PlatformIO, open source ecosystem for IoT
platformio.org
platformio.org
Is there something extra I'm missing that you can't do with the existing tools? Or is it to get some kind of lock-in and promote a commercial product? I've only done a very little programming for Arduino and the ESP8266 (where you can get the normal toolchain running... with some effort), but I got the impression that the embedded world has been trying to move away from magic "platforms" that do things for you in the background
I've been in a row with their founder over the same criticism you offer.
PlatformIO is really nice and smooth and seamless, as long as you do exactly what they think you should be doing.
It would be a great product if it un-bundled all the functions. Of course that would make it not-really-a-product, and they couldn't charge money for it as easily. But instead it's a monolith that offers you little wiggle-room.
For instance for the longest time it didn't expect you would do unit testing. There was no way to include it in your build, other than hacking PlatformIO itself. They've since added unit testing, but only using a framework of their choice. Which is nice, unless you already have a code base, or a preference for a framework.
My friend put it quite succinctly:
"I understand it from a vendors POV: they know that customers feel the pain of integrating disparate systems, so the sell the product that has done all that for them. Once both sides have bought in, they are reluctant to acknowledge that what they have is a monolith that makes the integration problem harder rather than making it go away."
> Is there something extra I'm missing that you can't do with the existing tools?
There isn't. And in fact there are other build systems you can use without having to use either the vendor's tools, or start form scratch.
> Or is it to get some kind of lock-in and promote a commercial product?
Pretty much. They're a company, and are (legitimately) trying to make a profit.
Could you explain in details? Do you know that you can create own development platform or add support for new embedded board wihout our support? - http://docs.platformio.org/en/latest/platforms/custom_platfo...
Yes, PIO Core 2.0 was monolitic. We rewrote it from scratch and announced decentrizlied PIO Core 3.0 in September, 2016: - http://docs.platformio.org/en/latest/history.html#id7
> They've since added unit testing, but only using a framework of their choice. Which is nice, unless you already have a code base, or a preference for a framework.
Which framework/workflow do you use for unit testing?
Mayb ein a separate post. That's a lot details :-)
> Do you know that you can create own development platform or add support for new embedded board wihout our support?
But you would still force me to use PlatformIO as my build system, correct?
Now, assuming I accepted that:
How much effort is it for someone who doesn't know PlatformIO that well, doesn't know (or car about) SCons? Is it more or less effort than just using make/cmake/build system of my choice instead?
Can I be certain that I won't find out too late that I'm now locked in somehow? Are its processes truly flexible enough for what I need, and might need in the future?
Can I even (with reasonable effort) create new commands (e.g. platformio static_code_analysis) to shape PlatformIO to my needs?
> Yes, PIO Core 2.0 was monolitic. We rewrote it from scratch and announced decentrizlied PIO Core 3.0
Nice. Can you give me a quick pointer on how to just use the uploader portion of PlatformIO given I already have binaries at some location?
> Which framework/workflow do you use for unit testing?
The one that makes the most sense. If I get to pick, I like to use Catch. But since the decision tends to have already been made for me (through existing test suites, or coworker's preferences), I use whichever is most practical.
Or even several in parallel, as is the case on my current project.
Free users don't get to unit test? Ivan, please tell me I'm misreading this.
PlatformIO wasn't built as a commercial project. What is more, we didn't re-invent wheel or clone other project/idea. PlatformIO is a practical solution for the problems which were resolved in http://smartanthill.org (the place where PlatformIO was born). The problem, where you need to build the same source code into the multiple embedded development platforms/architectures using any OS (Windows, Linux, Linux ARM, macOS, BSD, etc) WITHOUT ANY DEPENDENCIES to a system. The only one requirement is Python interpreter.
If you try PlatformIO today on a clean machine, it will take ~1-2 minute to build/deploy the whole project. Our build system is super fast, ~4-6secs per generic Arduino AVR project. It's based on SCons construction utility. We have own optimization for it and, also, provided PR as contribution to SCons.
> but why didn't they do all this using the normal tool-chain like CMake/git/GNU-tools/crosstool-NG/etc.?
