Home » Modeling » TMF (Xtext) » Please lower the barrier
| Please lower the barrier [message #36946] |
Sun, 05 April 2009 14:29  |
Eclipse User |
|
|
|
I've been spending some time this weekend with XText, both in the oAw and
TMF flavours, reading the forums and trying to get some demos up and
running.
This has been a rather frustrating experience, and I'm still not sure
whether XText is just another poorly documented project that creates more
problems than it solves, or a valuable addition to any developer's toolset.
If you want to reach a wider audience, please consider the following
issues:
- Provide a one-stop install.
- Prove by example that TMF can be applied to non-trivial languages, e.g.
Java.
- Provide some decent documentation, explaining how to add behaviour to
generated models and how to apply TMF to existing DSLs.
- Get out of that limbo between oAw and Eclipse. This is utterly
confusing, and given the choice between a stable version which is poorly
documented and lacking features and a beta version which has some more
features and documentation in some areas, but large gaps in others, the
only safe bet for potential users is to come back in a year's time.
Anyway, I'll stop moaning now and create separate threads for some of the
problems I've faced.
Regards,
Harald
|
|
|
| Re: Please lower the barrier [message #37050 is a reply to message #36946] |
Sun, 05 April 2009 17:29  |
| Eclipse User |
|
|
|
Hi Harald,
thanks for your feedback. Please find comments below.
Harald Wellmann schrieb:
> I've been spending some time this weekend with XText, both in the oAw
> and TMF flavours, reading the forums and trying to get some demos up and
> running.
>
> This has been a rather frustrating experience, and I'm still not sure
> whether XText is just another poorly documented project that creates
> more problems than it solves, or a valuable addition to any developer's
> toolset.
We are aware of the fact, that we have to improve the state of the
documentation. What have you missed at most?
>
> If you want to reach a wider audience, please consider the following
> issues:
>
> - Provide a one-stop install.
Please have a look at
http://oaw.itemis.com/openarchitectureware/language=en/660/d ownloads
I think we may not put a link to the download page in the wiki and I'm
aware of the fact, that the workspace setup is a little cumbersome (and
even outdated as it contains superflous steps).
>
> - Prove by example that TMF can be applied to non-trivial languages,
> e.g. Java.
I think, that a dsl should not be as complicated as Java itself, as it
would not provide any meaningful abstraction. We are favouring another
approach, that is to integrate better with the JDT and provide
cross-language services such as code completion of java types etc. But
you are right, we should come up with some serious examples. Currently
we are very busy and the focus is on other tasks, such as scalability,
code completion, code formatting and stuff.
>
> - Provide some decent documentation, explaining how to add behaviour to
> generated models and how to apply TMF to existing DSLs.
As we don't do something special with EMF models, we will most likely
not bind resources with writing documentation on how to work with EMF
model instances. We try to integrate as seemlessly as possible with the
concepts behind EMF and the currently rapidly evolving ecosystem.
You can add behaviour to your model in the EMF way, meaning don't use
generated metamodels, but import them and use JMerge to add handwritten
code to your model implementation. Alternativly you could use the
external visitor to achieve what you need. Again, both are core EMF
ideas and have nothing special to do with Xtext.
>
> - Get out of that limbo between oAw and Eclipse. This is utterly
> confusing, and given the choice between a stable version which is poorly
> documented and lacking features and a beta version which has some more
> features and documentation in some areas, but large gaps in others, the
> only safe bet for potential users is to come back in a year's time.
Please be more concrete. What do you mean with large gaps? Are you
refering to your other posts, or is there something else?
>
> Anyway, I'll stop moaning now and create separate threads for some of
> the problems I've faced.
>
> Regards,
>
> Harald
>
Thanks again, for your detailed feedback,
Sebastian
>
>
>
>
|
|
|
| Re: Please lower the barrier [message #37152 is a reply to message #37050] |
Sun, 05 April 2009 18:58 |
| Eclipse User |
|
|
|
Hi Sebastian,
>> - Prove by example that TMF can be applied to non-trivial languages,
>> e.g. Java.
>
> I think, that a dsl should not be as complicated as Java itself, as it
> would not provide any meaningful abstraction. We are favouring another
> approach, that is to integrate better with the JDT and provide
> cross-language services such as code completion of java types etc. But
> you are right, we should come up with some serious examples. Currently
> we are very busy and the focus is on other tasks, such as scalability,
> code completion, code formatting and stuff.
+1 for examples of non-trivial languages. TCS has its ZOO.
EMFText has its own Java 5 grammar
http://st.inf.tu-dresden.de/reuseware/index.php/EMFText_Conc rete_Syntax_Zoo_Java_5
One look on it and you know what is possible with those projects.
