14 ms·

Pain we forgot

364 points by Morgawr 12y ago

SSLy 12y ago
Edit: I've mismatched deps and imports, indeed that could be useful. Disregard my post. If I select a function from autocomplete, its dependencies should be automatically added to the project without any fuss. For the point, that that post makes, this one is at least available if you use IntelliJ with Java or Scala.
- RussianCow 12y ago
  IntelliJ does it with Python and other languages as well, although you have to explicitly hit Alt+Enter -> "Import this name".
  - jamii 12y ago
    That's true for pulling functions from other installed namespaces, but to my knowledge you can't install a library in the same way (eg search for 'parse html', drag a result into the editor, get Beautiful Soup added to your virtualenv and imported into the repl).
    - RussianCow 12y ago
      Ah, I didn't realize this is what the parent was talking about. That's pretty damn cool.
      - jamii 12y ago
        I don't know if the parent was talking about that, but it's what I was trying to get at in the post.
- edanm 12y ago
  You know, I've recently started thikning - why do we even need explicit imports? Isn't the compiler/etc smart enough to figure out the right import 99% of the time? Every file in a large project starts with dozens of redundant import lines. Why don't we have imports go the way of the inferred type?
  - riquito 12y ago
    What if foo() is present in two different libraries? (is this the 1% you're talking about?) What if tomorrow I add another library which implement his own foo()? (this is trickier) When I read the code, what is foo()? (here the IDE may be smart and show me the right documentation). All that said, I feel that the pythonic "Explicit is better than implicit" is better. It's also true that when I see a file whose first 30 lines are used for including dependencies it doesn't sound right. The problem lies with the fact that we use text to store structured data. If we were to save source code as binaries we could have the source code in a "file|object|item" and the dependencies in another item, linked to the source code (I'm not advocating this, source code in text files has many advantages, but also his drawbacks).
    - jamii 12y ago
      We give each name a unique id under the hood. At edit time the user chooses which one they want from a list. The actual AST only ever refers to ids and the names are effectively just documentation. This has the neat side-effect that ids are globally unique so you can figure out what libraries you need to pull down just by looking up ids in your package repository.
  - Cacti 12y ago
    Depends on what you're looking for. The issue isn't really "importing" the code references as much as it is a scoping issue. That is, imports are largely used to define scope for the file and mostly for the programmers benefit (so you're typing, say, Button.DoSomething instead of System.Windows.Controls.Interactive.Button.DoSomething). That and the odd namespace clash that a compiler can't really resolve, but that is more due to implementation and system issues, not programming proper.
ggchappell 12y ago
What a wonderful article. I'm not allowed to forget a lot of this pain. I teach programming, so I see it anew every semester. And of course I still experience much of it in my own work. So, yes, let's deal with these issues better. One little disagreement. In the "What?" and "Why?" sections the writer present some ideas for debuggers. While these are good ideas, I prefer to think of the inadequacies of existing debuggers as motivation for good practices: modularity, loose coupling, unit testing, etc. Certainly, it would be nice (say) to be able to examine the entire state of a running program easily. But I would rather code in such a way that I do not need to. So to those who would write the World's Greatest Debugger, I say, "Good for you." But even better would be to turn your efforts to producing languages, libraries, frameworks, and programming environments that make such a debugger superfluous.
- jamii 12y ago
  > I would rather program in such a way that I do not need (say) to examine the entire state of my code. But if you could examine the entire state, why wouldn't you? People build correct intuitions by comparing their mental model to reality. If you don't ever show see reality, or only peek at it through sprinkled print statements, then it becomes harder to build an accurate mental model. This applies whether we are talking about a beginner trying to understand how pointers work or an expert grappling with a complex system. Programming is always limited by our ability to comprehend complexity. So enabling simpler systems is a win, sure, but so is getting better at understanding things.
  - ggchappell 12y ago
    > But if you could examine the entire state, why wouldn't you? Well, I would, of course. As I said, I think his debugger ideas are good ones. OTOH, there is a limited amount of effort available for improving programming, and I would prefer that such effort be directed toward other ends. Part of the problem is that, in my experience, debuggers tend to encourage sloppy thinking. And the better the debugger is, the sloppier is the result. Using a debugger, together with thinking hard about a problem, can be a powerful combination. But all too often, I see debuggers used as substitutes for thinking hard; the way debuggers are designed seems to encourage this. To put it in the terms you used, constructing a mental model and comparing it with what the debugger shows, is a great idea. But in practice, I'm faced with a choice: work on my mental model now or fire up the debugger now. The latter is less taxing, and is often chosen; this is likely harmful in the long run.
fragmede 12y ago
> I want to just type 'email' and see a list of functions and libraries relating to email. Do people even remember what life was like before Google? I type 'email python' and get back a link to an email module. Am I being closed-minded in thinking that can't get much easier? Yeah, I need to understand a bit as to how email works, smtp and imap/pop and what not, and how to send vs receive email, but some level of understanding is just necessary.
- cjslep 12y ago
  I have a similar experience with Go in my spare time projects (very refreshing coming from a FORTRAN/C++ day job) which is a boon, but falls short of what the article describes. The way I interpreted it, even having to do the search in the first place is a problem to be solved. Being able to start typing "email" and have a context menu with functions available from a list of third-party libraries relating to email would be the goal. That way when you select "SendEmail()" in your code, it automatically pulls the library, installs the binary where it needs to be located (or pulls source and builds) without interrupting your workflow. With package management systems standardized within a language (such as Go) the problem domain is reduced, but extending the solution across multiple languages sounds like another third-party tool than a language feature.
  - _random_ 12y ago
    You mean something like Bing Code Search but with NuGet support? That would be cool. http://techcrunch.com/2014/02/17/microsoft-launches-smart-visual-studio-add-on-for-code-snippet-search http://techcrunch.com/2014/02/17/microsoft-launches-smart-vi... http://visualstudiogallery.msdn.microsoft.com/a1166718-a2d9-4a48-a5fd-504ff4ad1b65 http://visualstudiogallery.msdn.microsoft.com/a1166718-a2d9-...
    - jamii 12y ago
      Yes, exactly like that. I can't wait to see how Bing Code Search evolves as it starts to learn from its users.
  - 12y ago
rch 12y ago
I thought this project was about improving the programming experience for programmers, not students (or interns) who will in all likelihood never excell as programmers. Edit: after downvotes I decided to read the whole thing afterall, but remain unswayed. This guy has an important opportunity and I sincerely dislike seeing it squandered.
- alecdbrooks 12y ago
  I've been programming long enough that I know the magic incantations to get started in a few languages. But I am still in Anon's shoes any time I learn a new language or library. As the article says, "Even for experts, programming is an exploratory process." So, I'd say everything in this article applies to programmers, too. Categorizing people into programmers and non-programmers neglects the people who need programming but not at an expert level. Some interns and students won't excel as programmers, but they still can bring programming into what they do instead, especially if efforts like Light Table succeed. I suppose I shouldn't be surprised by this kind of pushback against reimagining programming — I've seen it as a response to some of Bret Victor's projects. Skepticism is great: Maybe the ideas behind Light Table are wrong or badly implemented. But we gain nothing from glib dismissals that seem to regard any attempt at improving or simplifying programming as beneath them.
  - rch 12y ago
    I disagree with language like 'reimagining programming' in this context almost as much as I regret seeing Bret Victor's name being dragged in for no reason -- to the best of my knowledge, Bret Victor didn't design Light Table, and doesn't inform its progress. There is progress to be made, and I'll be excited again when I see some.
    - alecdbrooks 12y ago
      I've seen the same "oh, this is only for beginners" reaction to some of Victor's ideas on making data visible and making libraries "discoverable" that seems needlessly dismissive. That's why I mentioned him. Also, Light Table's initial concept was inspired by his work[0]. I'm not always impressed by Light Table, but I appreciate the work the team has put into the problem and think they have good ideas. [0]: http://www.chris-granger.com/2012/04/12/light-table---a-new-ide-concept/ http://www.chris-granger.com/2012/04/12/light-table---a-new-...
SolarNet 12y ago
Personally, I think this is an extremely important topic, we need to change the way we program. It isn't the 70's anymore, why do we still program like it is. Why does it take teams of developers to create and manage applications which do really simple tasks in the grand scheme of things (and I realize the amount of complexity in building applications is staggering, but large portions of it could be better automated). Where are the auto generated GUIs, where is the ability to ship execution control to arbitrary devices, where is a hypermedia layer with independent view and presentation code? I'm approaching this from a different angle than the light table guys appear to be (I agree with everything they are saying). My angle is an attempt to build a cross platform module system (where platform includes runtime and programming language as well operating system and architecture): https://github.com/OffByOneStudios/massive-dangerzone https://github.com/OffByOneStudios/massive-dangerzone My argument is: before we can build the next generation of useful tools, we need a framework for managing both generations. massive-dangerzone is an attempt at bringing next generation usage (like that described in the article) to existing tools. It's still a big work in progress though, and is barely useful at the moment.
