Is the --ansi option still significantly affecting the performance as documented?

[ericbn]

Hi!

Version 0.26.0 change log mentions:

Significant performance improvement in ANSI code processing

but the performance topic in the documentation says:

--ansi tells fzf to extract and parse ANSI color codes in the input, and it makes the initial scanning slower.

Is the performance documentation still accurate, or can we use --ansi starting with version 0.26.0 without worrying about the performance hit anymore?

[tjex]

I don't feel like there has been a performance hit after enabling --ansi in my default ops.
However, it may need to come down to a benchmarking test, which I have no skills or know how to do...

[junegunn]

It depends on a few parameters, but it still can make fzf slower by 20~30%.

Is it noticeable? Probably not, unless you're using fzf with millions of input lines. But there's no point in doing unnecessary extra work so it's not on by default.

[ericbn]

Thanks a lot for the benchmarks and the explanation. It makes perfect sense this is not on by default.