Nothing "earthshattering" in either of these releases. But progress is definitely being made.
I have nothing to really back this up, but it looks like they are in the "death by 1000 cuts" mode of trying to speed things up. Little improvements here or there, little speedups across the board, refactoring bits of code to skip processes that aren't necessary when possible. Those kinds of things. If you are waiting for a big event like "hey we did X and now it's 5x faster across the board" i don't think you are going to see it any time soon (and if it does happen, you can be sure it'll be in the title!).
That being said, this release does bump the node version up, and the npm version. A lot of the startup slowness is from finding and loading node modules (at least from the bit of profiling I did months ago), and NPM3's flatter structure should allow maximum "reuse" of modules meaning less time reparsing identical modules.
I don't know if this will have any actual visible changes, but again, death by 1000 cuts.
I don't think it is, this is the only thing I'm interested in hearing from Atom. Every time I've tried it has been unusable to me, I need it to be at least close to Sublime performance and resource usage to matter.
There are still little things that I notice are slow. For example, when I search for text within a file, I hit Cmd-f, then paste whatever I'm searching for.. And then my previous search finally loads and overwrites my actual search. In other words, I'm quicker at pasting than the editor is into the search window, so I have to deliberately slow down my paste.
I'm actually trying VisualStudio Code now, mostly because Atom handles indentation terribly. Trying to copy and paste and preserve indentation is a lost cause as of now.
Similar real line: Does it still eat battery like a mid-westerner at a buffet?
I've thought about using underlying tech for a desktop app. I don't know how easy it will be to sell given that it would drain battery at probably 1.5 the rate of a normal Java or QT app.
not sure, but i think for buffet restaurants on the coast, "all you can eat" is at the customer's discretion; it's not a legal requirement like it is in the MidWest
I think it's unlikely the performance issues are related to the underlying tech (that is, Electron), since it's basically a custom build of Chromium that omits the normal web browser controls. Performance-wise, if your app is suitable for running in a browser, it should be fine for bundling with Electron.
As noted in a sibling thread, other Electron apps such as Spotify and Slack do suffer from performance issues, and the general notion is that Electron apps are "slow by default" - that is, slow unless you take explicit effort to make them fast.
Since when are programming editors handling large files and IDE-like features "suitable for running in a browser"?
I know there have been several (browser based programming editors) over the years, but all have been equally crappy and slow on medium to large files, and not that better at normal workflows either.
250mb of RAM for a chat app that can display 4k assets, renders every known written language under the sun side-by-side in multiple fonts, supports a stupid number of emoji, has an inline image viewer, realtime sockets kept open for notifications, includes syntax highlighting for a bunch of languages, a wysiwyg-ish editor for posts, one-on-one audio chat, and can work effortlessly with hundreds of people in a single channel (which is about as many as i've seen, but it could do more).
I think that's more than reasonable, and trying to get that number down to anything lower would be such a waste of resources for almost no payoff (what percentage of users do you think will have their lives measurably improved by having slack use 100MB less RAM?)
Oh, and mine is only using 100MB and has been running for a few days at least.
>250mb of RAM for a chat app that can display 4k assets, renders every known written language under the sun side-by-side in multiple fonts, supports a stupid number of emoji, has an inline image viewer, realtime sockets kept open for notifications, includes syntax highlighting for a bunch of languages, a wysiwyg-ish editor for posts, one-on-one audio chat, and can work effortlessly with hundreds of people in a single channel
Not of what you wrote is even remotely difficult to be achieved, even in a 50MB-using native app.
In fact most of those things come for free with the OS and relevant native libs are already loaded into memory:
"renders every known written language under the sun side-by-side in multiple fonts", "supports a stupid number of emoji" (the OS font-renderer can give all that to all apps for nearly free)
"can display 4k assets", "has an inline image viewer" (I can do that in a trivial app that's 1MB or less).
"includes syntax highlighting for a bunch of languages" ...
Really, none of these features, and neither all of them combined, warrant 250mb of RAM.
But again, would it really make any significant difference in the vast majority of users?
If slack had taken the time to devote another team to developing another application (because they still need to support the web), possibly multiple applications for each platform, would they be where they are today?
Would that time and money pay off in any way? I really believe it wouldn't. To save a few hundred mb of RAM, you are looking at triple or quadruple your development and maintenance costs, not to mention making it that much more costly to add new features. That doesn't seem like a good payoff.
Obviously in places where it did make a difference (mobile platforms) they went completely native, and it shows in both the positive and the negative. Their mobile apps look and feel native, and work quite well, but they lag behind the "web based" clients in features. (most notably the new voice chat options are not in the mobile versions)
In a perfect world every application would be custom developed for each and every platform, and no expense would be paid to ensure that there are 0 wasted resources, but we are far from a perfect world.
Having a slack app that runs on any desktop platform, feels the same between all of them, and allows Slack as a company to exist is a win in my book, even if it uses a few hundred mb more than it should (which again, for me it only uses 100mb of ram, that's 1/320th of my total ram on this machine, and only like 40mb more than the tab that i'm writing this comment on is taking).
>Would that time and money pay off in any way? I really believe it wouldn't. To save a few hundred mb of RAM, you are looking at triple or quadruple your development and maintenance costs, not to mention making it that much more costly to add new features. That doesn't seem like a good payoff.
I don't think writing native apps is so much more difficult that a web app like Slack. If anything, it's probably the opposite.
There are cross-platform (native) apps with custom UIs, like DAWs and NLEs that are an order of magnitude more complex in both their UI and their logic code, and yet are done with teams and by companies that have an order of magnitude less funding than a company like Slack.
It might not be more difficult, hell it is probably easier for each individual platform, but even if each one is 20% quicker than a web based one, all together you would end up spending 240% as much.
Plus then there is ongoing maintence costs, probably a ton more documentation costs (unless they all look and act the same), more coordination needed with new features to ensure parity, and more bugs and security issue surface area.
That's not a trade-off I'd want to make unless it was required for some other reasons.
I'm not particularly familiar with the relative costs of native vs Electron, but it should be noted that: a) many DAWs are single-platform; b) DAWs don't have backends, which is at least a significant portion (if not the majority) of what a service like Spotify or Slack spends development time on.
There's no doubt both of them are not resource-effective at all, especially compared to other apps made in other tech stacks.
However, both are pretty successful services, and since the parent comment expressed concern if an Electron app "would sell", these are examples that such apps are comercially viable (even with all the problems they have).
At the very least, one can use them to go to market faster and/or validate an idea. I do hope that Electron (or other similar techs) apps reach a point where they are viable long-term in terms of performance and battery-efficiency. I also dislike slow, memory-hungry and battery-draining apps as much as pretty much anyone, but one should not dismiss Electron as useless.
What were you doing with it that gave you performance problems? Basic navigation has millisecond response time. Auto-complete / live syntax checking for Python is instantaneous.
My peeve is with font rendering. The same fonts look better in pycharm. I hope they improve this.
At least moving around tabs is fast now. Other than that I feel Atom is pretty fast on my computer. Still some problems with large files compared to VSCode but I still like Atom more.
One thing I just tried: opening a bundle.js file. I accidentally do this at least once a day and it crashed Atom every time. The 1.11 beta handles it fine now. Good job Atom people!
I use it daily for PHP development, and I don't ever notice performance slow-downs unless I accidently click on a >1mb image and it tries to write the binary to display.
Apologies if this comes off as trolling. I really want it to be snappy so I can switch over.