weland 12y ago
I also wish that programming were a lot different today than it was when I started learning it. That being said, a lot of this article's points are things I've heard before. They led to the development of Visual Basic & co., mostly by people who had no contact with the Smalltalk and Lisp environment in the 80s, while people who did were shrugging and throwing tantrums like WHY THE FUCK DIDN'T YOU FUCKING LIKE IT TEN YEARS AGO? IMHO, all these things went down to the bottom of history because things like these: > Anon the intern needs to be able to open up Programmingâ„¢ and click 'New Web Form' are adequate for people who usually don't program, and extremely inadequate for people who usually do. Generally, and for good reasons, programmers will dislike a tool that hides implementation details for ease of operation. Past a certain level of complexity, the time spent manually doing the right cruft becomes significantly smaller than the time spent manually cleaning up after a smart tool. I sympathize with Anon the intern, but perhaps he should rethink his expectations about complexity; if discoverability is a problem, perhaps he could switch to something that's better documented? And at the risk of sounding like an elitist schmuck who rants about how things were back in his day, maybe he ought to start with something other than web programming. The size and complexity of that tech stack is humongous, to the extent that a large proportion of those who use it don't understand it more than two layers of abstraction down. Programs are also hard to pack and the environment that runs them is hard to setup. Because it involves at least two servers, possibly with several add-ons in order to allow the server-side languages to run, learning at least three languages (assuming server-side JS is an option), two of which (HTML and CSS) aren't quite being used for their original purpose. This is a beginner's nightmare and it has exactly nothing to do with the development tools. And then there are things that are far harder to solve than they originally seem: > I want to just type 'email' and see a list of functions and libraries relating to email. Related how :-)? Should MIME-related functions, needed to reason about attachments, also come up here? HTML parsing/converting, in case you need to deal with HTML email? Information cluttering does nothing to alleviate the opposite problem of information breadth: if Anon the intern's problem is he doesn't know how to Google for libraries or how to make efficient use of documentation, an IDE that presents him with a gazillion of possibly related things won't help him. Especially when, like all beginning programmers, one of his main difficulties is correctly defining the problem he's working on which, in turn, makes it likely for the solutions presented by the IDE to be nowhere even close to the one he needs, because the IDE (like Anon himself) thinks Anon is trying to solve another problem. There is, on the other hand, a lot more truth in this: > Tightening the feedback loop between writing code and seeing the results reduces the damage caused by wrong assumptions, lightens the cognitive load of tracking what should be happening and helps build accurate mental models of the system. I do think that the real resolution to this problem is writing simpler programs whose state is easier to track. On the other hand, programming tools today suck considerably at presenting program meaning. Things like evaluating what an expression comprising entirely of constants, or at least evaluating it based on the default values of the variables involved, are well within reach for today's tools, and yet programmers' calculators are still employed because 99% of the available IDEs couldn't evaluate ADDR_MASK & IO_SEGMENT if the life of every kid in Africa depended on it. This is wicked cool: http://repository.cmu.edu/cgi/viewcontent.cgi?article=1165&context=hcii http://repository.cmu.edu/cgi/viewcontent.cgi?article=1165&c... . However, I also find myself thinking that the very fact that we need debuggers that are this smart is proof enough that we don't reason about our programs well enough. Except for the fringe case of having to quickly debug a (possibly horrible) codebase I haven't written, I'd much rather prefer being good enough a programmer to avoid the need for a debugger that can tell me why foo is not 8 despite the fact that I fucking said foo = 8 ten lines above, than being a programmer with good enough tools to help me when I'm stupid.
logicallee 12y ago
I'll delete this if you don't like it, but I think this is why non-engineers like Steve Jobs, Jony Ives, and a person at HP I won't name, were able to make remarkably good consumer products without knowing how they work. It may be better for a non-engineer to design a laptop than for someone who actually knows how it is put together and exactly how it works. Of course, in a narrow sense, this means such a person isn't really designing it at all: the real engineers are, which may cause resentment. There is a very good chance that a non-programmer can design a programming IDE that is two or three orders of magnitude better (by whatever standard) than the status quo. This means such a person can't actually implement any part of it, or even know exactly what it's doing. Quite a surprising conclusion. By the way I have experienced this myself, when designing for a target I didn't know yet: after/while I was learning it, the resulting design iteration process was much worse than when I didn't know the implementation details. It's harder to think from the user's perspective, after you have been forced to think from the implementation's perspective.
- jamii 12y ago
  I'm starting to agree with this. If I was working on this alone I would have arrived at something just as complicated and baroque as what came before it. It was only when we started putting our prototypes in front of real users that I started to realise just how much of what I do every day is completely unrelated to the problem at hand. Having the that constant feedback is really vital to challenging existing preconceptions about what programming should look like.
- dreamfactory2 12y ago
  I don't think it's quite ignorance of the implementation domain is the key here, but rather outside-in vs inside-out thinking. Nearly all engineers solve problems from the inside in my experience.
  - logicallee 12y ago
    It's not just that though. Something may be impossible to do in the application, except with knowledge of x, but trivial for someone who knows x. If you know x you might totally forget that without this knowledge the task is basically impossible for the user. So if building the application engenders learning x, then you might totally miss this design problem, unless you step away long enough to forget x again yourself - and that may never happen. It is a bit too much to ask for one person to learn all of the details and totally own the implementation space, and yet be able to instantly forget all those details and go back to when they didn't know anything, to look at things from the user's perspective.
unwind 12y ago
I was somewhat surprised, but very happy, to see Verse mentioned in this context (although he did get Eskil's last name wrong, it's "Steenberg"). I co-developed the initial version of Verse with Eskil; I think it was a bit before its time perhaps. It was hard to get real traction for it, but at least the things Eskil has gone on to build on top have gotten some real attention. Great, and very well deserved!
- jamii 12y ago
  Oh man, I am full of typos today :S Verse was a huge inspiration for me and it's a shame to see it largely ignored. We're exploring similar ideas for getting our language tooling away from plain text and towards structured, networked representations. I would love to hear your thoughts on how to do that, or any advice you have from building Verse - jamie@scattered-thoughts.net EDIT I also thinks Eskil's ideas on 'what we can do' vs 'what we can get done' applies just as well to the rest of the programming industry. We're so focused on scaling up to millions of lines of code that we end up with environments where lunch_app takes a week to build and deploy.
cclogg 12y ago
Oh man, I've definitely felt this pain many times lol: "The samples will be missing lots of implicit information such as how to install the necessary libraries and how to deal with missing dependencies and version conflicts. Transcribing and modifying the examples may lead to bugs that suck up time. It's not terrible, mostly thanks to sites like stackoverflow, but it's still a lot of unnecessary distractions from the task at hand." So many times, the actual programming isn't tough, it's just getting all of the stuff around it setup that is hard.
- tluyben2 12y ago
  This is what I have mostly with Rails; when I started working with it I was good at Ruby but didn't use Rails and for a beginner it's horrible. There tons of options for everything, everything is badly documented, most things scratch an itch so when you use it, some of the completely logical things you expect are missing. Samples of gem usage randomly skip steps which can be a hair pulling exercise for a beginner. Then when you finished your app and try to update gems a week later, everything breaks completely. And this is not an attempt at trolling; I now use Rails daily and build rather huge projects in it and this still upsets me. When things work and you know how they work, all is easy and great, installing and using. When either or both of those are not true, you are in absolute hell, especially as a beginner. And nothing of that has to do with actual programming and it shouldn't be needed; your environment should solve this for this imho.
dclowd9901 12y ago
Are these guys ever going to actually finish the IDE, or postulate on what it means to program, ad infinitum. I'm glad they're being considerate in their design, but this whole project is starting to reek of over-aggrandized vaporware.
- seanmcdirmid 12y ago
  When Chris and company announced the light table project, they had very high goals. I knew then that this would take a long time, and some of the goals would even be unachievable without programming model changes. The advantage of being incremental is that you can reliably produce something in a reasonable amount of time. This is not incremental, and it's quite risky; be patient.
- woah 12y ago
  Are you mad that they're blogging?
- jamii 12y ago
  https://github.com/LightTable/LightTable/releases https://github.com/LightTable/LightTable/releases There will be another Light Table release in the next few weeks, featuring CodeMirror 4 (better performance, multiple cursors). We've only been working on Aurora since February. Designing a new language takes time. We may show some early demos in the next few months, depending on how the current prototype works out.
  - nazgob 12y ago
