Is the rails 2.0 scaffold system philosophically ( not technically? ) broken?

rubyonrails-talk
1

Hello,

I've read the (many) re-posts about problems around scaffolding in Rails 2.0 and have followed a number of tutorials and fully understand "how to scaffold" from a technical perspective, but I don't understand the *mindset* of how to use the new scaffolding. It seems like a productivity- / agility- regress and I'm thinking I may have failed to properly grok the new setup. In the interest of full disclosure, I'm coming back to Rails after being in other toolkits for about 9 months.

Thanks to the intrepid work of Sean Lynch at (

) I found a tutorial that would familiarize me with the raw "how to scaffold" material.

I followed his tutorial's step of:

``ruby script/generate scaffold Movie''

Great! From that point I filled in the "columns" in the migration as I had done in Rails 1.x. All I should need to do is run ``rake db:migrate'' and try adding a new record via the dynamically-created view.

When I started the server and navigated localhost:3000/movies I had the "create new" button. When I pushed that button there were no text widgets to enter *despite having defined the columns that corresponded to said widgets* having been added to the migration ( I have a lengthy blog post about how my diagnostics went, for anyone else's edification at http://stevengharms.net/?p=1063 ). In short the scaffold that had been created knew nothing of the columns I had added in the migration and, as such, the 'new' view had no widgets.

This struck me as well, wrong. On Sean's post another user confirms the same experience. I have tried it with sqlite3 / mysql / postgres connectors.

Research showed that the scaffold had remained static relative to the time that I had done the original aenemic invocation. Per ``script/ generate scaffold --help'':

./script/generate scaffold post` # no attributes, view will be anemic

To fix this I had to re-issue the script/generate command with all the attributes in "final draft" mode ( ``script/generate scaffold movie title:string text:description one_sheet_url:string'' ) and then over- write the old templates ( output stored below, for legibility, Fig. 1).

The solution implies: - You have to get the script/generate command's "attributes" arguments *perfect* at time of creation OR - You do this overwriting thing that I describe below.

As I recall Rails 1.x's dynamic scaffolding allowed us to use a scaffold flexibly strictly based on migrations and rake db:migrate. This flexibility allowed us to "sketch" ideas very rapidly. Or is it considered a "Good Thing" that you get a "perfected" ``generate scaffold'' command at some point? If so, what's the reasoning? Am I missing some sort of rake command that "refreshes" the scaffold templates?

Based on the comments at Sean's site and some of the questions in the comments to DHH's Rails 2. announcement I think there are others grappling with this quandry as well. Can anyone help?

Steven

==Fig. 1== bash-3.2$ script/generate scaffold movie title:string text:description one_sheet_url:string       exists app/models/       exists app/controllers/       exists app/helpers/       exists app/views/movies       exists app/views/layouts/       exists test/functional/       exists test/unit/ overwrite app/views/movies/index.html.erb? (enter "h" for help) [Ynaqdh] y        force app/views/movies/index.html.erb overwrite app/views/movies/show.html.erb? (enter "h" for help) [Ynaqdh] y        force app/views/movies/show.html.erb overwrite app/views/movies/new.html.erb? (enter "h" for help) [Ynaqdh] y        force app/views/movies/new.html.erb overwrite app/views/movies/edit.html.erb? (enter "h" for help) [Ynaqdh] y        force app/views/movies/edit.html.erb    identical app/views/layouts/movies.html.erb    identical public/stylesheets/scaffold.css   dependency model       exists app/models/       exists test/unit/       exists test/fixtures/    identical app/models/movie.rb    identical test/unit/movie_test.rb         skip test/fixtures/movies.yml       exists db/migrate Another migration is already named create_movies: db/migrate/ 001_create_movies.rb

2

Scaffolds != Rails

They're a starting point, and as such just give you something to start with. Scaffolds aren't meant to be your whole application, so the code is treated just like code that's written independent of them: If your object model changes, then you need to change your views and controller logic to match.

