Pharo MOOC
mooc.pharo.org
mooc.pharo.org
All three of the professional environments were very good. When Pharo kicked off, it was supposed to have been people building a free Smalltalk you could use to build real world software, as opposed to the research-oriented Squeak community. They didn't stick to that path, tho.
As another post described, Pharo is a mess, and it doesn't get more reliable from release to release. It's too bad. The idea that any Smalltalk environment today is state-of-the-art is just misguided. I would rather use Intellij IDEA any day.
Smalltalk is still a beautiful thing, tho.
Can you share this post, please?
And saying that VA or VW give a better user experience when coding vs Pharo, I beg to differ.
It is hard to lose code with Pharo. I have always recovered everything 99.999% of the time.
You are happy with VW, fine with me. I cannot stand it.
I also don't like the approach "let's write some code that passes tests and maybe works 80% of the time, and then release a few fixes every week for the next few months". Some of my classmates had to wait for 2 weeks just to get a working version of iceberg for the team project. And by working i mean it managed to clone the project, we kept getting "message sent to nil" and "object X does not understand message Y" errors for the rest of the semester and not just from iceberg. Any of those would be trivially caught in a static language.
Yes, there's a recovery feature. I found it slightly buggy too but it worked most of the time (and yes, i used it a few times). Still doesn't help when the whole image gets corrupted and pharo segfaults on startup.
Stephane Ducasse wrote:
> I told them NOT to use github.
> I told them that Bloc was alpha. They got a large list of topics and could have picked another one.
> Now of course we should improve and they too.
OTOH, please note that iceberg was only the tip of the iceberg (i am a huge fan of the name), the IDE itself did crash many times and corrupted my image once. The formatter moved comments into other blocks which changed their meaning and combined with CTRL+Z it deleted code after cursor in some cases reproducibly.
One extra funny though harmless bug was that shift+click didn't select a range of items items in some listboxes. It selected just the new one but left the old one highlighted. I clicked a few items and sometimes it selected the whole range. When I released shift and clicked more items, it played back the clicks i made earlier and selected those items instead. We joked that smalltalk did not only invent OOP but also command queuing in strategy games. Yeah, it was mean but we couldn't imagine what code had to be behind the listboxes to cause this behavior.
So yeah, maybe someone else thought OOP students would make a good typechecker for iceberg and designed the class accordingly but what i think and say about pharo still stands.
* Bloc is not a "Released" subsystem. While its development group report Bloc is stabilizing and encourage more people to start exercising it, Bloc is not on the Pharo 7 Release roadmap. Pharo 6 is current stable release.
* For a long long time, Monticello was Pharo's main source code control system. Responding to general criticism that Smalltalk sometimes lives in its own shoebox and doesn't interface well to "standard" systems, Pharo is spending significant effort on Iceberg to adopt Git as its main version control system (I don't know of any other Smalltalk that has a GUI Git tool). But Iceberg is quite a new subsystem with a Technical Preview release in Pharo 6.0. It was initially slow to draw new users from Monticello as many users were not familiar with git, so bugs remained that were only exposed with broader use. Last December I was trying Iceberg and also experienced crashes with it.
One thing to keep in mind is that Pharo's usually great stability is due to everything being implemented in itself on a garbage collected virtual machine, i.e. memory-violation errors eliminated. Iceberg makes use of the "libgit" external C library via FFI. While this aligns with Pharo's policy of not reinventing the wheel (another oft criticism of Smalltalk) but does allow memory errors that normally could never occur. So the Iceberg subsystem is not a good benchmark to judge the stability of the rest of Pharo.
btw, Lately I've found Iceberg stability quite improved and I started using it daily. Also, based on constructive feedback an improved GUI is planned.
* Around December there was also bug in the new Garbage Compactor related to the conversion of the VM from 32-bit to 64-bit. This is a complex system that took a while to identify, diagnose and correct.
Unfortunately I think you landed in a perfect storm. If the presenter was trying to show off the latest features I guess you got caught up on the bleeding edge of development. Part of the problem here is Pharo bleeding edge is usually quite stable (I use it daily), but we of the Pharo choir better placed to work around any hiccups in order to push Pharo forward. But these should not be exposed to newcomers and you don't have time to deal with them in a classroom setting. I'm sorry this damaged your impression of Pharo, but please understand how your broad statements damage Pharo in turn.
I see Pharo as two universes: the stable (now Pharo 6) and the development (now Pharo 7.) I work on the stable one and I don't concern myself at all with what is happening on the development one. The yearly release cycle of development-to-stable feels about right to me.
I have had to spend some time working out how to fit Pharo into my wider project build system, and how to store Smalltalk source code on Github as part of a larger repository such that you can handle merges etc. This made me curse a lot. It worked out fine in the end though.