    I think that some people (including me) are a bit disappointed that some(most?) of your effort goes to Aurora instead of LightTable. Don't get me wrong, Aurora sounds cool but for professional programmers LightTable is much more appealing.
    - jamii 12y ago
      I can sympathise with that. The reason for our decision is that there is a low ceiling on how much we can improve matters with Light Table. The problems we are trying to solve turn out to be pretty baked in to existing languages and toolchains and we can't hack around them in the editor. Aurora is a riskier project but it has much more potential to help people. Also, the fact that Aurora is intended to be simple and easy does not mean that we are building a toy. Quite the opposite - the intention is to remove some of the incidental complexity in programming to leave you free to focus on the actual problems. We intend to bootstrap the compiler and editor by way of dogfooding. I look forward to doing most of my programming in Aurora in the future.
politician 12y ago
"The samples will be missing lots of implicit information such as how to install the necessary libraries and how to deal with missing dependencies and version conflicts. " Docker pretty much solves this problem. Writing samples for your new project? Start with step 1 `docker pull yourproj/your_tutorial_image`.
gedrap 12y ago
I totally agree. It's true even for us, experienced Web guys (and gals), in this mess ruled by packages and their managers. For example, I am trying to make my first Rails app outside a tutorial. Okay, I want to use bootstrap. Should I use gem? Guess that's the ruby way. Okay seems like there a few of them. Tried one, another one. Doesn't work. Don't know why, since I am just copying half-cryptic stuff because I am new to Rails. A friend suggests to use Bower. It's easier. Right! I had totally forgotten about bower, it rocks! Google: bower rails. Okay, there is that thing sprockets which I apparently need to configure. Googled, a few blog posts opened, they offer a bit different advice each, let's try. Doesn't work. Nope. Not really... Google: bower rails bootstrap sprockets. And yay! Works. Two hours later, I have included Bootstrap. Properly.
- joevandyk 12y ago
  Download bootstrap css, put into app/assets/bootstrap. Should be all that's needed.
  - gedrap 12y ago
    Yeah, it would have worked. But I am too lazy (among real disadvantages) to download all the libraries manually :) Real disadvantages such as updating becomes so boring that you will probably never do that. And 'bower search <something>' followed by 'bower install <found it!>' is less distracting than googling and unzipping etc. Being lazy pays off.
    - Cacti 12y ago
      If you're a programmer and you can't even bother to download some files and unzip them, then... Lazy pays off because you innovate in the hopes of automating something, not that you just are actually completely lazy.
  - igravious 12y ago
    And repeat for every update of Bootstrap? Why?
naturalethic 12y ago
Pretty sweet article. Needs a once over from the author for some edits though. "if you don't have something useful to say, don't say nothing at all" made me giggle.
- bridger 12y ago
  I can't figure out what that one is supposed to mean!
Goopplesoft 12y ago
> We still program like it's 1960 because there are powerful path dependencies that incentivise pretending your space age computing machine is actually an 80 character tty. We are trapped in a local maximum. Love this (although I don't like extra long lines). He's talking to you Pep8.
dreamfactory2 12y ago
Great article but isn't a lot of this more about using different levels of abstraction i.e. frameworks, DSLs and specific to those? That could either take the form of being sufficiently focused that a simple text editor is adequate (e.g. high level commands, no cruft), or a full-featured managed environment (state, docs etc) specific to the framework. I'm not seeing how this is a task for a generic IDE unless I missed something about light table and they are creating a new language.
- jamii 12y ago
  > unless I missed something about light table and they are creating a new language. http://www.lighttable.com/2014/03/27/toward-a-better-programming/ http://www.lighttable.com/2014/03/27/toward-a-better-program... The short version is yes, we are making a new language. The kind of workflow I discussed in the OP was the original goal for Light Table and it turns out that it can't just be hacked onto existing languages.
  - rwallace 12y ago
    Fair enough. I won't say your decision to make a new language is right or wrong - I don't know how it will turn out - but given that you have decided to do so, and given your objectives for it, my one big recommendation would be to make relational databases central to the language, the way Perl makes string processing central and PHP makes web pages central. Some possible ideas are outlined at http://geocities.com/tablizer/top.htm http://geocities.com/tablizer/top.htm - you don't have to go with those particular ideas of course, but I think it's important to make relational data central one way or the other.
    - jamii 12y ago
      We are heavily inspired by http://shaffner.us/cs/papers/tarpit.pdf http://shaffner.us/cs/papers/tarpit.pdf
mkozlows 12y ago
Articles like this one annoy me, because it's easy to diagnose big problems -- here, watch me do it: "Why should compilers choke if you forget a semicolon? If they can diagnose a syntax error, can't they also fix it? Can't functions be smarter about seeing when they're used improperly and tell you while you're writing the code instead of when you're running it? Why can't the code be understandable to anyone who knows English?" What's hard -- and often impossible -- is fixing those big problems, because a lot of times they're genuinely intractable; and when they're not, they're often so difficult that they might as well be. So just sitting around and complaining about them sounds insightful, but it doesn't really get anything done. And yeah, I know that they're allegedly working on "fixes" for these issues, but based on the track record so far (LightTable promised all sorts of revolutionary views of what an IDE could be; it's delivered... not a whole lot of revolution), I don't have any faith that Aurora is going to amount to much either. And I don't want to be too negative, because sometimes a previously-intractable problem turns out to now be tractable, and it takes someone who was willing to question long-accepted pain to find that out. So I'd be pleasantly surprised if this project delivered something that had even as much effect as the development of, say, git or xunit. But I'm not holding my breath.
lemming 12y ago
I'm glad to see IntelliJ getting some love there for making much of this easier - few companies have worked so hard at it, and the results are amazing. Looks like Microsoft is making some great steps too, Bing Code Search gets a bad rap for pandering to blub programmers but I'd love something like that in my editor.
ilaksh 12y ago
Can't be sure but it seems like the surveying has resulted in most of the right conclusions and the right direction for the project. HOWEVER, I still think most everyone is missing the REAL issue here. As evidenced by the powerful tools cited in this article that address various aspects of the "programming" problem, there have been numerous efforts to move the state-of-the-art in software development forward. And to a great degree those efforts have _proven_ quite a few superior paradigms. And yet, we haven't seen those new paradigms become truly mainstream for most programmers. Why? I do NOT believe it is because the approaches haven't integrated the right new concepts, because there are so many existing useful tools combining many different new ideas effectively which have failed to become mainstream among programmers. I think many of those new approaches could and should have become normal operating mode for programmers. I think the reason they did not is this: the core definition of "programming" (and by extension "software engineering" etc.) is an antagonistic and largely manual process of creating complex textual source code that can be used to describe system behavior. Period. For example, if I want to call myself a web "developer" I DARE NOT use an interactive GUI tool to generate and maintain the source code for my web site. Web developers will always say, for example, they avoid this because the source code generated that way is less maintainable by hand. In many cases that may be true. In some cases with advanced code generation it is not. Regardless, I don't believe that is the actual reason. Its just a rationalization. What do we call someone who's entire job entails creating a web page using a graphical user interface? In other words, this is a hypothetical person who has found a hypothetical GUI tool that can accommodate all of his web site design and implementation needs without any manual edits to source code. If he builds a web site or web application this way, and writes zero lines of code, do we refer to him as a very smart and advanced web "developer"? Or do we call him a web designer or simply a WordPress user (for example)? We do NOT refer to him as a web "developer". He has no right to refer to himself as a "developer" or "programmer" because he has not wrestled through an antagonistic manual process to create complex textual source code. And that's what people are not understanding. The reason we don't call him a programmer is NOT because he didn't create an effective program or website. Its because the way he did it wasn't hard enough and doesn't match our outdated definition of what "programming" is. To create a POPULAR system (among "real" "programmers") that makes programming more practical or easier using new paradigms, you must either redefine programming to include the possibility of new paradigms and a non-antagonistic process, or perhaps somehow trick programmers into thinking what they are doing is actually harder than it is. Maybe if there are a few places to type a command to generate source code, that will be sufficiently complex to still be considered "programming". If you are too successful without doing those things then you will just have another tool that only "users" or "beginners" would ever admit to using.
zem 12y ago
solving this problem was part of REBOL's dream.i really wish it had been properly open-sourced from day one and built up a good community and momentum around it; it was a very promising language.
shortstuffsushi 12y ago
The analogy to me seems similar to that of programmers and compilers. I want to write programs to do xyz, but I don't particularly care to know the lower-level to machine level implementation. Instead, I know a roughly-english set of instructions to tell the compiler what I want to do, and it takes care of the rest for me. Could it not be possible to take this even further, to the point where I say "I want to do ..." and all of the "code," be it high or low level is generated for me. I realize making something capable of this level of abstraction would be incredibly difficult, but it would certainly be fun to try.