--Jeremy

3

So your take could be summarized as: "The scaffold system is as good as it needs to be because you're only going to use it really briefly before you start fleshing out proper views".

4

Jeremy McAnally wrote:

Scaffolds != Rails

No, but the simplicity of the Rails 1.x scaffold is what 'sold' Rails to a lot of people (e.g. via the famous DHH Blog app video on the Rails site). Personally, I think it would have been nice to have kept the 'backwards compatibility' intact so that newcomers would have ready access to all the Rails 1.0 tutorials available rather than trying to follow those tutorials and immediately running up against the buffers, so to speak... :wink:

best wishes Huw

http://www.sapphiresteel.com Ruby In Steel for Visual Studio

5

I believe you can still install it from a plugin if you're really interested in it.

I think dynamic scaffolding was a crutch that kept people from really getting Rails from the start (i.e., you didn't have to build views so they were missing out on that). I think making them generate the code gets them elbow deep in the sort of stuff they'll be writing quicker.

--Jeremy

6

Breaking backward compatibility is a luxury that only open-source developers can afford. It costs nothing to lose customers if they aren't paying. If you need to maintain your customer base (like, for example, Microsoft does) then you do anything to avoid breaking backward compatibility.

See for example: http://blogs.msdn.com/oldnewthing/archive/2006/11/06/999999.aspx

I personally was quite shocked to see that Rails 2 knowingly broke things. Extracting to a plugin I can deal with. Outright removal is shocking.

fredistic

7

Even though this is a bit off-topic, Rails didn't arbitrarily break things. Developers who use Rails use the code: it's exposed, we manipulate it, and it's what we use in our applications. Therefore, it's in the best interest of everyone involved if Rails cuts out the cruft while pushing towards better solutions. If someone wants to keep using old feature, they're welcome to keep using the version of Rails they're using.

I would hate to end up with a 35MB framework that could easily be 2MB or less but has kept so much stuff around in the interest of backwards compatibility.

--Jeremy

8

Um, this looks interesting: http://wiki.rubyonrails.org/rails/pages/Scaffolding+Extensions+Plugin

f

9

Fortunately, open-source projects are not run by money. They respect their users, and that's why there's a cycle of deprecation/removal going on. Warnings about deprecated stuff all over the place, documentation, etc.

A major release is allowed to break things, that's what the 2.0 signals. You can put the version of Rails your application is known to run OK under vendor/rails, or revise and upgrade.

To polish and continue improving something you need to add, but you need to cut as well. A major release allows cutting.

-- fxn

10

And there's no one holding a gun to anyone's head forcing them to use the new versions. You can wait until you are ready, the old versions are still there.

Applications using frameworks like Rails are tied to the implementation of the version of the framework they use. This is okay as long as you can control if and when you move to a new version. Some years ago, there was a lot of interest in the idea of making framework-based operating systems. Here's a war story from those days: http://talklikeaduck.denhaven2.com/articles/2007/06/15/a-meeting-with-gill-bates

11

Breaking backward compatibility is a luxury that only open-source developers can afford.

Backwards compatibility is frequently very expensive. Microsoft in particular expends vast amounts of resources on backwards compatibility, and quite a bit of that effort is almost entirely useless to the vast majority of their customers.

Would you rather have those engineers working on new/improved functionality, or on bug-for-bug compatibility that's only interesting to a tiny minority of users?

Think of backwards compatibility as a tax that older users impose on newer users. That may be worth paying; newer users may themselves want backwards compatibility in the future.

But the community may also decide that tax isn't worth paying. Older users may be required to spend resources to use newer versions of the system in question. That's OK; they're getting the benefits of development resources applied to the newer versions too.

If you need to maintain your customer base (like, for example, Microsoft does) then you do anything to avoid breaking backward compatibility.

Not at all. Older users just have to spend some resources making sure they're good on the newer system. It's one of those engineering/business decisions that people make every day.