I have been working on this for one year now and I haven't had problems with Pharo bugs or instability.
It's a nice tool! https://www.youtube.com/watch?v=e_e3_3MyTBY
btw, which was the CUIS prototype framework you used? There should be little reason it couldn't be made to run on Pharo.
https://github.com/hernanwilkinson/Cuis-Smalltalk-Denotative...
For those who don't know what I am talking about, there are several families of languages where the state of your program is kept within an image. The upside of this is that you can integrate everything together in a really neat way and build neat tooling with introspection. The downside is that you have just made what's going on not visible from the filesystem, and have to replace all of the tooling that you're used to. You know, things like your favorite editor, command line programs, source control...
If you switch languages regularly (which most developers do in my experience), being able to use your favorite toolset everywhere turns out to be more important than being able to build an ideal toolset for only one language.
I independently arrived at the same conclusion. The main problem with Pharo is the esoteric tooling; life is too short to struggle with tools.
That said, I consider Smalltalk, Objects and Design by Chamond Liu to be one of the best books I ever read on the subject of object oriented programming. Heartily recommended.
And I actually find things like Spotter, Code Rewriter, GTInspector and Calypso pretty neat.
IntelliJ is not simpler and is closed source. It is neat, true. We even use its icons in Pharo.
And that's the problem. The tools are different. They require learning. And once learned, they are only useful for dealing with one language. And that is a language that nobody around me uses.
The result is a barrier to entry. A barrier to entry of a type that I'm not bothering with again unless I have a good reason to.
Furthermore the small size of the community means that I am unlikely to get to choose tools that fit me. I'll be stuck with whatever the personal preferences were of the people who wrote the tooling. Even if I like them, my neighbor probably won't.
You bring up IntelliJ. However I don't use IntelliJ. I prefer vi and command lines. The guy sitting across from me likes Visual Studio. I couldn't tell you how many other IDEs are in use by programmers in the same space as me. No single toolset is going to be to all of our tastes. And it doesn't matter with a filesystem based language, because we are each free to use whatever we want.
User experience is one area where Smalltalk seems like it could use some focus on. Compare the screenshot of the system browser on pharo.org compared to figure 17.1 (page 293, PDF page 315) of the "blue book":
http://sdmeta.gforge.inria.fr/FreeBooks/BlueBook/Bluebook.pd...
Apparently that layout was perfected in 1980. Has anyone ever experimented with showing more than one method at a time in a window? And taking up less screen real estate with the class hierarchy? Seems ripe for some new thinking here.
And working on a fuse thing that exposes my running image as a filesystem so that I can such tools as well.
But saying that the tools are different and you cannot be bothered to use them? Look, Visual Studio is a monster beast, with tons and tons of features, including Azure ones. It takes a serious while to master.
And vim also takes a serious while to use well. Especially when we use a couple of plugins.
An engineer knows the tools. Then a specialist masters them. The master makes them. Or changes them.
Pharo allows one to do that. Without strings attached.
The benefits of learning a particular tool are essentially proportional to how often you use that tool. The greater flexibility of other tools such as vim means that the cost of learning it can be amortized over essentially as many projects as you wish. The costs for learning Pharo tooling are only spread over as many Pharo projects you do. If you don't believe you'll be doing all that much in Pharo, it makes it a fairly daunting proposition to invest that time.
But Pharo is all about the deep integration of the tooling. Its what facilitates such a tight Code->Compile->Run->Debug cycle. It can take a while for the penny to drop. If you do a drive-by taste of Pharo you probably didn't get chance to really get into the flow.
Some say... "The only programming languages worth learning are those that change the way you think about programming." And Pharo/Smalltalk tooling does that.
Smalltalk is much more than the OO model.
On the other hand if your intent is to learn about its collection classes, what MVC looked like in its original inspiration, and so on, then you should learn Smalltalk.
There is more to learn than any of us has time for. If you know what you hope to get out of learning something like X, then you should pick the most convenient X around to learn it.
While Pharo (and Squeak, etc.) have baked in editors and source control, Pharo has integrating with general tools as a design goal. It includes an integrated Git client and can load packages from or move them into a Git repository.
Integrating with general tools may be a design goal. Yet it clearly isn't fully there. A filesystem based language does better without even trying.
Iceberg is breaking new ground compared o Kraken/SourceTree etc.
Pharo took the risk of integrating git support in the tools.
It takes a few iterations.
So what? Either one is a user or one is inventing new stuff. When was it you wrote a full wrapper for libgit? And made it available to any app? Yes, any Pharo app now has the ability to use libgit features. Awesome and opening quite some doors.