- GFischer 12y ago
  There are companies trying. In Uruguay, a tool called Genexus enjoys a high level of success promising something like that :) . For simple CRUD apps, it's pretty much magical - and it looks awesome in demos, heck, it can generate code for every mobile phone out there just throwing around some controls, specifying your data fields and writing a few rules, but if you want to make it do something that isn't explicitly considered, it's very annoying (and it has some UI and UX problems). And don't ever try to touch the autogenerated code :P . That said, it's probably the easiest way to get from 0 to a complete CRUD app, once you learn how it works. http://www.genexus.com/products/genexus?en http://www.genexus.com/products/genexus?en
zan2434 12y ago
Saved. Can't wait to look back at this in a few years.
webmaven 12y ago
While I am watching the development of LightTable with interest, it seems to me that the example used in the OP is missing the forest for the trees. It isn't hard to imagine Anon Intern managing to cobble together a solution for lunch_app that isn't an 'app' at all, just a form created with a builder like WuFoo or Google Forms, and some 'recipes' in a tool like IFTTT that has built in integration points for email, the accounting system, etc. Arguably this wouldn't much of an improvement over the VB6 + Access status quo (or Visual FoxPro, or FileMaker Pro, etc), except that the individual components can be much more robust and scalable, monitored, auditable, and so on, without Anon Intern having to worry about any of that.
- jamii 12y ago
  I mentioned that near the beginning: > There are probably dedicated apps that cover this particular example but we are more concerned with how an end-user would solve this kind of problem in general... VB6, Access and Excel are exactly the kind of tools we seek to emulate on the usability front. Half of the world runs on Excel precisely because it is easy to get started, easy to understand and yet is capable of handling fairly complex simulations. If we can make something equally approachable but backed by a sane, maintainable and powerful language then we can give programming powers to a lot of people who need them.
  - webmaven 12y ago
    Except that I am not talking about a dedicated app at all, but general purpose tools/services for parts of the problem that can be combined arbitrarily, just like libraries can. IFTTT is exactly that sort of service.
rafaelferreira 12y ago
Not often I can read an essay that leaves me at the same time so excited and so disappointed. Excited because I found myself agreeing with every point: isn't it obvious to everyone that the programming experience could be so much better? We created wonderful tools for graphical expression and number crunching, that keep getting better [1] [2], while our own day-to-day tools remain basically in a rut. Disappointed because it seems LightTable is foregoing an incremental path to reach the goal, choosing the boil-the-ocean approach. We do need more long term start-from-scratch rethink everything kind of projects; like what VPRI's STEPS project aims to achieve [3]. But there is a lot that can be done to improve the programming experience today, and I don't see enough work in this area. IMO, the latest meaningful improvement in software development tooling was Intellij Idea around 2001 (arguably the functional programming renaissance represents another meaningful improvement, but the real breakthroughs there happened in the 70s). LightTable moving to the Blue plane leaves the PinkPlane unattended. [1] http://www.adobe.com/technology/projects/content-aware-fill.html http://www.adobe.com/technology/projects/content-aware-fill.... [2] https://www.youtube.com/watch?v=UccfqwwOCoY https://www.youtube.com/watch?v=UccfqwwOCoY [3] http://www.viewpointsresearch.org/html/writings.php http://www.viewpointsresearch.org/html/writings.php
- jamii 12y ago
  We aim to have some demoable in the next few months and actually usable this year. We're purplish, at worst :)
tim333 12y ago
Regarding Anon the intern wanting to build lunch_app I've been in a similar position learning programming from the basics and wanting build some apps. I'll put in a vote for web2py as being a good tool that gets around a few of the gripes in the article. It's largely one click to install server, editor, db, debugger etc. and was specifically designed for teaching beginners (it has versioning as well using mercurial but I wasn't able to get that going with one click). It also does most stuff you'd need and is all in Python so you can use the libraries and rewrite bits if you can't get the framework to do something. Even so it would take Anon a while to make his app but I think it's some sort of progress. The creator talks about it here youtube.com/watch?v=iMUX9NdN8YE
- jamii 12y ago
  I'll check it out, thanks.
iammyIP 12y ago
Maybe every tool that we invent and use is two-faced: it makes life both easier and harder, and therefore the wish to make programming easy could result in one of the most difficult journeys so far, comparable with inventing an AI indistiguishable from a human. If i imagine a success in 75 years, then i would probably sit infront and talk to a human-like android: "play me some music based on Bach with the voice of Eddie Murphy, mellow, with a hint of tragedy, but not too sad" - then it would in the blink of an eye play me some music and sing along with it. Chances are i would change Eddie Murphy to some other singer that is readily available at the huge human voice archive this machine has access to. But i am bored with archived and old fashioned music, i want something new and exciting. So i try to teach the machine a new voice. As it turns out it is pretty good at mixing characteristics of multiple voices and synthesizing a new one out of it. I let it do some hundred random variations with hints on what parameters i prefer. "that rough whiskey vibe - keep that and mix it with Presleys guttural hiccups"... "no , not like this, listen". I try to sing what i have in mind, but i am no good singer, so the next test comes out worse than before. "go back ... one more". We spend the rest of the day analyzing Presleys archive and filtering out the guttural characteristic that i meant. I heard i am an expert programmer, my grandma called me that, and i sit down with this android for a whole month and explain in all details the music i want to hear, constantly listening and tweaking my instructions in a flawless instant feedback cycle. The most time spent, or what i sometimes feel - wasted, is actually searching for the right words in my own head to describe what i mean. After 2 weeks i found out that we have broken down the characteristic of a singers voice into twohundredandeighty relevant parameters. Relevant not for everyone of course, but for me and this music i am working on. Maybe i am taking this too far. "delete MelodramaticChorusAccentuationTongueModulationMode two to four". I spend the next week with condensing these parameters down to thirtythree. At some days in the next week i am really without any inspiration and just let the android play thousands of random variations based on the current version that i judge on a 1 to 10 basis while doing some gardenwork. Sometimes i would only change a little note, or the single expression of a syllable. "uah insiide meeee... you see the "siide" must sound more desperate because this guy is on the verge of losing his lifelong dream at that moment." - we spend the rest of the day tweaking that "desperate" thing. This android is amazing. After a month and three weeks the music is ready. I save it and do some other things, mostly gardenwork and watching the drones fly at the evening - they really got some nice new formation techniques that i enjoy greatly. The next day i speak with Andreas (another real human) and tell how good the android has learned my personal preferences for singing voices and how fractalising Bachs harmonic structures to the fifth degree is worth it but no further without structural simplification at the base while keeping the dimensionalitys denominator intact (that was really a complicated talk, i cant get into all the details here). We agree to exchange our Androids preference patterns - my bach-presley11786 for Andreas painting-images9904, and our androids get automatically updated. Next day i try it out. This may take a while, Andreas has warned me. But over the next days i can watch the android painting a really amazing photorealistic but nonetheless astonishingly dreamlike image with Andreas pattern, although the process is, as he said quite slow by design, and one image at a size of four square meters takes about 3 weeks to finish since it includes a complicated technique of iterated de- and reconstruction with additional time for the oil paint to dry up inbetween. But the result is definately worth it. After the image is finished, i call Andreas and congratulate him for his exceptional good taste on imagery (afterall he worked two and half a years on it). He also thanks me for the musical pattern and asks if he can use the guttural Elvis thing in one of his next Crazy Donkey Singalong Performance. That asking just being a polite convention, i naturally give him full permission. We agree to hold another meeting to talk in detail about his structural approach to synthesize dreams without crossing over to kitsch in two weeks. The rest of this day i am back in my garden, the drones fly really low this week... maybe a thunderstorm is coming... but this is great! Every human is a programmer now, which essentially means that he tells other beings what to do, in more or less detail, and of course the fact, that these to be told are not humans anymore, but human-like machines. So, if one of these humans does not tell a machine what to do, he usually enjoys not telling other humans what to do (besides the endless debates about how to tell another machine best what to do (and not to forget - the debates about how it could be better to directly think to a machine rather than speak to it and why it has not been implemented yet)).
perrygeo 12y ago
Or the pain we remember with fond memories. Like anything good in life, the struggle to do it well, to fully understand it, to master it, is what makes it worthwhile. It's also what makes programmers effective - we revel in solving complex but tractable problems. There is an inherent tradeoff in creating any system: you can make the system very simple and it becomes too simplistic for the advanced users who want more control and detail. You allow too much control and detail and you alienate users who want simplicity. A programming language/IDE is a system that affords advanced users the ability to solve an immense set of problems. That unrestrained ability comes at the expense of simplicity. The more constrained and specific that ability becomes, the simpler it can be (i.e. Excel is more constrained but far simpler than a Hadoop cluster)

