I recently¹ discovered ugrep, and I have benefited a lot in using this to search my logseq journals. But recently I noticed I was getting fewer results and also confirmed some misses. My specific strange issue was that if I pointed directly to my miss, ug found it but not otherwise. I was consulting with microsoft copilot about this mystery.

running into a weird ugrep issue where, ug -%% 'some blah' /some/foo/path , comes up dry , but if I am specific ug -%% 'some blah' /some/foo/path/more/specific/file.py , then I get results. Is there some index that needs to be poked for rebuilding?

Insights included turning on --hidden and --all , to search in .gitignore and other hidden files, as well as following symlinks. Those did not work for me. I went through a few more back and forths in the conversation but nope nothing. One cool suggestion was to use --stats. This was helpful because ug will report how many directories and files were searched and when I tried this I saw way fewer directories than expected.

And then the answer became obvious 🤦‍♂️, ughh, ug was likely not looking recursively. My original success was on my logseq journal which is just a flat directory and so I did not encounter this issue. So I looked at the ug --help again and indeed saw there was a -r recursive option. I tried it and problem solved.

It took me a while to grok this one, but in my defense I had a solid 30 days of benefit where I just so happened to only use this on a flat directory 😆. And for the past several years prior, I had been living on ag, the silver searcher, which is recursive by default and so I kind of assumed this would be true for ug too.