12

Philosophical arguments aside, if they were going to take it out, they should have, well, just taken the whole thing out. I should just get an error when I try to create a scaffold if it isn't going to get made properly.

All this about backward compatibility is all fine and good; but at the very least i should get a deprec message instead of having to hunt through the erb files in the vestigial scaffold remnants that don't work, re-rake, see that nothing changed, question my own sanity, then do a Google search and come here. It doesn't make sense.

13

All this about backward compatibility is all fine and good; but at the very least i should get a deprec message instead of having to hunt through the erb files in the vestigial scaffold remnants that don't work, re-rake, see that nothing changed, question my own sanity, then do a Google search and come here. It doesn't make sense.

I suggest to delegate the scaffolding to a plugin which specializes with this task, for instance http://activescaffold.com/.

(yeah that doesn't answer the philosophical question for sure!)

best wishes !

Thibaut Barrère / LoGeek

14

So, Many older tutorials (and books) suggested a method of working models and relations like: generate model, scaffold, migrate, check, migrate, check... (e.g. the original blog-video and the depot tutorial)

That is all deprecated and replaced by... what? How will the rewrites look? What is the new preferred way of working?

I was quite comfortable with this way of working myself and I haven't really found something to fill the void yet. I don't want to sound critical. I would just love to get the scoop on what has made scaffolding more or less obsolete in the eyes of the core team.

My current guess is that I should start using the console more when being interactive with models.

cheers. Martin Westn

15

Dynamic scaffolds give you squat. You got one line of code doing some magical things so your browser renders some magical other things - don't tell me you can learn the framework by staring at that line long enough until it conveys meaning. It won't.

There's a place for it of course: marketing material. Shiny "oh look this is so great" screencasts, which imply that your next big Web2.0 Buzzword-Compliant Social Networking app is just ten minutes away.

What you can do is script/generate scaffold Foo bar:text - that'll give you stuff to look at, and if you don't like what you see you actually have the chance to change stuff. Also, it does what the name implies ("scaffold", remember?), which is A Good Thing.

In other news, the "preferred way of working" is still, after all those years, to actually writing code while knowing wtf is going on.

Oh, and another reason: everytime someone writes scaffold into their text editor or irc client, god kills a kitten. true story.

Martin wrote:

16

Sooooo....I'm guessing there's no more code generation huh?

These sorts of decisions usually make tons of sense to core developers on a framework. However, to the masses, it may be a quite different.

Here is a newbie's perspective: I've been into web frameworks for about a year now. And the "magic" of code generation is what drew me in. CakePHP has a "bake" feature, which doesn't do a 1/4 of the stuff I've seen RoR do in the 12 hours I've been messing with it. I stayed a way from RoR this long for the simple fact that I didn't see the need tl learn Ruby, due to it's "interesting" syntax.

Recently I saw a screen cast of the RadRails plugin and I was sold on RoR. Along with the fact that it's more established and has a larger following. From all the tutorials I've been reading the code generation was light years ahead of CakePHP. So I'm thinking, I'll install this guy, install that cool IDE, and get going with the Agive Web Dev. Book and a few tutorials. I figured the quickest way to get into RoR is to port some of my CakePHP apps: import that databases (schema conventions seem to mimic RoR), slap on some scaffolding to generate my MVC's, then peer into the code. There are sooooo many plugins there, I couldn't wait to get started on more complicated things like AJAX, Auth, etc.

12 hours later, I'm still trying to scaffold that stupid cookbook2 with two tables in it. I thought I was doing something wrong. I figured, it already got the schema of the database, how hard could it be to build the stupid models, controllers and views. To my great dismay, I've come to learn they have been removed? Wow.

Sorry for the rant, but I don't really understand. What's the point of all the visual database design tools, if at the end of the day, I still gotta write everything in the migration syntax? I see where all the migration stuff can come in handy for "revisions", but when starting off, I don't get it. Yes, scaffolding will give you a lot more stuff than you need. But, IMO, it's much easier to sit on the delete key for a while, than it is to go write code that you don't understand.