364 points by Morgawr 12y ago

SSLy 12y ago

Edit: I've mismatched deps and imports, indeed that could be useful. Disregard my post. If I select a function from autocomplete, its dependencies should be automatically added to the project without any fuss. For the point, that that post makes, this one is at least available if you use IntelliJ with Java or Scala.

RussianCow 12y ago
IntelliJ does it with Python and other languages as well, although you have to explicitly hit Alt+Enter -> "Import this name".
- jamii 12y ago
  That's true for pulling functions from other installed namespaces, but to my knowledge you can't install a library in the same way (eg search for 'parse html', drag a result into the editor, get Beautiful Soup added to your virtualenv and imported into the repl).
  - RussianCow 12y ago
    Ah, I didn't realize this is what the parent was talking about. That's pretty damn cool.
    - jamii 12y ago
      I don't know if the parent was talking about that, but it's what I was trying to get at in the post.
edanm 12y ago
You know, I've recently started thikning - why do we even need explicit imports? Isn't the compiler/etc smart enough to figure out the right import 99% of the time? Every file in a large project starts with dozens of redundant import lines. Why don't we have imports go the way of the inferred type?
- riquito 12y ago
  What if foo() is present in two different libraries? (is this the 1% you're talking about?) What if tomorrow I add another library which implement his own foo()? (this is trickier) When I read the code, what is foo()? (here the IDE may be smart and show me the right documentation). All that said, I feel that the pythonic "Explicit is better than implicit" is better. It's also true that when I see a file whose first 30 lines are used for including dependencies it doesn't sound right. The problem lies with the fact that we use text to store structured data. If we were to save source code as binaries we could have the source code in a "file|object|item" and the dependencies in another item, linked to the source code (I'm not advocating this, source code in text files has many advantages, but also his drawbacks).
  - jamii 12y ago
    We give each name a unique id under the hood. At edit time the user chooses which one they want from a list. The actual AST only ever refers to ids and the names are effectively just documentation. This has the neat side-effect that ids are globally unique so you can figure out what libraries you need to pull down just by looking up ids in your package repository.
- Cacti 12y ago
  Depends on what you're looking for. The issue isn't really "importing" the code references as much as it is a scoping issue. That is, imports are largely used to define scope for the file and mostly for the programmers benefit (so you're typing, say, Button.DoSomething instead of System.Windows.Controls.Interactive.Button.DoSomething). That and the odd namespace clash that a compiler can't really resolve, but that is more due to implementation and system issues, not programming proper.

ggchappell 12y ago

What a wonderful article. I'm not allowed to forget a lot of this pain. I teach programming, so I see it anew every semester. And of course I still experience much of it in my own work. So, yes, let's deal with these issues better. One little disagreement. In the "What?" and "Why?" sections the writer present some ideas for debuggers. While these are good ideas, I prefer to think of the inadequacies of existing debuggers as motivation for good practices: modularity, loose coupling, unit testing, etc. Certainly, it would be nice (say) to be able to examine the entire state of a running program easily. But I would rather code in such a way that I do not need to. So to those who would write the World's Greatest Debugger, I say, "Good for you." But even better would be to turn your efforts to producing languages, libraries, frameworks, and programming environments that make such a debugger superfluous.

jamii 12y ago
> I would rather program in such a way that I do not need (say) to examine the entire state of my code. But if you could examine the entire state, why wouldn't you? People build correct intuitions by comparing their mental model to reality. If you don't ever show see reality, or only peek at it through sprinkled print statements, then it becomes harder to build an accurate mental model. This applies whether we are talking about a beginner trying to understand how pointers work or an expert grappling with a complex system. Programming is always limited by our ability to comprehend complexity. So enabling simpler systems is a win, sure, but so is getting better at understanding things.
- ggchappell 12y ago
  > But if you could examine the entire state, why wouldn't you? Well, I would, of course. As I said, I think his debugger ideas are good ones. OTOH, there is a limited amount of effort available for improving programming, and I would prefer that such effort be directed toward other ends. Part of the problem is that, in my experience, debuggers tend to encourage sloppy thinking. And the better the debugger is, the sloppier is the result. Using a debugger, together with thinking hard about a problem, can be a powerful combination. But all too often, I see debuggers used as substitutes for thinking hard; the way debuggers are designed seems to encourage this. To put it in the terms you used, constructing a mental model and comparing it with what the debugger shows, is a great idea. But in practice, I'm faced with a choice: work on my mental model now or fire up the debugger now. The latter is less taxing, and is often chosen; this is likely harmful in the long run.

fragmede 12y ago

> I want to just type 'email' and see a list of functions and libraries relating to email. Do people even remember what life was like before Google? I type 'email python' and get back a link to an email module. Am I being closed-minded in thinking that can't get much easier? Yeah, I need to understand a bit as to how email works, smtp and imap/pop and what not, and how to send vs receive email, but some level of understanding is just necessary.

cjslep 12y ago
I have a similar experience with Go in my spare time projects (very refreshing coming from a FORTRAN/C++ day job) which is a boon, but falls short of what the article describes. The way I interpreted it, even having to do the search in the first place is a problem to be solved. Being able to start typing "email" and have a context menu with functions available from a list of third-party libraries relating to email would be the goal. That way when you select "SendEmail()" in your code, it automatically pulls the library, installs the binary where it needs to be located (or pulls source and builds) without interrupting your workflow. With package management systems standardized within a language (such as Go) the problem domain is reduced, but extending the solution across multiple languages sounds like another third-party tool than a language feature.
- _random_ 12y ago
  You mean something like Bing Code Search but with NuGet support? That would be cool. http://techcrunch.com/2014/02/17/microsoft-launches-smart-visual-studio-add-on-for-code-snippet-search http://techcrunch.com/2014/02/17/microsoft-launches-smart-vi... http://visualstudiogallery.msdn.microsoft.com/a1166718-a2d9-4a48-a5fd-504ff4ad1b65 http://visualstudiogallery.msdn.microsoft.com/a1166718-a2d9-...
  - jamii 12y ago
    Yes, exactly like that. I can't wait to see how Bing Code Search evolves as it starts to learn from its users.
- 12y ago

rch 12y ago

I thought this project was about improving the programming experience for programmers, not students (or interns) who will in all likelihood never excell as programmers. Edit: after downvotes I decided to read the whole thing afterall, but remain unswayed. This guy has an important opportunity and I sincerely dislike seeing it squandered.

alecdbrooks 12y ago
I've been programming long enough that I know the magic incantations to get started in a few languages. But I am still in Anon's shoes any time I learn a new language or library. As the article says, "Even for experts, programming is an exploratory process." So, I'd say everything in this article applies to programmers, too. Categorizing people into programmers and non-programmers neglects the people who need programming but not at an expert level. Some interns and students won't excel as programmers, but they still can bring programming into what they do instead, especially if efforts like Light Table succeed. I suppose I shouldn't be surprised by this kind of pushback against reimagining programming — I've seen it as a response to some of Bret Victor's projects. Skepticism is great: Maybe the ideas behind Light Table are wrong or badly implemented. But we gain nothing from glib dismissals that seem to regard any attempt at improving or simplifying programming as beneath them.
- rch 12y ago
  I disagree with language like 'reimagining programming' in this context almost as much as I regret seeing Bret Victor's name being dragged in for no reason -- to the best of my knowledge, Bret Victor didn't design Light Table, and doesn't inform its progress. There is progress to be made, and I'll be excited again when I see some.
  - alecdbrooks 12y ago
    I've seen the same "oh, this is only for beginners" reaction to some of Victor's ideas on making data visible and making libraries "discoverable" that seems needlessly dismissive. That's why I mentioned him. Also, Light Table's initial concept was inspired by his work[0]. I'm not always impressed by Light Table, but I appreciate the work the team has put into the problem and think they have good ideas. [0]: http://www.chris-granger.com/2012/04/12/light-table---a-new-ide-concept/ http://www.chris-granger.com/2012/04/12/light-table---a-new-...

