Comment by hmwhy | Hacker News Reader

Comment by hmwhy | Hacker News Reader

Parent Full thread

I second this.

I have some experience in both web development and scientific computing in various fields (not bioinformatics). Things felt just as hard at the beginning in all cases—I don't recall things being any better in scientific computing compared to web development.

I agree that documentation is terrible most of the time, but that's not something limited to web development in my own experience. (On the contrary, check out the documentation of the website in question in this thread: https://news.ycombinator.com/item?id=23031408, it's absolutely fantastic).

Some things become second nature when you have developed expertise in that field, and it's easy to criticise something that you have comparatively less knowledge in and pass it off as "too difficult", particularly when the web development part seems to be a means to an end (which seems to be the case here).

Edit: missing conjunction, tidied up the last sentence for clarity.

Oh, interesting. I just wrote a reply saying the exact opposite. I remember learning scientific computing as being much less frustrating. Sometimes you'd have some error spawning out of some ancient C/Fortran library. But that was rare for me. Even after a couple of years on the web, I still feel like I encounter issues where I just don't even know how to start.

Same, and they're not.

If anything, programming for academia is more of a mess than web development. Case in point - I spent the better part of a day rewriting what's essentially a parallelized for loop because the library I'm using in Julia to open HDF5 files has a well known memory leak. I have the impression that something similar in any popular web dev framework or library would have been ruthlessly purged a long time ago.

Bugs in libraries and code that overpromises are maddening (and too common), but I think it's different.

You can get the data out of an HDF5 file with

    using HDF5
    c = h5open(filename, "r") do file
        read(file, "data")
    end

And you're done. You're not somehow locked into plotting data with the Plotly backend, but not GR. You know exactly where the data came and can trivially point it somewhere else with some 'magic' breaking.

HDF5.jl suffers from a memory leak when HDF5 files are opened in succession.

Bug report: https://github.com/JuliaIO/HDF5.jl/issues/349

Right, but that's a bug.

It's annoying it isn't fixed and working around it is undoubtedly annoying. This seems different from a system where there's lots of complexity in just choosing what frameworks to start with and if/how they play well together, and all of that rapidly changing over time. (I'd bet, for example, that many of those examples used to work).

Reply on news.ycombinator.com