	+

# JS uses default encodeURIComponent
encodeURIComponent(" ")        # Space Encoding	%20

Difference about * (W3C Baggage allows both * and %2A for encoding *)

python quote/quote_plus encode to %2A
js encodeURIComponent doesn't encode to %2A
ruby encode_uri_component/encode_www_form_component doesn't encode %2A

[arielvalentin]

Per our slack convo I think we agree that uri is the most correct per the spec. Let's use that and set this as a fix:

[arielvalentin]

Actually how do you feel about splitting this or into separate ones.

Have the URI fix in a separate PR from ostruct so it's more explicit