llxisdsh

用redis也会出现主机宕机, 然后回拨

分享另一个方法: 如果ExecCmds生成蓝图不全, 用下面: UnrealEditor-Cmd.exe xx.uproject -TestExit="Fixed up redirectors for" -TestExitExecCmds="Puerts.Gen" -nullrhi -log -nosplash -nosound -nopauseonsuccess -nocontentbrowser 思路就是-TestExit检测到某个log(编辑器收集完bp), 然后退出, -TestExitExecCmds指令同时捕获到退出信号, 去执行某个命令"Puerts.Gen" -TestExitExecCmds 指令实现需要加点代码到: Puerts\Source\DeclarationGenerator\Private\DeclarationGenerator.cpp: ``` cpp public: void StartupModule() override {...

Has尚能支持, 但HasBy(func)如果不是通过key, 那就会遍历values, 这就O(n)了 说实话这些都是外围函数, 不过你可以试试 https://github.com/llxisdsh/pb.git

我的readme里有性能测试，我已经找不到比我pb.MapOf更快的了，包括这个，无论读和写，我的测试环境是x86 64位机型。我的项目也在使用。

> > Has尚能支持, 但HasBy(func)如果不是通过key, 那就会遍历values, 这就O(n)了 说实话这些都是外围函数, 不过你可以试试 https://github.com/llxisdsh/pb.git > > 好的，那这个库和pb库哪个性能更高呢？除了HasBy(func)，其他函数计划什么时候支持呢 加了呢，so easy

> > A dedicated `ComputeRange` makes sense to me since it'd have better atomicity guarantees when compared with `Range` + nested `Compute`. The thing is that the each bucket `Range`...

> > Strictly speaking, when the table changes, we have to restart the loop. > > In case of a intensive writes to the hash table, this would lead to...

I’ve implemented a progressive Copy-On-Write (COW) for mapof, where Store performance is about 20% slower than the standard implementation. However, when combined with parallelized resize, Store becomes 5% faster than...

Additionally, my COW mapof implementation is 20% slower than the standard one, partly because my buckets store 6 entries, making them slightly deeper.