Hacker News Reader

Comment by 0x00000000 | Hacker News Reader

Don't remind me of how awful GitHub search is.

I guess code search is just too niche with no marketability to ever improve

GitHub is most aweful search that I have regularly used. It doesn’t even have basic dedup capability, forget about any thing slightly less basic like limiting search to file names. And it’s not that it is hard to derank potential dedups given that they already know the fork graphs. I have spent good chunk of my productivity manually sweeping through pages upon pages of exact same search results to find the snippets I was looking for and hating creator(s) of this functionality from the bottom of my heart. These people shouldn’t be allowed to build any search again without a mandatory year long rehabilitation training camp (unless they are doing this for enemy states during war times).

How about finding multiple hits per file? Tell me how many matches there are in each file, PLEASE. I should not have to click into the match to know there's more than one match in that file. I shouls also be able to have a view that shows me every match (a la grep/ag). If I need to search, I pull down the repo, index it quickly with ag and do the search. That's really unacceptable, in my eyes, when the basic search is there already.

I find myself using source graph; It seems to work well enough.

yes, source graph looks great. Using docker, every easy to deploy on any OS

Github search also only searches content hosted on github.com which while still large is only a fraction of the code available on the internet.

GitHub search? Try nytimes.com search.

lol i worked in the same department as the search team. what a nightmare - some of the code in there went back to the 90s

Is it worse than reddit?

Try the search feature in many CMS systems. Those tend to really suck, with WordPress' being awkward enough that a whole ecosystem of plugins cropped up to replace it.

There's also at least one version of the OpenCart admin search which was designed so poorly that it didn't actually search based on whether the term was anywhere in the page title/content, but whether the title started with the search term (I think they'd screwed up the MySQL syntax). That was interesting.

Either way, search on a system not designed as a search engine is usually okay at best, absolutely terrible at worst.

Reply on news.ycombinator.com