Use weaver for semantic convention codegen

[jsuereth]

I wasn't sure what you might want to change/fix about codegen when moving to weaver, so this attempts to keep the status quo as best as possible.

TODOs

• Import markdown / notes -> javadoc filter to the Jinja weaver uses.
• Finish supporting all types of template keys or throw an exception if an unsupported one shows up.
• Remove previous jinja template
• filter unwanted semconv groups in JQ expression in weaver config.
• figure out what to do w/ K8S convention camel case.
• Use type_mapping in weaver.yaml instead of macros.
• Add stability: deprecated attributes back to incubating classes.

Possible Improvements thanks to weaver

• We could generate all the code (for incubating + stable) in the same weaver config, with the same shared set of templates. I kept it as-is for now so you can more easily compare what's happening.
• We can register name conversions for method/class/etc. for easier to understand case conversions.
• We can register acronymns, for consistent capitalization.

[jack-berg]

What's the difference between build tools and weaver? Is the community trying to transition away from build tools?

[jsuereth]

Update this, but we're now getting issues where weaver can't appropriately "javadoc-friendly" a comment, e.g.:

[ant:checkstyle] [ERROR] /home/joshuasuereth/src/open-telemetry/semantic-conventions-java/semconv/src/main/java/io/opentelemetry/semconv/ExceptionAttributes.java:26: <p> tag should be preceded with an empty line. [JavadocParagraph]

[jsuereth]

@jack-berg This is ready for review now. We have the javadoc/comment fixes in place for codegen.

I haven't consolidated down to calling weaver ONCE in gradle yet, but I can do that in a follow up PR.

[jsuereth]

Here are two minor suggestions:

1. Attributes are sorted by default, so most of the sort filters can be removed.

Fixed

2. The file name definition can be simplified, as detailed below.

I recommend using the new, simpler method to specify file names. For example, in IncubatingSemanticAttributes.java.j2, instead of:

{%- set my_file_name = ctx.root_namespace | pascal_case ~ "IncubatingAttributes.java" -%}
{{- template.set_file_name(my_file_name) -}}

you could define the file name directly in weaver.yaml within each template object like this:

templates:
  - template: "incubating_java/IncubatingSemanticAttributes.java.j2"
    filter: ...
    application_mode: ...
    file_name: ctx.root_namespace | pascal_case ~ "IncubatingAttributes.java"

FYI - the file_name is actually a JINJA template, not an expression, so it's:

templates:
  - template: "incubating_java/IncubatingSemanticAttributes.java.j2"
    filter: ...
    application_mode: ...
    file_name: "{{ctx.root_namespace | pascal_case}}IncubatingAttributes.java"

[jsuereth]

FYI - I can remove this in the template when there aren't enum definiitons if you'd like to see that.

[jack-berg]

I don't think its a big deal. Your choice.

[jsuereth]

Let's leave as-is for now.

Once we're on weaver (the important piece of this PR) there's incremental improvements we can begin making - but I'd like to focus on progress.

[jsuereth]

FYI - there's a conversation on weaver about whether we need a feature to force ending . in comments, see: https://cloud-native.slack.com/archives/C0697EXNTL3/p1725530857893809

I have concerns fixing this may just cause more problems in comment generation, and these one-word descriptions should be fixed upstream, but as an FYI - I wasn't able to safely fix the . vs. "no dot" here.

[lquerel]

The GH issue for tracking open-telemetry/weaver#353

[jsuereth]

thanks @lquerel !

[jack-berg]

@jsuereth, @lmolkova, @lquerel, and @open-telemetry/java-approvers - I'll merge this tomorrow if there are no additional comments.