I bet more people are looking for blogs than docs.

My suspicion is that blogs have ads and documentation doesn't, and that's why the results are like this.

Totally agree! My point is that they both probably contain the same answer and I bet that's all most people care about. I think most of us here have strong preferences for picking the "most appropriate" source, but frankly, I suspect we're weird.

In my experience, blogs are mostly young students trying to build an online presence or something similar, and contain bad advice and wrongness.

If I wanted the docs, I would get the docs. The blogs are useful because the documentation is largely unhelpful due to being overly simplicity or extremely nondescript. The blogs are where it's at because you can see how someone else used that thing

That's too generalized and IME not even true for say, Python docs, or regexes, or Gnu tools, or HTML. Most bootcamp/academy/coder-type blogs/copycats are aimed at giving only a tiny fraction of the information, but with examples (sometimes only very simple ones that cover only the most basic use-cases, and no links or mention of the other missing 90% of information (e.g.: "to read more about regexes, see ...").

"site:docs.python.org/3/ term" gets much better results. (You have to add the site: qualifier, due to SEO it's no longer just enough to say "Python docs". Hence "If I wanted the docs, I would get the docs." is not as trivial to do as you suggest. Things actually went beyond annoying to downright dangerous when searches for "Python 3.10 <topic>" were drowned out by older e.g. (3.6-3.8) version SEO stuff, instead of the latest Python docs.)

Queries like "awk by example" or "egrep by example" or "jq by example" seem to hit the sweetspot between completeness and working example: they give you third-party non-commercial unaffiliated expert bloggers not stuffed with upsell links to a bootcamp. Just pure information sites.

Having mentioned all this, now I've probably inadvertently cursed those sites...

To me what this shows is that it's really painful to try and use an "intelligent" algorithm for "content mediation" as if it were an actual search engine. Anything that spectacularly fails the "exact substring match" use-case is well on the way to becoming a content-mediator, and potentially as bad a content-moderator/thought-police/etc.

Unfortunately even before anyone started talking about "semantic web" or whatever, substring-match was declining in utility due to spam, so a substring-matcher was only as useful as it's ranking algorithm. So some cleverness in ranking is necessary, but when ranking becomes too clever, or irrelevant due to the arms race with spam, it's an almost inevitable slide from necessary rankers to mediators.

Since all the search/spam arms race is pretty much played out with giants like twitter/google anyway at this point, it is kind of hard to understand why we can't get and keep a separate and truly basic internet search engine that just works. I don't have a handy example, but in my experience even DDG & friends seem to routinely fail the "exact substring match" test for unusual phrases that I know I read verbatim last week.

What I (and I think others) really want is just substring-match tool plus a very minimal amount of extra cleverness for basic stuff like spelling variations, some fuzzing for dates/digits, directly adjacent stuff according to a simple thesaurus. Not my area of expertise, but I would guess that there's a combination of problems that prevent this tool from existing.

a) Other search companies are all trying to do fancy-but-fuzzy "semantic distance" searches to emulate Google b) It's still true in 2023 that just keeping an up to date index is a super hard problem no matter how many crawlers are in your army c) Despite bots everywhere, the deeper web in social-media is just too login-walled for the rule-abiding crawlers to deal with d) There's just no money in do-what-i-say-not-what-i-mean search tools

DDG seems unusually susceptible to highly ranking AI-generated overly verbose garbage. I can't recreate the same issue with google or bing, so I am not sure where it pulls it from. I can find those same results in google, it's just they are on the 2nd or 3rd page.

I've never really found docs particularly useful. Usually, I just want a few simple examples to get the gist of how something works and docs tend to be very bad job of that. Meanwhile, the W3Schools SEO spam HN disdains tends to good job of that.

These days I use ChatGPT though.

Eh, I've had the opposite experience. If you search for anything slightly obscure in a language like Java, you'll get dozens of spam sites aimed at beginners that answer a much more basic question than the one you asked. That's because Java searches cater to a large population of people who can barely code.

Java is particularly rough because there are several populations of Java users (students, Android developers, enterprise developers, among others) and their coding standards are universes apart.

That and it's been around so long that the language and it's standards have been through so many iterations

> These days I use ChatGPT though.

I usually get non-working code snippets that pass the wrong parameters.

uBlocklist removes a lot of the crud, thankfully