All developers are users of someone else's software and knowledge. I'm OK with that, and happy to not reinvent perfectly good wheels. You go ahead and enjoy wrapping libgit.
My preference is to understand how to use data to help businesses. For example a decade ago I gave a tutorial on A/B testing which resulted in multiple companies adding many millions to their own bottom line.
It is a big world. There is room for all kinds.
Nobody forces one to use Pharo and I actually find it indeed useful to have an FFI binding to libgit.
Pharo has had git support for years with git filetree. There is no urgent need for a built in tool. Iceberg is an integrated tool which is not even yet released for the version of Pharo it is intended for (7).
The project is useful to test drive UFFI significantly. As such it is a really good test bed for all kinds 7of data structures and callbacks etc. Coupled with the 64 bit VM + image it is even more so.
I am in the Pharo ecosystem as a long play. Expedient solutions are plentiful. But they all lack this QWAN in Pharo that makes it enjoyable to me.
I am interested in understanding how things do work. Pharo allows me to do it in a way that suits me.
Given the count of "me" above, it is clear that I follow my own path and I couldn't care less about popularity.
Truth be told, I find it somewhat useful to have a harder to approach community. It keeps the signal to npise ratio at an acceptable level.
Yes, I'd like a cookie. But not for libgit, but libstrophe. It was easier (and clean!) to bind with Pharo than with a lot of other technologies. The ability to pin objects is also very usable.
Whoever designed the GUI also probably never used even git, or they couldn't stand their own tool. From the top of my head: 1) it shows all commits in a linear list (with broken unicode) completely ignoring branches 2) deletions are green and additions red.
I find little point in learning Smalltalk today, except maybe read a ten minute article to check out its odd syntax.
https://johnkoerner.com/csharp/the-c-interactive-window-in-v...
Previously there was the Immediate window but it is limited:
https://msdn.microsoft.com/en-us/library/f177hahy%28v=vs.110...
btw, see the section here on "Blub Paradox" (http://www.paulgraham.com/avg.html ) which means I'm not really competent to comment of C#'s dynamic objects.
It seems C# work by subclassing the DynamicObject class. What about dynamically extending all the other existing C# classes?
What if I subclassed a C# Queue as MyQueue and then while my days long simulation was running I wanted add an instance variable 'trackCount' to all existing MyQueue instances? In Pharo all you do is this...
Queue subclass: #MyQueue instanceVariableNames: ’trackCount’ classVariableNames: ’’ poolDictionaries: ’’ category: ’Example’
and program keeps running, this definition is updated in the editor.
1. write a test
2. run the test
3. test fails; debugger pop up
4. implement/fix the method inside the debugger
5. repeat at 2 until the test succeed.
it's usual that a coding session is entirely done in the debugger which is like an interactive coding environment where absolutely everything is inspectable.
Very stable. Quite powerful.
And it was/is a pleasure to work work with it.
One has to invest time and effort to master it. I would say that this is best suited to people who actually know what they are doing. Syntax is easy but frameworks are very advanced. So, do not go there without support.
I see comments that are not that kind to Pharo. I do to care to be honest. I do not think that Pharo is for the masses.
It has imperfections but nothing is at the core. And things are moving forward very well.
As a Pharo consortium member I vote with cash and usage.
These days for graphical programming, I prefer C with nuklear (https://github.com/vurtun/nuklear) over smalltalk.
If I remember right, they had a separate image per customer and it took a moment to bring up the VM if hadn't been to the site in a while.
It's a bit sad, I appeared to be a Google sheets meets office 365 flows on steroids before either was any good:
(i think that's from 2006 or so, archive.org has more info, but it's a bit painful to navigate on mobile)
Avi Bryant comments on hn from time to time:
https://news.ycombinator.com/user?id=avibryant
As I understand it, each customer got an image/vm - and data was just stored as Smalltalk objects. It'd been interesting to see where they might have gone if they coupled the dabbledb front end with gemstone/s.
[1] http://seaside.gemtalksystems.com/docs/OOCL_SuccessStory.pdf
However, only one. Make of that what you will.
Here are commercial companies that believe in Pharo (in the money on the table way of measuring belief)
- 9 bronze level (1k Euro/year)
- 5 silver level (2k Euro/year)
- 11 gold level (4k Euro/year)
- 5 platinum level (8k Euro/year)
http://consortium.pharo.org/web/reports
...any ideas what that means? (i.e. they stopped publishing them for privacy reasons, the consortium no longer has meetings, the consortium is no longer active, etc.?) Any ideas on what projects they are currently funding, and how many man-hours of programmer time are being purchased for Pharo in 2018?
