Alan de Freitas
Alan de Freitas
We are extracting more dependency symbols than we need. - We extract dependency symbols used in `NamedTypeInfo` so their ID points to the proper symbol. This is technically correct but...
Some people expressed an interest in a JSON generator. Once the XML and Handlebars generators are in sync regarding the DOM, a native JSON generator would be trivial to provide...
ExprInfo is currently only a string with the expression as written.
Doxygen's language: - It's poorly documented enough to develop a second independent implementation in Clang, and there will always be conflicts. - Has lots of historical cruft - Is *unique*,...
The lack of tests for `Interface` is a problem. We might consider emitting it as part of the XML.
As LLVM is a large project, it would be nice to see how mrdocs handles it in terms of memory consumption (#718). We should add it to the demos. (not...
Clang processes one TU at a time while Mr. Docs accumulates data with each TU. For instance, if you compile 100 TUs and each produces 100 unique `Info`, then Mr....
Investigate what's stopping a boost::urls::url_view from being `constexpr`. We might be able to make it conditionally `constexpr` only under c++14. The obstacle might be that parsing is heavily buffer-oriented.
We should have a function to handle unicode urls: ```cpp result parse_unicode_url(...) ``` To convert to a valid URL, the host would be converted with punycode, and other components would...
Can someone derive from `std::string_view` as an alternative to deriving from `boost::urls::grammar::string_view_base`? `string_view_base` stores a `core::string_view` and is a proxy for its member functions. In principle and semantically, it's guaranteed...