Can someone please tell me that I'm mistaken. And that there HAS to be a way around this? How do I get tons of SQL into an "initial" migration, so that I can generate MVC's from there?

ThanX in advance.

17

I guess the Rails core team has become a little out of touch with the newbie developer. But, frankly, that's a necessary step for the maturation of a project like this. 3 years ago, Rails needed all the mindshare it could get, and the infamous screencast was a powerful hook to draw people in.

These days things are different. The core team is trying to Rails as powerful as possible for it's large userbase without bloat. So far the vision has been maintained extremely well. Maybe the decisions with scaffolding don't sit well with everyone. The problem is when debate happens in the core community, scaffolding is the last thing on everyone's list to worry about.

I really sympathize with the effect of undocumented changes and out of date tutorials--I've been burned plenty. But scaffolding really is such an insignificant part of Rails that its flaws should have no bearing on your decision of whether or not to use Rails. Rails isn't a visual toolkit, it's a serious development framework. You just have to make it over two humps in the learning curve: the Rails API hump and then later the Dynamic Ruby hump and you'll be golden.

18

Hi Everyone,

I've read each post in this thread & I cannot believe the misunderstanding of 'scaffold' in rails 2.0.2

As Johannes puts it so well...

Quote

Dynamic scaffolds give you squat. You got one line of code doing some magical things so your browser renders some magical other things - don't tell me you can learn the framework by staring at that line long enough until it conveys meaning. It won't. <<<<<<<< End Quote

Can I suggest those who are complaining about this change follow "Akita's" tutorial on rails 2, which should enlighten you to what the change is all about in a real context !

HTH - Dave Porter

19

Maybe the use of the term scaffolding is wrong. I meant the code generation from the migration.

Everyone here is always going to say that it’s not hard to write your own controllers. I’m not saying it’s hard. I’m saying, as a newbie developer in the framework, it’s a great helping tool. For me, web development is about getting stuff done.

But seriously, how does one “import” and already existing database design into a Rails Project? I’m talking about 100+ tables, with (on average) 20+ fields.

Does one need to specify script/generate model (and all 20 fields)?

20

This is the first time I have been unpleasantly surprised by a comment on this forum. I wonder if this reflects the views of the core team.

Working on creating 2 large apps based on rails 1, I am in the process of moving them to Rails 2. But I sense a significant distance between the core Rails Movers and "the Rest of us". So much of the documentation (which was in any case not ideal) has now been confused, and left behind. As an existing Rails developer, I can just about cope with the transition of how to convert to plugins for things like pagination, auto_complete etc. (But for a good summary of the changes I was forced to buy the peepcode pdf) I dont mind paying , but I would if I was just trying rails out for the first time . So many of the tutorials are now going to just not work for Rails 2 and are going to point in the wrong direction, particularly as regards REST (and consequently the scaffold) and routing which is getting a fairly high profile. Even the Agile Rails book is no longer being quoted as the Primary resource. I suspect if I were coming as a noob, then this amount of confusion, on top of the lack of a single documentation resource, would cause me to give up in disgust.

Now I do not want to be a complainer. I do not have any problems with the Rails 2 direction, and getting to grips with REST routing and the various concepts is really good for me. But if there is no leg up for those entering the fray for the first time,and it is felt that maturing the product necessitates losing touch with the noob, there is a danger that Rails could become an old boy's club. (Purely my own opinion, but that would be a real shame - I have been a strong advocate of Rails.)

Being a transition from 1 to 2, means this is Rails first experience in version transition. The rails team has got so much right that it would be a shame to dismiss the significance of the version transition. I think there is merit in the view that need for total backward compatibility can be dispensed with. But if that is done without clear documentation for the transition and its implications (prominantly on the Rails home site - cos that is the logical first place to look), then it is just going to leave confusion to eventually do its worst.

comments made with good intentions - reflecting my own feelings along the way. Tonypm