SolarNet 12y ago

Personally, I think this is an extremely important topic, we need to change the way we program. It isn't the 70's anymore, why do we still program like it is. Why does it take teams of developers to create and manage applications which do really simple tasks in the grand scheme of things (and I realize the amount of complexity in building applications is staggering, but large portions of it could be better automated). Where are the auto generated GUIs, where is the ability to ship execution control to arbitrary devices, where is a hypermedia layer with independent view and presentation code? I'm approaching this from a different angle than the light table guys appear to be (I agree with everything they are saying). My angle is an attempt to build a cross platform module system (where platform includes runtime and programming language as well operating system and architecture): https://github.com/OffByOneStudios/massive-dangerzone https://github.com/OffByOneStudios/massive-dangerzone My argument is: before we can build the next generation of useful tools, we need a framework for managing both generations. massive-dangerzone is an attempt at bringing next generation usage (like that described in the article) to existing tools. It's still a big work in progress though, and is barely useful at the moment.

weland 12y ago

I also wish that programming were a lot different today than it was when I started learning it. That being said, a lot of this article's points are things I've heard before. They led to the development of Visual Basic & co., mostly by people who had no contact with the Smalltalk and Lisp environment in the 80s, while people who did were shrugging and throwing tantrums like WHY THE FUCK DIDN'T YOU FUCKING LIKE IT TEN YEARS AGO? IMHO, all these things went down to the bottom of history because things like these: > Anon the intern needs to be able to open up Programmingâ„¢ and click 'New Web Form' are adequate for people who usually don't program, and extremely inadequate for people who usually do. Generally, and for good reasons, programmers will dislike a tool that hides implementation details for ease of operation. Past a certain level of complexity, the time spent manually doing the right cruft becomes significantly smaller than the time spent manually cleaning up after a smart tool. I sympathize with Anon the intern, but perhaps he should rethink his expectations about complexity; if discoverability is a problem, perhaps he could switch to something that's better documented? And at the risk of sounding like an elitist schmuck who rants about how things were back in his day, maybe he ought to start with something other than web programming. The size and complexity of that tech stack is humongous, to the extent that a large proportion of those who use it don't understand it more than two layers of abstraction down. Programs are also hard to pack and the environment that runs them is hard to setup. Because it involves at least two servers, possibly with several add-ons in order to allow the server-side languages to run, learning at least three languages (assuming server-side JS is an option), two of which (HTML and CSS) aren't quite being used for their original purpose. This is a beginner's nightmare and it has exactly nothing to do with the development tools. And then there are things that are far harder to solve than they originally seem: > I want to just type 'email' and see a list of functions and libraries relating to email. Related how :-)? Should MIME-related functions, needed to reason about attachments, also come up here? HTML parsing/converting, in case you need to deal with HTML email? Information cluttering does nothing to alleviate the opposite problem of information breadth: if Anon the intern's problem is he doesn't know how to Google for libraries or how to make efficient use of documentation, an IDE that presents him with a gazillion of possibly related things won't help him. Especially when, like all beginning programmers, one of his main difficulties is correctly defining the problem he's working on which, in turn, makes it likely for the solutions presented by the IDE to be nowhere even close to the one he needs, because the IDE (like Anon himself) thinks Anon is trying to solve another problem. There is, on the other hand, a lot more truth in this: > Tightening the feedback loop between writing code and seeing the results reduces the damage caused by wrong assumptions, lightens the cognitive load of tracking what should be happening and helps build accurate mental models of the system. I do think that the real resolution to this problem is writing simpler programs whose state is easier to track. On the other hand, programming tools today suck considerably at presenting program meaning. Things like evaluating what an expression comprising entirely of constants, or at least evaluating it based on the default values of the variables involved, are well within reach for today's tools, and yet programmers' calculators are still employed because 99% of the available IDEs couldn't evaluate ADDR_MASK & IO_SEGMENT if the life of every kid in Africa depended on it. This is wicked cool: http://repository.cmu.edu/cgi/viewcontent.cgi?article=1165&context=hcii http://repository.cmu.edu/cgi/viewcontent.cgi?article=1165&c... . However, I also find myself thinking that the very fact that we need debuggers that are this smart is proof enough that we don't reason about our programs well enough. Except for the fringe case of having to quickly debug a (possibly horrible) codebase I haven't written, I'd much rather prefer being good enough a programmer to avoid the need for a debugger that can tell me why foo is not 8 despite the fact that I fucking said foo = 8 ten lines above, than being a programmer with good enough tools to help me when I'm stupid.

logicallee 12y ago

I'll delete this if you don't like it, but I think this is why non-engineers like Steve Jobs, Jony Ives, and a person at HP I won't name, were able to make remarkably good consumer products without knowing how they work. It may be better for a non-engineer to design a laptop than for someone who actually knows how it is put together and exactly how it works. Of course, in a narrow sense, this means such a person isn't really designing it at all: the real engineers are, which may cause resentment. There is a very good chance that a non-programmer can design a programming IDE that is two or three orders of magnitude better (by whatever standard) than the status quo. This means such a person can't actually implement any part of it, or even know exactly what it's doing. Quite a surprising conclusion. By the way I have experienced this myself, when designing for a target I didn't know yet: after/while I was learning it, the resulting design iteration process was much worse than when I didn't know the implementation details. It's harder to think from the user's perspective, after you have been forced to think from the implementation's perspective.

jamii 12y ago
I'm starting to agree with this. If I was working on this alone I would have arrived at something just as complicated and baroque as what came before it. It was only when we started putting our prototypes in front of real users that I started to realise just how much of what I do every day is completely unrelated to the problem at hand. Having the that constant feedback is really vital to challenging existing preconceptions about what programming should look like.
dreamfactory2 12y ago
I don't think it's quite ignorance of the implementation domain is the key here, but rather outside-in vs inside-out thinking. Nearly all engineers solve problems from the inside in my experience.
- logicallee 12y ago
  It's not just that though. Something may be impossible to do in the application, except with knowledge of x, but trivial for someone who knows x. If you know x you might totally forget that without this knowledge the task is basically impossible for the user. So if building the application engenders learning x, then you might totally miss this design problem, unless you step away long enough to forget x again yourself - and that may never happen. It is a bit too much to ask for one person to learn all of the details and totally own the implementation space, and yet be able to instantly forget all those details and go back to when they didn't know anything, to look at things from the user's perspective.

unwind 12y ago

I was somewhat surprised, but very happy, to see Verse mentioned in this context (although he did get Eskil's last name wrong, it's "Steenberg"). I co-developed the initial version of Verse with Eskil; I think it was a bit before its time perhaps. It was hard to get real traction for it, but at least the things Eskil has gone on to build on top have gotten some real attention. Great, and very well deserved!

jamii 12y ago
Oh man, I am full of typos today :S Verse was a huge inspiration for me and it's a shame to see it largely ignored. We're exploring similar ideas for getting our language tooling away from plain text and towards structured, networked representations. I would love to hear your thoughts on how to do that, or any advice you have from building Verse - jamie@scattered-thoughts.net EDIT I also thinks Eskil's ideas on 'what we can do' vs 'what we can get done' applies just as well to the rest of the programming industry. We're so focused on scaling up to millions of lines of code that we end up with environments where lunch_app takes a week to build and deploy.

cclogg 12y ago

Oh man, I've definitely felt this pain many times lol: "The samples will be missing lots of implicit information such as how to install the necessary libraries and how to deal with missing dependencies and version conflicts. Transcribing and modifying the examples may lead to bugs that suck up time. It's not terrible, mostly thanks to sites like stackoverflow, but it's still a lot of unnecessary distractions from the task at hand." So many times, the actual programming isn't tough, it's just getting all of the stuff around it setup that is hard.

tluyben2 12y ago
This is what I have mostly with Rails; when I started working with it I was good at Ruby but didn't use Rails and for a beginner it's horrible. There tons of options for everything, everything is badly documented, most things scratch an itch so when you use it, some of the completely logical things you expect are missing. Samples of gem usage randomly skip steps which can be a hair pulling exercise for a beginner. Then when you finished your app and try to update gems a week later, everything breaks completely. And this is not an attempt at trolling; I now use Rails daily and build rather huge projects in it and this still upsets me. When things work and you know how they work, all is easy and great, installing and using. When either or both of those are not true, you are in absolute hell, especially as a beginner. And nothing of that has to do with actual programming and it shouldn't be needed; your environment should solve this for this imho.

dclowd9901 12y ago

Are these guys ever going to actually finish the IDE, or postulate on what it means to program, ad infinitum. I'm glad they're being considerate in their design, but this whole project is starting to reek of over-aggrandized vaporware.

