brainstorm: code topic to support advanced definition search

[dna2github]

Thanks for the awesome work for code search!

what is a topic?

• for example, i have 2 micro services written in golang in 2 repos.
• i want to create a topic to describe how the 2 repos are connected. for example, MicroService 1 running a restful API server and provide "/api/v1/test"; MicroService 2 calls the API using HTTP lib; then a topic can be microservice2|customer_oriented_act() -> mircorservice2|call_request_to_api_v1_test -> microservice1|register_api_v1_test -> microservice1|handle_api_v1_test; i can name the topic like how does SystemA deal with a customer request of specific action

why need topic?

• topic is a concept for people learning and it can become a knowledge base. for example, i would like to learn how openssl integrate with a repo to think about if we can replace openssl with gnu tls lib in the repo; when we have such data, we can search in natural language like how does docker create a container -> ... syscall clone() ... (topic -> code())
• topic can support more accurate definition search; for example, almost IDE implement Go to definition in Java, when it meets an interface usage, it redirects user to the interface instead of tracking data flow and redirecting user to the real implementation. another example; in C/C++ we have macro token concatenation and we can define #define func(x) func_##x, how we can reversely find func_hello definition to the macro usage of func(hello)? in i18n area, the search is also helpful: in en.strings
string.hello=hello; sp.strings string.hello=ola; cn.strings string.hello=你好; and in code <button>{{string.hello}}</button>; how we can connect the strings and the usages? in summary, it is helpful for code() -> topic; e.g. func_hello -> func(hello) -> func_hello is a function
• even it can support topic -- code() -> doc; then we can easily know how the function works or how we design it ...

definitely, i know there is a similar issue: #8593

but it does not resolve the real advanced definition even we use LSP or LSIF. machine learning may help, but currently NLP is at baby stage.

tech challenge:

• if user mark a position, how do we track the position in source code changes