Yes, I care about faster conversions between string types.
On Monday, May 11, 2015 at 10:52:28 PM UTC-4, Scott Jones wrote: > > I would like to get the performance of string handling in Julia improved > (and also correct a number of flaws in handling Unicode). > Currently, many of the operations are very slow (because of the way > strings are represented, and also simply > because the algorithms used were not as fast as they could be). > For communicating with Java where the String type is always UTF-16, C/C++, > where wide strings are either UTF-16 or UTF-32, > databases (which frequently use UTF-16 in Asia, because UTF-8 can take > more space, depending on the data), or using Unicode > APIs in Windows or the ICU libraries which use UTF-16, the performance of > the conversions between the ASCIIString/UTF8String > types that Julia normally uses for strings and UTF-16 can be critical for > an applications performance. > > I have a PR (#11004) to fix the performance issue #10959 I raised (in the > GitHub repository ScottPJones/julia branch spj/fast_utf), > if anybody would like to try this out. > My own testing shows that it gives about a 2-10x improvement (most of the > time, 10x), you can get my latest benchmark results > along with the code I used to benchmark it at my gist, > https://gist.github.com/ScottPJones/bb712f7b85d1d8d91a9a. > > I am curious if these sorts of performance improvements (along with better > error handling and Unicode input validation) > would be important to anybody else, as I am trying to convince people to > merge this PR into the Julia base... > > Thanks, Scott > > >
