Hacker News Reader

> scientific software ... where interfaces are notoriously bad.

I think the problem here is two fold: bad UIs and bad APIs. In the science world, there doesn't seem to be any abstraction between backend and frotnend -- the whole thing is one intertwined mess.

Scientists don't always write great software; they write scripts that get the job done quickly in a very specific way. Pair that with the lack of time that they have to wrap things in a GUI and you get a Tkinter/Qt app that is unmaintanable.

I'm currently volunteering with the NZ Department of Conservation. Here is a tiny preview [0] of an app that I have just started making for them. It's purpose is to aid scientists with processing camera trap images using machine learning. I'm building the UI using ElectronJS and Svelte, whilst the backend is a separate Python project.

[0] https://streamable.com/0ji1ii

You realise it's possible to write a decent UI in Qt that isn't a dumpster fire from the codebase perspective right?

I know the fad these days is to turn everything into a web app but my experience working with Qt really just demonstrated to me that a lot of web UI development is reinventing wheels from desktop UI but poorly and with more bugs and frustration.

That's just my perspective though :-)

My personal experiences differs.

I recently spent a year working on a Qt app for drawing circuit diagrams, aimed at nuclear engineers. I've lost count of the number of times I found myself wishing for some feature that React has implemented for years.

The parts of the code that were easiest to maintain were the ones that reinvented reactive UI: every modification to the underlying data called some "updateGraphicsView" function that would then re-render the entire thing. Not great for performance, great for avoiding bugs.

Seriously, for me Qt goes in the "never again" category.

> I'm building the UI using ElectronJS and Svelte, whilst the backend is a separate Python project.

Thanks for sharing this. I have a relatively simple Python script that I'm trying to package for a non-technical Windows user so, as someone whose experience skews toward web development, your comment caught my attention.

I just googled "electron svelte python" and found this write up:

https://medium.com/@abulka/electron-python-4e8c807bfa5e

Which references this Github repo:

https://github.com/fyears/electron-python-example

It all sounds pretty brittle. How are you managing communication between your ElectronJS frontend and Python backend? Any references you recommend?

I think I came across that as well. I didn't like it.

As far as I understand, there are two ways to handle communication between the Python and ElectronJS:

1) Use inter-process communication. Basically, your Python app outputs to stdin and your Electron app consumes the output. Also works the other way. This is what I went with.

2) Use a Flask (or equivalent) local web server and have your app send requests to it.

I find method 2) a bit hacky and unintuitive for a desktop app. If you are planning on hosting your backend in the cloud this would be a decent solution, but I am not.

I had already done some work to turn my Python app into a standalone CLI tool using `click` [0] which made the interaction with ElectronJS very simple. I used `python-shell` [1] for the inter-process communication. I do think that I might not need this once I package my Python app into an .exe -- then Electron should be able to interact with it in some way without caring that it's built in Python.

You can take a look at my repo here, though it's still in its infancy: https://github.com/petargyurov/megadetector-gui

[0] https://click.palletsprojects.com/en/7.x/

[1] https://www.npmjs.com/package/python-shell

Why not use something like gRPC or cap’n’proto to make the request/response stuff easier to manage?

Because it still requires request and response. The way you are doing request/response, or encoding the data is immaterial.

Some time ago I helped build a software that heavily used some python GIS libraries (GDAL, rasterio, etc) and we needed a UI for some windows users. We chose cefpython[1] to the job, as we could interface easily between the python and JS parts. The development experience is not so polished as of Electron though.

Nonetheless, is a great project and not so difficult to use, although not perfect, it worked perfectly for our needs.

[1] https://github.com/cztomczak/cefpython

As far as quickly standing up an effective gui for python, I'd recommend gooey.

https://github.com/chriskiehl/Gooey

NZer here, interested to hear more about how you ended up doing that.

Pasting my answer from a similiar comment above:

> I sent an email to a specific conservation project, asking if they happen to need any software help at the moment and they did. We got chatting, agreed on some goals and I got working. I signed their volunteer agreement form which sets out some conditions (nothing unsurprising). That's it really.

For what it's worth I am not even based in NZ; I'm in the UK. Happy to chat in more detail if you want to know more (my email is in my profile)

This is very pretty. Great job on the frontend.

That actually looks really cool! Are you planning to display the output in the UI as well?

Thanks! Yes, I am. The intention is for the user to also be able to correct the output of the model.

After that I want to give the user an easy interface for training their own models (transfer learning using an existing base model).

How is Electron doing btw? I'm itching to use it in one project but it doesn't feel like it's getting any traction in the community for a long time now. Since it's a long term project I kind of get mixed feelings about using it.

