On 10/9/22 03:08, Siarhei Siamashka via Digitalmars-d-learn wrote:
Does the difference really have to be two orders of magnitude for you to acknowledge that there might be a performance problem in Phobos? [...] Except that similar one-liners implemented using other programming languages are faster and more versatile (can handle any input data without catastrophic performance pitfalls).
Oh, I get all that, there's no reason to argue with me or win me over, I can see that the implementation is subpar. Since I'm never hitting this performance bottleneck in my code, and being a regular dev and not a core maintainer, I simply haven't been motivated enough to contribute an improvement. Change like that isn't happening in the forums. Various optimizations have made it into Phobos in the past, don't think you would get any pushback if you can show that a new implementation improves the situation in almost all cases while maintaining compatibility.
