Geosite questions

[PoneyClairDeLune]

I've been wondering... Why doesn't Xray use a tree-based matcher for domain:, dotless: and full: matchers? The memory usage should be improved, but I'd have no idea by how much.

The tree would be built in the reverse order of domains, with TLDs forming the direct children of the root node. Whenever domain:, dotless: or full: was matched, the corresponding leaf node would be marked with such flags, as well as which lists do those flags apply. I currently use such a structure to implement a domain de-duplicator for merging several blocklists. Other types of matchers though will have to use the current design.

The assumption was made from the file structure of geosite.dat.

[Meo597]

geosite 和 geoip 是祖传屎山代码，即便以前有些优化也只是在局部解决问题

现在重构后，移动平台的 domain: 使用 trie
而没有内存压力的地方，完美哈希始终是最优解

目前重构后的版本仍有一些可以微优化的地方
比如 compact 自己有一套 idx 因此底层的 strmatcher 没有必要持有 values 估计还能压几 mb 下去