This is my first project with it so I'm probably not in a position to provide much insight. So far development with it has been almost identical to coding a website, but maybe because I am yet to start on any OS dependant features.

What makes you say that it isn't getting traction? AFAIK, some huge companies are using it to create their apps.

I think it gets a lot of underserved hate, just because it's JS and because it isn't "truly native". People also love to chime in about its performance but 9/10 times that's down to a bad implementation, in my humble opinion.

On top of that, I already have a browser installed and each Electron app is a step for turning the Web into ChromeOS.

Is there any benefit to using electron vs a website with the same code? Especially if you want it to be free and accessible to everyone

I have built a couple Electron apps, one of which will be available as a commercial product soon.

The main advantages (over executing your app in-browser) to my knowledge:

1. You can execute native code outside normal browser sandbox

2. You get your own application window

You can even have multiple application windows! But VS Code is the only one I can think of that bothers.

God forbid you want to be in two Slack orgs at the same time, perhaps on different virtual desktops. Off to the browser version with you!

That's just bad usability on Slack's part, not a technical limitation of the tech. Also, afaik Slack is/has moving away from Electron.

Agreed on the bad usability, but that's par for the course with most Electron apps. You get the laziest possible "stuffed a webpage in a window" UI. And maybe it's even a nice UI in some ways, but it's usually pretty clear that it's not native software.

Re: moving away from Electron, not as far as I know. They had a major update a year or so ago to make some big efficiency gains by not running a separate Electron instance for each organization, but the new version is still in Electron: https://slack.engineering/rebuilding-slack-on-the-desktop/

One can hope that they'll take the iPad version and make some improvements to let ARM Mac users get that instead, but currently they've chosen to disable downloading it from the Mac App Store.

Wait, you can execute say c++ using electron?

Yep: https://www.electronjs.org/docs/tutorial/using-native-node-m...

Spotify is an Electron app that's doing well. Also my own company's experiences with it have been good.

Spotify isn't an Electron app. It predates Electron and uses CEF ( https://bitbucket.org/chromiumembedded/cef , https://en.wikipedia.org/wiki/Chromium_Embedded_Framework ), like Steam.

I should have said Slack.

As a curious New Zealander, how does one volunteer with a government department? I had always assumed it would have to be strictly a contractor type relationship

I sent an email to a specific conservation project, asking if they happen to need any software help at the moment and they did. We got chatting, agreed on some goals and I got working. I signed their volunteer agreement form which sets out some conditions (nothing unsurprising). That's it really.

Ah, interesting! I had just assumed that most departments wouldn't have a process for handling this kind of stuff or that it would be considered more costly in terms of getting a signed agreement, coordinating etc etc

Games just as often as not have terrible UIs. Take Destiny 2 for example. It's a nightmare of right click (no indication that you'd want to right click). And in particular games are flash for flash sake. I love it for games but I wouldn't want my app I'm trying to be productive with to animate every tiny motion or throw out sparks when I click a button.

Further, much of the database is joypad controlled UIs

Right. But there are many games, many experiments to draw from.

|Further, much of the database is joypad controlled UIs

Scientific software requires exploring data, one could imagine many interfaces where joypads, VRs controllers, etc. could be useful and more intuitive ways to do this exploration.

I disagree. Designing ergonomic interfaces is too often too far from designing playful interfaces that support the gameplay. Game designers != UX designers. UX designers already are the "game designers" of applications. And unfortunately they are not that often part of game development teams.

I’ve worked with dozens of “UX designers” and they are almost always just graphic designers who realized that they can make more money calling themselves UX designers.

Most of them don’t do much user testing, and the ones that do seem to use user testing just to provide support for their preexisting ideas.

Real UX experts, and real rigorous UX design methods exist because I’ve taken classes in grad school that teach them. But it’s basically impossible to find people who know how to do that in the sea of graphic designers calling themselves UX experts.

This is my experience too. True insights take hard work and time to gain.

I'm pretty sure (well certain) it's possible to do. We built a demo VR project to explore insect anatomy. It didn't get very far, was ultra crude, and we had no clue what to consider in the UI/UX of VR.

It melted our fellow scientist's minds, without exception they were floored by the potential, and this a complete hack. https://github.com/bioip/vronto

Sorry for the late reply, I lost track of my comment. Serious VR applications are around for quite some time now. There is a pretty huge research body on VR, usability in VR etc. Recent consumer VR makes it certainly more accessible but the already existing body of VR research is still applicable. IMO it's simply not a new frontier that game designers are necessarily the best experts in. Especially when you consider under which constraints games are developed.