Regards,
Krzysztof Kowalczyk
|
|
|
| Re: Please lower the barrier [message #37186 is a reply to message #37152] |
Mon, 06 April 2009 04:36 |
| Eclipse User |
|
|
|
we highly welcome any contributions ;-)
Sven
Krzysztof Kowalczyk schrieb:
> Hi Sebastian,
>
>>> - Prove by example that TMF can be applied to non-trivial languages,
>>> e.g. Java.
>>
>> I think, that a dsl should not be as complicated as Java itself, as it
>> would not provide any meaningful abstraction. We are favouring another
>> approach, that is to integrate better with the JDT and provide
>> cross-language services such as code completion of java types etc. But
>> you are right, we should come up with some serious examples. Currently
>> we are very busy and the focus is on other tasks, such as scalability,
>> code completion, code formatting and stuff.
>
> +1 for examples of non-trivial languages. TCS has its ZOO.
> EMFText has its own Java 5 grammar
> http://st.inf.tu-dresden.de/reuseware/index.php/EMFText_Conc rete_Syntax_Zoo_Java_5
>
>
> One look on it and you know what is possible with those projects.
>
> Regards,
> Krzysztof Kowalczyk
|
|
| | | |
| Re: Please lower the barrier [message #38196 is a reply to message #37050] |
Mon, 06 April 2009 17:18 |
| Eclipse User |
|
|
|
Sebastian Zarnekow wrote:
>>
>> If you want to reach a wider audience, please consider the following
>> issues:
>>
>> - Provide a one-stop install.
> Please have a look at
> http://oaw.itemis.com/openarchitectureware/language=en/660/d ownloads
No package for Linux x86_64...
>>
>> - Prove by example that TMF can be applied to non-trivial languages,
>> e.g. Java.
> I think, that a dsl should not be as complicated as Java itself, as it
> would not provide any meaningful abstraction.
Well, at least I need scopes, packages and multiple translation units.
Java was just an example.
> As we don't do something special with EMF models, we will most likely
> not bind resources with writing documentation on how to work with EMF
> model instances.
Now that's just the point: What is the audience you are trying to reach?
Only people who've been breathing EMF in and out for at least two years?
Or any Java developer trying to stay in touch with current trends?
Pardon my sarcasm, but it sometimes seems to me there are two groups of
developers, those who do modelling and those who do real work in real
projects, and they are almost disjoint...
To the average project guy, the model guys seem to be talking Latin all
the time, and once in a while when he comes across a tutorial which is at
least sort of readable, it is about some toy example and some
transformations that seem like art pour l'art.
So you can go on celebrating High Mass in Latin, talking to the initiated
only. Or you start evangelizing in plain English, giving the project guys
some real life examples so they will realize that there is some benefit
for their daily work to be gained from all this modelling.
In the companies I've worked with, there's an estimated 5%-10% of
developers using UML on a regular basis, which means they draw a couple of
class diagrams once a month, nothing else.
Writing a model in XML, validating it against a schema and generating code
and/or documentation from it may seem natural to you and me, but I've had
a hard time making some people see the point in this approach. I doubt
they would have been more susceptible had I been talking about M2T
transformations based on an XSD metamodel.
XText and the accompanying toolchain have the potential to make the
project guys realize that modelling is more than Abstract Nonsense. But
you still have to wrap it up and present it in a form that Joe Developer
can digest.
Regards,
Harald
|
|
|
| Re: Please lower the barrier [message #38295 is a reply to message #38196] |
Mon, 06 April 2009 18:26 |
| Eclipse User |
|
|
|
Harald,
Comments below.
Harald Wellmann wrote:
> Sebastian Zarnekow wrote:
>
>>>
>>> If you want to reach a wider audience, please consider the following
>>> issues:
>>>
>>> - Provide a one-stop install.
>
>> Please have a look at
>> http://oaw.itemis.com/openarchitectureware/language=en/660/d ownloads
>
> No package for Linux x86_64...
Wouldn't it just run on that JVM?
>
>
>>>
>>> - Prove by example that TMF can be applied to non-trivial languages,
>>> e.g. Java.
>
>> I think, that a dsl should not be as complicated as Java itself, as
>> it would not provide any meaningful abstraction.
>
> Well, at least I need scopes, packages and multiple translation units.
> Java was just an example.
>
>
>> As we don't do something special with EMF models, we will most likely
>> not bind resources with writing documentation on how to work with EMF
>> model instances.
>
> Now that's just the point: What is the audience you are trying to reach?
> Only people who've been breathing EMF in and out for at least two
> years? Or any Java developer trying to stay in touch with current trends?
As many as possible. More of that latter than the former actually...
>
> Pardon my sarcasm,
Why should it be pardoned? Either you care about what people think of
your sarcasm or you don't. You seem to be on the fence. I suppose
that's a good sign.
> but it sometimes seems to me there are two groups of developers, those
> who do modelling and those who do real work in real projects, and they
> are almost disjoint...