seanmcdirmid 12y ago
When Chris and company announced the light table project, they had very high goals. I knew then that this would take a long time, and some of the goals would even be unachievable without programming model changes. The advantage of being incremental is that you can reliably produce something in a reasonable amount of time. This is not incremental, and it's quite risky; be patient.
woah 12y ago
Are you mad that they're blogging?
jamii 12y ago
https://github.com/LightTable/LightTable/releases https://github.com/LightTable/LightTable/releases There will be another Light Table release in the next few weeks, featuring CodeMirror 4 (better performance, multiple cursors). We've only been working on Aurora since February. Designing a new language takes time. We may show some early demos in the next few months, depending on how the current prototype works out.
- nazgob 12y ago
  I think that some people (including me) are a bit disappointed that some(most?) of your effort goes to Aurora instead of LightTable. Don't get me wrong, Aurora sounds cool but for professional programmers LightTable is much more appealing.
  - jamii 12y ago
    I can sympathise with that. The reason for our decision is that there is a low ceiling on how much we can improve matters with Light Table. The problems we are trying to solve turn out to be pretty baked in to existing languages and toolchains and we can't hack around them in the editor. Aurora is a riskier project but it has much more potential to help people. Also, the fact that Aurora is intended to be simple and easy does not mean that we are building a toy. Quite the opposite - the intention is to remove some of the incidental complexity in programming to leave you free to focus on the actual problems. We intend to bootstrap the compiler and editor by way of dogfooding. I look forward to doing most of my programming in Aurora in the future.

politician 12y ago

"The samples will be missing lots of implicit information such as how to install the necessary libraries and how to deal with missing dependencies and version conflicts. " Docker pretty much solves this problem. Writing samples for your new project? Start with step 1 `docker pull yourproj/your_tutorial_image`.

gedrap 12y ago

I totally agree. It's true even for us, experienced Web guys (and gals), in this mess ruled by packages and their managers. For example, I am trying to make my first Rails app outside a tutorial. Okay, I want to use bootstrap. Should I use gem? Guess that's the ruby way. Okay seems like there a few of them. Tried one, another one. Doesn't work. Don't know why, since I am just copying half-cryptic stuff because I am new to Rails. A friend suggests to use Bower. It's easier. Right! I had totally forgotten about bower, it rocks! Google: bower rails. Okay, there is that thing sprockets which I apparently need to configure. Googled, a few blog posts opened, they offer a bit different advice each, let's try. Doesn't work. Nope. Not really... Google: bower rails bootstrap sprockets. And yay! Works. Two hours later, I have included Bootstrap. Properly.

joevandyk 12y ago
Download bootstrap css, put into app/assets/bootstrap. Should be all that's needed.
- gedrap 12y ago
  Yeah, it would have worked. But I am too lazy (among real disadvantages) to download all the libraries manually :) Real disadvantages such as updating becomes so boring that you will probably never do that. And 'bower search <something>' followed by 'bower install <found it!>' is less distracting than googling and unzipping etc. Being lazy pays off.
  - Cacti 12y ago
    If you're a programmer and you can't even bother to download some files and unzip them, then... Lazy pays off because you innovate in the hopes of automating something, not that you just are actually completely lazy.
- igravious 12y ago
  And repeat for every update of Bootstrap? Why?

naturalethic 12y ago

Pretty sweet article. Needs a once over from the author for some edits though. "if you don't have something useful to say, don't say nothing at all" made me giggle.

bridger 12y ago
I can't figure out what that one is supposed to mean!

Goopplesoft 12y ago

> We still program like it's 1960 because there are powerful path dependencies that incentivise pretending your space age computing machine is actually an 80 character tty. We are trapped in a local maximum. Love this (although I don't like extra long lines). He's talking to you Pep8.

dreamfactory2 12y ago

Great article but isn't a lot of this more about using different levels of abstraction i.e. frameworks, DSLs and specific to those? That could either take the form of being sufficiently focused that a simple text editor is adequate (e.g. high level commands, no cruft), or a full-featured managed environment (state, docs etc) specific to the framework. I'm not seeing how this is a task for a generic IDE unless I missed something about light table and they are creating a new language.

jamii 12y ago
> unless I missed something about light table and they are creating a new language. http://www.lighttable.com/2014/03/27/toward-a-better-programming/ http://www.lighttable.com/2014/03/27/toward-a-better-program... The short version is yes, we are making a new language. The kind of workflow I discussed in the OP was the original goal for Light Table and it turns out that it can't just be hacked onto existing languages.
- rwallace 12y ago
  Fair enough. I won't say your decision to make a new language is right or wrong - I don't know how it will turn out - but given that you have decided to do so, and given your objectives for it, my one big recommendation would be to make relational databases central to the language, the way Perl makes string processing central and PHP makes web pages central. Some possible ideas are outlined at http://geocities.com/tablizer/top.htm http://geocities.com/tablizer/top.htm - you don't have to go with those particular ideas of course, but I think it's important to make relational data central one way or the other.
  - jamii 12y ago
    We are heavily inspired by http://shaffner.us/cs/papers/tarpit.pdf http://shaffner.us/cs/papers/tarpit.pdf

mkozlows 12y ago

Articles like this one annoy me, because it's easy to diagnose big problems -- here, watch me do it: "Why should compilers choke if you forget a semicolon? If they can diagnose a syntax error, can't they also fix it? Can't functions be smarter about seeing when they're used improperly and tell you while you're writing the code instead of when you're running it? Why can't the code be understandable to anyone who knows English?" What's hard -- and often impossible -- is fixing those big problems, because a lot of times they're genuinely intractable; and when they're not, they're often so difficult that they might as well be. So just sitting around and complaining about them sounds insightful, but it doesn't really get anything done. And yeah, I know that they're allegedly working on "fixes" for these issues, but based on the track record so far (LightTable promised all sorts of revolutionary views of what an IDE could be; it's delivered... not a whole lot of revolution), I don't have any faith that Aurora is going to amount to much either. And I don't want to be too negative, because sometimes a previously-intractable problem turns out to now be tractable, and it takes someone who was willing to question long-accepted pain to find that out. So I'd be pleasantly surprised if this project delivered something that had even as much effect as the development of, say, git or xunit. But I'm not holding my breath.

lemming 12y ago

I'm glad to see IntelliJ getting some love there for making much of this easier - few companies have worked so hard at it, and the results are amazing. Looks like Microsoft is making some great steps too, Bing Code Search gets a bad rap for pandering to blub programmers but I'd love something like that in my editor.

ilaksh 12y ago

Can't be sure but it seems like the surveying has resulted in most of the right conclusions and the right direction for the project. HOWEVER, I still think most everyone is missing the REAL issue here. As evidenced by the powerful tools cited in this article that address various aspects of the "programming" problem, there have been numerous efforts to move the state-of-the-art in software development forward. And to a great degree those efforts have _proven_ quite a few superior paradigms. And yet, we haven't seen those new paradigms become truly mainstream for most programmers. Why? I do NOT believe it is because the approaches haven't integrated the right new concepts, because there are so many existing useful tools combining many different new ideas effectively which have failed to become mainstream among programmers. I think many of those new approaches could and should have become normal operating mode for programmers. I think the reason they did not is this: the core definition of "programming" (and by extension "software engineering" etc.) is an antagonistic and largely manual process of creating complex textual source code that can be used to describe system behavior. Period. For example, if I want to call myself a web "developer" I DARE NOT use an interactive GUI tool to generate and maintain the source code for my web site. Web developers will always say, for example, they avoid this because the source code generated that way is less maintainable by hand. In many cases that may be true. In some cases with advanced code generation it is not. Regardless, I don't believe that is the actual reason. Its just a rationalization. What do we call someone who's entire job entails creating a web page using a graphical user interface? In other words, this is a hypothetical person who has found a hypothetical GUI tool that can accommodate all of his web site design and implementation needs without any manual edits to source code. If he builds a web site or web application this way, and writes zero lines of code, do we refer to him as a very smart and advanced web "developer"? Or do we call him a web designer or simply a WordPress user (for example)? We do NOT refer to him as a web "developer". He has no right to refer to himself as a "developer" or "programmer" because he has not wrestled through an antagonistic manual process to create complex textual source code. And that's what people are not understanding. The reason we don't call him a programmer is NOT because he didn't create an effective program or website. Its because the way he did it wasn't hard enough and doesn't match our outdated definition of what "programming" is. To create a POPULAR system (among "real" "programmers") that makes programming more practical or easier using new paradigms, you must either redefine programming to include the possibility of new paradigms and a non-antagonistic process, or perhaps somehow trick programmers into thinking what they are doing is actually harder than it is. Maybe if there are a few places to type a command to generate source code, that will be sufficiently complex to still be considered "programming". If you are too successful without doing those things then you will just have another tool that only "users" or "beginners" would ever admit to using.

zem 12y ago

solving this problem was part of REBOL's dream.i really wish it had been properly open-sourced from day one and built up a good community and momentum around it; it was a very promising language.

shortstuffsushi 12y ago

The analogy to me seems similar to that of programmers and compilers. I want to write programs to do xyz, but I don't particularly care to know the lower-level to machine level implementation. Instead, I know a roughly-english set of instructions to tell the compiler what I want to do, and it takes care of the rest for me. Could it not be possible to take this even further, to the point where I say "I want to do ..." and all of the "code," be it high or low level is generated for me. I realize making something capable of this level of abstraction would be incredibly difficult, but it would certainly be fun to try.

