Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [ide-dev] Java IDEs comparison

As for the Eclipse UI Guidelines, it's good to revisit that, but I think they only address a limited portion of the UI issues: they help maintain consistency across Eclipse IDEs/plugins.

However, if we want to do something more fundamental, like change or redesign certain aspects of the Eclipse workbench UX , we can't do it with just the UI Guidelines. And without a "product manager" like Doug described it, it would be hard to get consensus on some of those changes. (not to mention developers to work on them)

For example, I think the workbench would work better with perspectives used less. Maybe reduced to just a General perspective and Debug, as was mentioned before. Or even further, to just one perspective, but that would require smarter view management (debug views or view-folders popping in and out according to debug sessions, etc.). Management of launch configurations also could use an overhaul. I'd like the see the name of the currently active launch configuration in the toolbar, like IJ, or CDT's LaunchBar. Having transient launch configurations is also a good idea. Another thing I liked with IJ's workbench was implicit auto-save (that's more than just auto-save).

On 9 September 2016 at 11:55, Bruno Medeiros <> wrote:

On 7 September 2016 at 16:38, Carsten Reckord <reckord@xxxxxxxx> wrote:
> It's surprising how the code completion of IntelliJ (checking target type,
> frequency of calls, chain completion...) is still listed as better than
> the one in Eclipse IDE. Wasn't Code Recommender providing that in all
> Java-based packages?

I don't have empirical evidence, just my subjective impression, but one problem with completion that I frequently encounter in Eclipse and that IJ seems to get better is the local context that I'm working in. If there are a dozen proposals for the current completion, we often don't have to look as far as global frequency, chains and all that:

Just prefer the type that is already imported in the current class over the one that isn't, the type or variable that I've used in the line just above where I'm asking for completion over other applicable ones, the method in my current class over some external type proposal, etc.

> Also, it's strange that IJ is congratulated for its "Polyglot
> development", where the Eclipse ecosystem has a much better offering. The
> reason is probably that people compare IDEs as they're shipped, not
> ecosystem.

In general, the IJ people shine in two "polyglot" areas here:

1. Consistency: editors and tools look and work the same across languages, whereas there are many (small but annoying) UX differences between different editors and language scopes in Eclipse.
2. Cross-Language features: I haven't tried this in Neon, but before, navigating between artifacts of different languages, getting usages etc was just way more seamless in IJ.

But I hope with the new extensible editors and the language server initiative this is about to improve dramatically :)

Indeed, particularly point 2. I think when the article referred to better "polyglot" support, it was referring to cross-language integration, not just number of different IDEs for different languages.

I don't do any Web work (be it backend or frontend), but developer acquaintances of mine who do, and whose judgement I trust, say that IntelliJ support here is vastly superior. Seamless integration of embedded Java/HTML/CSS/_javascript_/PHP/SQL/XML/etc code into other code.
I don't know the details myself, but I believe them. After all, IntelliJ has the PSI platform, a platform for building language engines which has cross-language integration as one of its design goals. Eclipse has nothing like that.


Back to the top