perf: switch conditionally to ripgrep or fd

[ja0nz]

resolves l3kn/org-fc#69

[l3kn]

Thanks for this!

At first I couldn't see the benefit because on my org-fc-directories (still containing thousands of files) the difference in runtime was so small. Trying it on my whole home directory, fd is much faster.

I think the greatest improvement is limiting the search to files actually containing flashcards. With the -exec flag of find, we might be able to avoid the extra xargs and I assume fd has something similar.

The commands will need some work though, so their results are exactly the same. (For example, the current rg command also finds org archive files).

I'll try to rewrite them and then do some end-to-end indexer benchmarks.

[l3kn]

After a lot of experimentation, these three commands seem to behave exactly the same regarding hidden files, hidden directories and upper/lower case '.org' extensions.

• "find -L %s -name \".*\" -prune -o -name \"[^.]*.org\" -type f -exec grep -l --null \"^:FC_CREATED\" {} \\+"
• "rg ^:FC_CREATED: -L -l --null -g '[^.]*.org' %s"
• "fd --type f -s -e org -g '[^.]*.org' -L %s --exec-batch grep -l --null \"^:FC_CREATED:\" {}"

As expected, rg is significantly faster than find. I have benchmarked these on my home directory and the subdirectory I use for org files:

• find: 2.53s, 0.18s
• rg: 0.36s, 0.095s
• fd: 0.84s, 0.24s

What's odd here is that fd is slower than find in one of the cases. -e org is made redundant by the latter glob pattern but it appears to reduce the runtime.

I think it's safe to assume that anyone who has installed fd would be fine installing rg as well, so until we can find a fd command that is faster than find in all cases, I'll only add rg as an alternative.