You can do that if you work with 1-2 board/MCU and use the same OS. If you have a project and your team uses different OS, or you need to be sure that your project has the same workflow not matter which system environment will be used, PIO is cool here.
You need to bother with make/cmake/build scripts/toolchains/upload tools and other things... Just tell PIO which board (currently, we support ~370 different boards/architectures/mcu) do you want to use and PIO will do all complicated work for you automatically. You can easy move from one OS to another, or even to work simultaneously from different places, IDEs.
PlatformIO that is not just a "multi-platform build system". It's own philosophy and own vision how embedded development should look. A lot of people do not agree with us and have an own vision and use own tools/instruments. That is normal, or better, we hear good criticism and trying to make PlatformIO better and better.
> Or is it to get some kind of lock-in and promote a commercial product?
Again, PIO was not built as a commercial project. We see our mission in an embedded world and do all work for free under 100% open source/Apache v2.0. Today, PlatformIO has own niche and own big community. We see big future in PlatformIO not only from a business point of view. PlatformIO allows makers, hobbyists, companies to focus on own ideas and own products and save time which could be spent to maintain multiple development platforms using different build scripts, tools, etc.
Today, we work on the other solutions for the business/developers who need to more professional instruments. We don't have an aim to monetize existing work (3 years) or even change the license. Whole this work should be moved to PlatformIO Foundation or a similar protected organization.
PlatformIO has own history, thousands of developers trusted in it and supported us. PlatformIO could not be successful without our community, they are our guide for the new achievements.
Regards, The PlatformIO Team
You know, it's funny. You explain this as a big advantage of PlatformIO, but it's actually the thing that worries me the most. I don't even know yet why -- kind of a gut feeling.
I guess it prompts me to ask question such as: will it really work properly for all 370 platforms? Or only kinda? Can it do everything I need it to? or does it limit all 370 boards to their lowest common denominator in terms of interface and exposed capabilities?
There's another argument though: Usually I need only one board. Maaaaybe two. So while setting up toolchains can be a thoroughly unpleasant experience, it's something I do once per project or so, and then I'm good for the next months or years. This vast support is nice, but rather irrelevant in practice.
We don't force you to use PlatformIO. If you like manually manage build scripts, toolchains, etc. and understand how to do that, you can switch PlatformIO to your own build system. I remember that I already explained to you how to do that (own `build_script` entry within PIO build environment).
> There's another argument though: Usually I need only one board. Maaaaybe two.
Thanks, I've heard your feedback. PIO is intended for developers who don't want to focus on the 1 board for 1 year. We don't live in 90-th, today each month we have new boards, MCUs, connectivity solutions, etc. PIO allows you an easy switch between them. If you don't know at 100% which hardware do you need for the next 1 year, PIO is the excellent tool here. You can experiment with dozens of boards and find better hardware according to the project requirements. They cost a few $$$ today.
Stay tuned for our news. We are working on the very interesting things which will be announced soon.
Yes, I vaguely remember. I think it was before 3.0 came out?
It's been a while, but IIRC it didn't feel like build_script was a good solution for me.
IIRC you said you were going to improve the documentation. I shall re-read it. Maybe it makes more sense to me now, and I'll change my opinion.
> Stay tuned for our news. We are working on the very interesting things which will be announced soon.
I'll be watching with interest.
Actually, no.
That was one of the points I was trying to make. Beause I have to trust you that PlatformIO works well enough on all 370 boards. And I won't know until I try, by which time I'm already deeply invested in PlatformIO.
It is! :)
http://docs.platformio.org/en/latest/librarymanager/index.ht...
Please note that you can manually update manifest and make PR to author's repository. Click on "Manifest" tab when looking at a library in Web Registry and you will see "Edit" button.
I'll provide the counterpoint. If I want to do serious work, then PlatformIO is too constraining, and doesn't offer a painless way out of these constraints. Other than not to use it.
Recently I installed it again and the instructions had changed to modify an existing installlation
I could not imagine embedded development without PlatformIO. With just so many types of hardware, each having slight dofferences and intricacies, it quickly becomes wasteful to use individual solutions. PlatformIO accomplishes so many goals at once, I am deeply grateful for this product!