This may have impact on compilation times, since the ddoc generation code will have to walk the AST twice, once to categorize the symbols, the second time to output them in sorted order.
Comment #2 by doob — 2014-06-26T21:05:12Z
How about having one buffer per category and then merging them at the end before outputting them. Will that help?
Comment #3 by hsteoh — 2014-06-26T21:12:36Z
That might work, though it does significantly change the way the ddoc code works right now. But I'm no expert in the ddoc code, so I could be wrong (I've only seen bits and pieces of it, just enough to fix some bugs, not enough to have an accurate overview of how it works).
Comment #4 by robert.schadek — 2024-12-13T18:15:49Z