101–110 of 263 posts
Anybody have experience using echarts with Vue 2.6? This looks way easier to implement than chart.js, but I’m worried about compatibility with such an old Vue build.
We [1] just added to ECharts as a charting library for our AI and are switching the default GUI charts over to it as well. We did a pretty extensive review before selecting it. ECharts won because it's excellent, and very pretty. 1. https://sourcetable.com
Earlier quoted context omitted.
I refuse to use it out of principle that they have intertwined it with observable. All the modern docs for d3 assume you are using observable. So unless you are relying entirely on AI it’s now even harder to grok the documentation. Not sure why this is so contentious. You can search online how many others this impacted as well. All of the modern examples for d3 were rebaked into the observable notebook pattern and it…
This is a very bizarre take. If you want to just blindly copy and paste d3 code, you may have issues with the docs being hosted on observable. But absolutely zero part of the d3 api (or core design patterns of use) have anything to do with observable. It’s like saying, “I refuse to use PyTorch because their docs are built with mkdocs.” Moreover, even if the coupling were limiting (which, again, it is not), it’s odd t…
Observable is cool when you want to build data notebooks. Observable is obnoxious if you want to add a D3 pie chart to your Vue application and have to untangle calls to D3’s API from reactive cell values, which look like ordinary JavaScript, but are not, and will cause compilation and runtime errors when copied.
The problem is that D3 resources mix D3 documentation with demos of D3 itself, and demos of Observable, into a single, inseparable combination. Nobody would complain were those three things separate resources; alas, they’re not.
Every time someone raises this issue, they are shot down by people like you, with the same nonsensical argument—just because the maintainers write cool demos in a fringe datavis DSL, that doesn’t automatically mean it’s helpful for the 99% use case of rendering charts in normal apps.
Do anyone have experience with this in comparison to vega?
I'm not an expert on either, but my impression is that vega tries to do literally everything using a custom json syntax which can get confusing and unwieldly. Echarts is a bit more straightforward at least for the simple case.
Vega has a questionable security track record which is concerning if you let untrusted users define graphs.
Earlier quoted context omitted.
This is a very bizarre take. If you want to just blindly copy and paste d3 code, you may have issues with the docs being hosted on observable. But absolutely zero part of the d3 api (or core design patterns of use) have anything to do with observable. It’s like saying, “I refuse to use PyTorch because their docs are built with mkdocs.” Moreover, even if the coupling were limiting (which, again, it is not), it’s odd t…
Look, it’s not “bizarre” to point out that Observable has deeply shaped the modern D3 ecosystem. The issue isn’t that the D3 API has changed, it’s that a huge number of learning resources, examples, and docs have been rebaked into Observable’s reactive notebook style, which, for people not using Observable, adds a ton of indirection and mental overhead. You might not notice it if you already know D3 inside and out, b…
I don't understand the relevance of any of this, but I think I've done a fair job outlining my points above. Let me give my best summary: the execution environment used by a library's (example) documentation is independent from learning the API of the library itself. I agree that newcomers to JavaScript may find plenty of confusion there, and I'm sure a decent chunk of D3's users may be new to web-programming in general, but it's not the job of D3 maintainers to account for that.
I actually think our back-and-forth is a perfect example of why open-source is so painful to work on.
D3 is one of the best documented libraries out there is. There are multiple books, hundreds of hours of youtube videos, and most importantly, dedicated maintainers (Mike Bostock, Philippe Riviere, etc.) who've poured hours into additional sources of documentation and are very responsive on GitHub issues.
The unfortunate outcome here is that users have come to expect this sort of high-quality support and documentation (again with the italics, who does this guy think he is???). Every D3 submodule has standard api documentation, sure, but thats expected of all libraries these days. However, the additional example documentation (again, nobody got paid to make this material) for the most recent releases has been migrated from bl.ocks.org (a now defunct open-source service users didn't pay for it) to Observable Notebooks. Now, Observable is a VC-backed business, yes, but the documentation is still completely free. In this thread, you mention you don't like this, so in at least one online conversation where D3 comes up, you actively advocate against using it out of principle! (Couldn't resist!).
Of course this is just my viewpoint on what has transpired, and I'm likely articulating it in a more-inflammatory-than-reality manner. But I'd prefer to have D3 documentation in the form of free, interactive Observable Notebooks rather than to have no documentation at all. Even more so if it helps out the authors of the open-source library.
As a tip for those who have difficulties going from the reactive Observable model to vanilla js -> you can always just take an observable notebook and ask AI to convert it to a set of vanilla files for you.
PS - Sorry for the novel, at an airport!
Earlier quoted context omitted.
Look, it’s not “bizarre” to point out that Observable has deeply shaped the modern D3 ecosystem. The issue isn’t that the D3 API has changed, it’s that a huge number of learning resources, examples, and docs have been rebaked into Observable’s reactive notebook style, which, for people not using Observable, adds a ton of indirection and mental overhead. You might not notice it if you already know D3 inside and out, b…
"You can use as many italics as you want but I don’t think it’s a wildly bold claim and your counter example is simply silly." I don't understand the relevance of any of this, but I think I've done a fair job outlining my points above. Let me give my best summary: the execution environment used by a library's (example) documentation is independent from learning the API of the library itself. I agree that newcomers to…
First, the italics. You opened with “This is a very bizarre take”—emphasis on very—which, whether you intended it or not, sets a smug tone. It’s the kind of rhetorical move that shuts down discussion before it starts. I’m not saying this to nitpick style, but to point out how quickly this moved from a discussion about tooling to a dismissal of my viewpoint. It is an easy tell the person writing gives away when they use it like you do.
Now, to the substance: yes, the D3 API is technically independent of Observable. But in practice, Observable is now the primary medium through which new users encounter and learn D3. The official examples, new documentation, and most teaching material are embedded in the Observable environment, which introduces its own model of reactivity, syntax, and execution. That’s not a minor detail, it’s a real obstacle for people who just want to understand and use D3 in plain JavaScript.
Your PyTorch/mkdocs analogy completely misses the mark. Mkdocs is a static site generator. It doesn’t change the code you’re learning—Observable, on the other hand, does. You can’t meaningfully learn D3 from many of these new examples without understanding how Observable cells work. That’s a tight coupling, and one I don’t think should be hand-waved away.
“Just use AI to convert the notebook” is a non-answer. It assumes everyone is okay with outsourcing understanding to a black box just to get a usable snippet. That’s exactly the kind of indirection I’m objecting to in the first place.
Lastly, yes, I’ve set a personal boundary: I don’t want to buy into an ecosystem that made a choice I don’t support. That’s not a call to boycott, it’s just me saying, this direction doesn’t work for me. If that’s enough to dismiss my view as bizarre, then I’m not sure what kind of discussion is even possible here.
ECharts is in my opinion the best out there. It surprisingly doesn't come up in a lot of lists or search for charting libraries. I've tried them all: chart.js, google charts, amCharts, Highcharts, ApexCharts. We use it in our tool/library https://docs.chartsql.com/
What about D3js? I know that D3js is low level, but with AI it's pretty easy.
The only real drawback that I've discovered is that it uses canvas to generate the charts. So, when you UI changes to dark mode, you need to reload your charts to update the color scheme ... which in the grand scheme of things is really minor.