I totally agree. Everybody who thinks game UIs are somehow good doesn't play games...

As long as you exclude UI/UX from "getting things done" it's going to be notoriously bad. Academia, government, enterprise all the same. There's no accountability to actual users and doing UX properly would turn the whole project on it's head so guess why no one cares...

| There's no accountability to actual users and doing UX properly would turn the whole project on it's head so guess why no one cares...

I don't think we're saying they don't care.

It's a balance, right? I've been around long enought to see the UI/UX world emerge and become more defined, but some of that world is common sense that comes with experience. In my experience some UI/UX is hot-air, and in fact gets in the way of getting things done, but other bits/and insights are super useful. What I'm suggesting is to tease out specifics with reference to systems that others are familiar with- games.

In that sense, personally, everyone cares. But when people have to operate in big org environments priorities change. Think about it, no one wakes up in the morning and thinks I'm going to make something terrifyingly hard to use today. No one invests or gives grants for something that "is going to be terrifyingly hard to use", right? And yet dominating majority of big org software is terrifyingly hard to use. The answer to why? I think is the point you are missing. Your suggestion is commendable, interesting but no number of references or examples are going to help make your next project better as well as that why.

As far as UI/UX world goes it has a problem of becoming fashionable, popular and somewhat profitable. This always attracts certain kinds of people and hot air. I'm sure you have plenty of your own examples in academia.

I'm admit I'm not totally clear as to your point. If it is that software is terrible because it's not accountable to its users then yes, I agree, this is often one of the reasons why.

In our case our group is funded primarily by an endowement. We use (dogfood) our own software, therefor we are accountable to ourselves. We are also developing it for our friends, collaborators, and other scientists who have traditionally not had access to the types of tools we work on, we are accountable to them. The bits that exist are because our users need them, have co-designed them, and have given us a ton of feedback as to how to iteratively improve them. We don't have to sell ($) them a product (yes, we understand there are long term issues with growth, support, and sustainability that will take hard work to deal with).

We're fortunate enough to have time to think and engage with concepts that startups that must have to sell products must master, thus we can "dream" about what game interfaces might bring to the table.

I am somewhat familiar with both, UI/UX and keep relatively apace with a few fields of research. Currently, I'm working to become good at front-end development.

This is an area I'd like to work on, if you have a project that could use someone like me please hit me up here (a comment with your email) or via twitter (@_areoform).

We just wrapped up our very first community building event. https://github.com/SpeciesFileGroup/taxonworks_together_2020. You can track us down from there if you're interested. All our software is open source, and we have a ton of open issues ;).

Thanks! Will do.

BTW, I'm fairly certain that the "github submit a pull request" model doesn't work for UI/UX. Consistency and tone matters, as does integrating feedback and cycling it. The idea of hitting issues piecemeal without some kind of overarching strategy feels like a recipe for failure.

Unless you have unified brand guidelines, a UI component library etc, such an effort seems like an exercise in futility.

That's a good point re PRs.

One possible way of interacting would be to explore the API, and develop a proof of concept UI element with a bit of functionality, that plays off it. Perhaps in a context like this: https://github.com/jlpereira/tw_api_test.

We do have use various semantics throughout the project, randomly documented. Contributions to how best to organize and get beyond this are also welcome, along the lines of this issue: https://github.com/SpeciesFileGroup/taxonworks/issues/85.

The issue is that it is hard to make money in scientific software, especially for life sciences. You usually get a bimodal distribution: really big, entrenched, and massive software firms and the rest have <100 headcount. There is no middle ground.

At First Principle we have tried to offer our design help for free for the scientific research community because we want to support the community. Whether robotics, ai or medical.

Most of the time there isn't really an interest from the people involved because they just want to do the research and not add the work of getting a proper interface working ontop of what they already have to deal with.

The offer still stands though.

See comments/links above in this thread re our first community building event. All feedback is good feedback.

I think this is a great idea. Imagine the amount of time and unnecessary confusion this could save.

Game UX and UI are often incredibly good, including seamless gradual on boarding.

Cheers to Pixo! As a UIUC alumn, it's great to see their work referenced on HN!

It was great to work with Pixo. They developed the logo, and homepage, and we ran an experiment together where they built out an interface against our APIs. I learned alot about all aspects of software development from the process, and they learned a little about our science. Would recommend.

a lot of things would benefit UI wise if paired with game UI devs

I wonder how many games have dedicated UI developers. GUIs beside the game itself are often clunky, complicated and at best playful.