I think we should let this C era meme die, the manuals are often terrible. I'm currently working with the AWS SDK Python documentation and it's a hot pile of garbage from all points of view (UX, info architecture, technical detail, etc.).
Python lang docs are "kind-of-OK" but when someone raves about them I'm left scratching my head. Information is not always well-organized, examples are hit-and-miss, parameter and return types not always clear, etc.
Referencing docs as a programmer is generally a nightmare and a time sink, and it's the one use case where ChatGPT is slowly becoming indispensable crutch for me. I can ask for very specific examples that are not included in the docs, or that cannot be included in the docs, for example combinatorial in nature: "how can I mock this AWS SDK library by patching it with a context manager"? Occasionally it will hallucinate, but even if it gets it 8/10 times right - and it's higher than that in practice - it will prove revolutionary at least for this use case.
> I'm currently working with the AWS SDK Python documentation and it's a hot pile of garbage from all points of view (UX, info architecture, technical detail, etc.).
I agree that pretty much all AWS documentation is woeful, and it's a travesty that the service is so expensive yet its documentation is so poor. I would gladly dump AWS and never use it again, as I hate paying top-dollar to decipher the AWS doc team's mistakes (not to mention that they are unresponsive to bug reports and feedback).
My point was made more in jest, and supposed to point out the irony of the communities' changing expectations of what documentation should be like. I predict that in a few years we'll be circling back to prioritizing writing software documentation well. (Kind of like how everybody was hating on XML for the past 20 years and it's now having a renaissance because it actually does what it's supposed to well very well.)
I'm amazed by how divisive it is. I've also been using it to significantly increase my productivity, be that documenting things or having it mutate code via natural language or various other tasks. I feel that if you keep in mind that hallucination is something that can happen, then you can somewhat mitigate that by prompting it in certain ways. E.g. asking for unit tests to verify generated functions, among other things.
I find this tool so useful, that I scratch my head when I read about how dismissive some people are of it.
I think one of the reasons why Python got such a reputation for good docs is because its primary competitors back in the day were Perl and Ruby. Ruby has horrible documentation to this day, and Perl has extensive docs that are difficult to navigate; in comparison with either, Python was definitely superior.
Python lang docs are "kind-of-OK" but when someone raves about them I'm left scratching my head. Information is not always well-organized, examples are hit-and-miss, parameter and return types not always clear, etc.
Referencing docs as a programmer is generally a nightmare and a time sink, and it's the one use case where ChatGPT is slowly becoming indispensable crutch for me. I can ask for very specific examples that are not included in the docs, or that cannot be included in the docs, for example combinatorial in nature: "how can I mock this AWS SDK library by patching it with a context manager"? Occasionally it will hallucinate, but even if it gets it 8/10 times right - and it's higher than that in practice - it will prove revolutionary at least for this use case.