I don't know what it means that there is not later reports. The consortium is definitely active. It arose because open source software is about people scratching an itch, and often the focus is on the fun stuff and not on the boring, difficult, dirty engineering required to make a product reliable and successful (e.g. operating a CI infrastructure). Pharo arose out of INRIA[1], one of France's publicly funded national research institutes. But their mission and governance structure is not suited to managing non-research engineers dedicated to working on Pharo. So InriaSoft[2] was created to fork off its successful software into consortiums funding pure engineering work on the software as an ongoing concern. At the moment the consortium has one full-time engineer, but hope to get another in the next couple of years if the consortium continues the current growth rate.
AFAIK (I am not directly involved), the main efforts decided by consortium members were:
* Stabilizing 32-bit to 64-bit conversion for Pharo, utilizing and contributing to the OpenSmalltalk-VM project.
* Stablizing Iceberg as our Git/GUI interface
* Maintaining CI infrastructure
btw, There is also the Pharo Association with 82 individual members. [3]
[1] https://www.inria.fr/en/institute/inria-in-brief/inria-in-a-... [2] https://www.inria.fr/en/news/news-from-inria/launch-of-inria... [3] https://association.pharo.org/
MOOSE, which is some sort of code analysis tool:
http://www.moosetechnology.org/
I've never been able to make heads or tails of it, but it's kind of an architect-level tool and I'm just a lowly engineer.
> This Mooc is demanding but rewarding...
So the language is easy to learn and yet the course to learn it is demanding? Interesting logic...
Nevertheless, probably a good exercise for anybody who wants to learn what real Object Orientation is about and who didn't learn Smalltalk yet.
Nevertheless, I find the argument on their website kinda odd and a bit confusing. I mean, I see a lot of Smalltalk advocates claiming the language is so simple to learn and at the same time I see a lot of programmers having trouble learning it.
Maybe it is just wrong to call it 'very easy to learn' just because the syntax is simple.
I think the first important step, is to understand the different columns of the 'System Browser'. From a quick search that video playlist might help you:
https://www.youtube.com/watch?v=0xF4fnGaE2w&list=PLqbtQ7OkSt...
Doing serious work is demanding.
But so is serious work in any environment.
"The only languages worth learning are those that change the way you think about programming."
The IDE randomly crashes, corrupts images (all your work is in the persistent image) and randomly ignores clicks making it insanely frustrating to use. Git integration is "WIP" which means it sometimes silently deletes files or you just don't get it to work at all. Branches are not implemented, glhf resolving conflicts on a team project. Oh, and it saves your passwords in plain text.
Error messages will keep opening until you kill the "IDE" and undo will sometimes delete half your code with no way back (maybe that's why you're supposed to keep methods small?). It's a bit ironic that Mr. Ducasse calls himself an "expert on software quality".
Here's our final project, the weekly reports document our "learning" experience: https://gitlab.fit.cvut.cz/taibrmar/sokoban-using-bloc
Most students had similar feelings and the response to the course was overwhelmingly negative.
There are a variety of proprietary language vendors with great products (Cincom Smalltalk, Dyalog APL, Wolfram language that is part of Mathematica), but it is a weird sale in the modern age. Why use a better tool if you have to pay for it and your entire business's IP is tied to another company pulling the strings. Even if they promise their tools are 10x more productive, is the risk acceptable?
Another problem with Pharo is the confusion with Bloc, Morphic, and Spec. With C# you have WinForms and that is pretty simple and Java has Swing. With Pharo, I can't seem to find a decent way to build a GUI. Is Spec supposed to replace Morphic?
And with Xamarin, a slew of other ways.
Not that less confusing.
I wouldn't have.
Because these things are still in flux.
These things are interesting and are the way forward. But not ready for end users.
But this has exposed you to the realities of creating new things: it is not as clean as one would want nor are users that cool when facing them. In a unusual environment like Pharo, it gives a kind of blurring effect since a user has a hard time distinguishing between solid ground and treacherous new terrain.
I would advise you to look back at Bloc and Iceberg when Pharo 7 is released.
The occasionnal freeze of the environment is indeed a pain and a new user would cringe more than once. I actually have a REPL server running in my images to restart the UI when such stuff occurs. Not the realm of a newcomer for sure.
Pharo is great for my uses. It is not the best tool for everything, granted. But when dealing with complex domains, I haven't found anything close.
Stephane Ducasse wrote:
> I told them NOT to use github.
> I told them that Bloc was alpha. They got a large list of topics and could have picked another one.
> Now of course we should improve and they too.
This is not saying that everything in Pharo is perfect, but in future could you please frame your criticism to indicate your pain with Pharo was with alpha features.