GFischer 12y ago
There are companies trying. In Uruguay, a tool called Genexus enjoys a high level of success promising something like that :) . For simple CRUD apps, it's pretty much magical - and it looks awesome in demos, heck, it can generate code for every mobile phone out there just throwing around some controls, specifying your data fields and writing a few rules, but if you want to make it do something that isn't explicitly considered, it's very annoying (and it has some UI and UX problems). And don't ever try to touch the autogenerated code :P . That said, it's probably the easiest way to get from 0 to a complete CRUD app, once you learn how it works. http://www.genexus.com/products/genexus?en http://www.genexus.com/products/genexus?en

zan2434 12y ago

Saved. Can't wait to look back at this in a few years.

webmaven 12y ago

While I am watching the development of LightTable with interest, it seems to me that the example used in the OP is missing the forest for the trees. It isn't hard to imagine Anon Intern managing to cobble together a solution for lunch_app that isn't an 'app' at all, just a form created with a builder like WuFoo or Google Forms, and some 'recipes' in a tool like IFTTT that has built in integration points for email, the accounting system, etc. Arguably this wouldn't much of an improvement over the VB6 + Access status quo (or Visual FoxPro, or FileMaker Pro, etc), except that the individual components can be much more robust and scalable, monitored, auditable, and so on, without Anon Intern having to worry about any of that.

jamii 12y ago
I mentioned that near the beginning: > There are probably dedicated apps that cover this particular example but we are more concerned with how an end-user would solve this kind of problem in general... VB6, Access and Excel are exactly the kind of tools we seek to emulate on the usability front. Half of the world runs on Excel precisely because it is easy to get started, easy to understand and yet is capable of handling fairly complex simulations. If we can make something equally approachable but backed by a sane, maintainable and powerful language then we can give programming powers to a lot of people who need them.
- webmaven 12y ago
  Except that I am not talking about a dedicated app at all, but general purpose tools/services for parts of the problem that can be combined arbitrarily, just like libraries can. IFTTT is exactly that sort of service.

rafaelferreira 12y ago

Not often I can read an essay that leaves me at the same time so excited and so disappointed. Excited because I found myself agreeing with every point: isn't it obvious to everyone that the programming experience could be so much better? We created wonderful tools for graphical expression and number crunching, that keep getting better [1] [2], while our own day-to-day tools remain basically in a rut. Disappointed because it seems LightTable is foregoing an incremental path to reach the goal, choosing the boil-the-ocean approach. We do need more long term start-from-scratch rethink everything kind of projects; like what VPRI's STEPS project aims to achieve [3]. But there is a lot that can be done to improve the programming experience today, and I don't see enough work in this area. IMO, the latest meaningful improvement in software development tooling was Intellij Idea around 2001 (arguably the functional programming renaissance represents another meaningful improvement, but the real breakthroughs there happened in the 70s). LightTable moving to the Blue plane leaves the PinkPlane unattended. [1] http://www.adobe.com/technology/projects/content-aware-fill.html http://www.adobe.com/technology/projects/content-aware-fill.... [2] https://www.youtube.com/watch?v=UccfqwwOCoY https://www.youtube.com/watch?v=UccfqwwOCoY [3] http://www.viewpointsresearch.org/html/writings.php http://www.viewpointsresearch.org/html/writings.php

jamii 12y ago
We aim to have some demoable in the next few months and actually usable this year. We're purplish, at worst :)

tim333 12y ago

Regarding Anon the intern wanting to build lunch_app I've been in a similar position learning programming from the basics and wanting build some apps. I'll put in a vote for web2py as being a good tool that gets around a few of the gripes in the article. It's largely one click to install server, editor, db, debugger etc. and was specifically designed for teaching beginners (it has versioning as well using mercurial but I wasn't able to get that going with one click). It also does most stuff you'd need and is all in Python so you can use the libraries and rewrite bits if you can't get the framework to do something. Even so it would take Anon a while to make his app but I think it's some sort of progress. The creator talks about it here youtube.com/watch?v=iMUX9NdN8YE

jamii 12y ago
I'll check it out, thanks.

iammyIP 12y ago

Maybe every tool that we invent and use is two-faced: it makes life both easier and harder, and therefore the wish to make programming easy could result in one of the most difficult journeys so far, comparable with inventing an AI indistiguishable from a human. If i imagine a success in 75 years, then i would probably sit infront and talk to a human-like android: "play me some music based on Bach with the voice of Eddie Murphy, mellow, with a hint of tragedy, but not too sad" - then it would in the blink of an eye play me some music and sing along with it. Chances are i would change Eddie Murphy to some other singer that is readily available at the huge human voice archive this machine has access to. But i am bored with archived and old fashioned music, i want something new and exciting. So i try to teach the machine a new voice. As it turns out it is pretty good at mixing characteristics of multiple voices and synthesizing a new one out of it. I let it do some hundred random variations with hints on what parameters i prefer. "that rough whiskey vibe - keep that and mix it with Presleys guttural hiccups"... "no , not like this, listen". I try to sing what i have in mind, but i am no good singer, so the next test comes out worse than before. "go back ... one more". We spend the rest of the day analyzing Presleys archive and filtering out the guttural characteristic that i meant. I heard i am an expert programmer, my grandma called me that, and i sit down with this android for a whole month and explain in all details the music i want to hear, constantly listening and tweaking my instructions in a flawless instant feedback cycle. The most time spent, or what i sometimes feel - wasted, is actually searching for the right words in my own head to describe what i mean. After 2 weeks i found out that we have broken down the characteristic of a singers voice into twohundredandeighty relevant parameters. Relevant not for everyone of course, but for me and this music i am working on. Maybe i am taking this too far. "delete MelodramaticChorusAccentuationTongueModulationMode two to four". I spend the next week with condensing these parameters down to thirtythree. At some days in the next week i am really without any inspiration and just let the android play thousands of random variations based on the current version that i judge on a 1 to 10 basis while doing some gardenwork. Sometimes i would only change a little note, or the single expression of a syllable. "uah insiide meeee... you see the "siide" must sound more desperate because this guy is on the verge of losing his lifelong dream at that moment." - we spend the rest of the day tweaking that "desperate" thing. This android is amazing. After a month and three weeks the music is ready. I save it and do some other things, mostly gardenwork and watching the drones fly at the evening - they really got some nice new formation techniques that i enjoy greatly. The next day i speak with Andreas (another real human) and tell how good the android has learned my personal preferences for singing voices and how fractalising Bachs harmonic structures to the fifth degree is worth it but no further without structural simplification at the base while keeping the dimensionalitys denominator intact (that was really a complicated talk, i cant get into all the details here). We agree to exchange our Androids preference patterns - my bach-presley11786 for Andreas painting-images9904, and our androids get automatically updated. Next day i try it out. This may take a while, Andreas has warned me. But over the next days i can watch the android painting a really amazing photorealistic but nonetheless astonishingly dreamlike image with Andreas pattern, although the process is, as he said quite slow by design, and one image at a size of four square meters takes about 3 weeks to finish since it includes a complicated technique of iterated de- and reconstruction with additional time for the oil paint to dry up inbetween. But the result is definately worth it. After the image is finished, i call Andreas and congratulate him for his exceptional good taste on imagery (afterall he worked two and half a years on it). He also thanks me for the musical pattern and asks if he can use the guttural Elvis thing in one of his next Crazy Donkey Singalong Performance. That asking just being a polite convention, i naturally give him full permission. We agree to hold another meeting to talk in detail about his structural approach to synthesize dreams without crossing over to kitsch in two weeks. The rest of this day i am back in my garden, the drones fly really low this week... maybe a thunderstorm is coming... but this is great! Every human is a programmer now, which essentially means that he tells other beings what to do, in more or less detail, and of course the fact, that these to be told are not humans anymore, but human-like machines. So, if one of these humans does not tell a machine what to do, he usually enjoys not telling other humans what to do (besides the endless debates about how to tell another machine best what to do (and not to forget - the debates about how it could be better to directly think to a machine rather than speak to it and why it has not been implemented yet)).

perrygeo 12y ago

Or the pain we remember with fond memories. Like anything good in life, the struggle to do it well, to fully understand it, to master it, is what makes it worthwhile. It's also what makes programmers effective - we revel in solving complex but tractable problems. There is an inherent tradeoff in creating any system: you can make the system very simple and it becomes too simplistic for the advanced users who want more control and detail. You allow too much control and detail and you alienate users who want simplicity. A programming language/IDE is a system that affords advanced users the ability to solve an immense set of problems. That unrestrained ability comes at the expense of simplicity. The more constrained and specific that ability becomes, the simpler it can be (i.e. Excel is more constrained but far simpler than a Hadoop cluster)