Sometimes it seem to me there are two groups of developers: those who
use rocks and pointy sticks and those who use high level tools. Yes,
they do seem disjoint. Do you think something like an API for reading
and writing XML Schemas is a real project or just some modeling stuff
that's light, fluffy, and barely releveant? I'd like to think XSD is a
real project. Silly me perhaps...
>
> To the average project guy, the model guys seem to be talking Latin
> all the time,
Hence the reason why I suggest folks stop using the word meta like it
was going out of style...
> and once in a while when he comes across a tutorial which is at least
> sort of readable,
And that fits his short attention span? :-P
> it is about some toy example and some transformations that seem like
> art pour l'art.
Of course if it were a more complex example, his attention span might be
strained and he might be inclined to think it's all very complicated.
Way too complicated. Stones and pointed sticks are simple...
>
> So you can go on celebrating High Mass in Latin, talking to the
> initiated only.
Preaching to the converted is a bit of a futile exercise... Preaching
in a language only the converted understand even more so!
> Or you start evangelizing in plain English,
What about German or French?
> giving the project guys some real life examples so they will realize
> that there is some benefit for their daily work to be gained from all
> this modelling.
Remember there is a balance here as well. You're talking to developers,
not religious zealots. We like to do real projects and all this
evangelizing stuff doesn't always sit with us nearly as well as you
might imagine. We might hope that people can think for themselves and
can project from simple examples to more complex tasks.
>
> In the companies I've worked with, there's an estimated 5%-10% of
> developers using UML on a regular basis, which means they draw a
> couple of class diagrams once a month, nothing else.
There's that old assumption that modeling == UML. So tiresome...
>
> Writing a model in XML, validating it against a schema and generating
> code and/or documentation from it may seem natural to you and me, but
> I've had a hard time making some people see the point in this approach.
I thought the world could imagine nothing finer than XML data with a
lovely schema to back it up? Is that not true? Is XML not the be all
and end all of human readable notations? I ask no pardon for my
sarcasm. :-P
> I doubt they would have been more susceptible had I been talking about
> M2T transformations based on an XSD metamodel.
Beer. Try to ply them with liquor first. It works well in countless
social settings.
>
> XText and the accompanying toolchain have the potential to make the
> project guys realize that modelling is more than Abstract Nonsense.
As opposed to the concrete tedium that makes their every day dreary.
Maybe we should remind people more often how completely tedious their
menial activities really are. Sure these activities keep you busy, but
if a machine can do it, is that not completely beneath you?
> But you still have to wrap it up and present it in a form that Joe
> Developer can digest.
Yes, that's an excellent point. So much is about marketing and the spin
associated with the technology. I'll bet Microsoft will spin Oslo
MGrammar extremely well. But don't forget that you're talking to
developers who like nothing better than to develop the next truly useful
feature, rather than to be preachers who like nothing better than to
hone a finer edge on next Sunday's sermon.. Somewhere there is a middle
ground. How best to find it...
>
> Regards,
>
> Harald
>
>
>
>
>
>
|
|
| | |
| Re: Please lower the barrier [message #39291 is a reply to message #39229] |
Tue, 07 April 2009 17:22 |
| Eclipse User |
|
|
|
This is a multi-part message in MIME format.
--------------060504090908060900050606
Content-Type: text/plain; charset=ISO-8859-15; format=flowed
Content-Transfer-Encoding: 7bit
Harald,
Comments below.
Harald Wellmann wrote:
> Ed Merks wrote:
>
>> Sometimes it seem to me there are two groups of developers: those who
>> use rocks and pointy sticks and those who use high level tools. Yes,
>> they do seem disjoint. Do you think something like an API for
>> reading and writing XML Schemas is a real project or just some
>> modeling stuff that's light, fluffy, and barely releveant? I'd like
>> to think XSD is a real project. Silly me perhaps...
>
> I was playing the devil's advocate.
He has an army of advocates. Why be one of them? :-P
> "Relevance" is context dependent.
Like a context sensitive grammar!
> It's quite obvious that working on any sort of modelling project is
> relevant to anyone who is doing it for fun, for research or for their
> income, if they are lucky enough to have an employer specialized in
> this area.
I consider myself to be among the lucky to be sure. Although I'm not a
big believer in luck. Good planning and good choices often lead to good
things...
>
> For people working on the other end of the value creation chain, there
> is always a pointy-haired boss who'd rather have them continue working
> with rocks and sticks than investing (read: wasting) time on tools
> outside of the application domain which are totally "irrelevant" to
> next month's deadline...
I ditched my last pointy haired boss. He was way too busy bringing me
down to his level.
< http://ed-merks.blogspot.com/2008/06/so-long-and-thanks-for- all-open-source.html>
http://ed-merks.blogspot.com/2008/06/so-long-and-thanks-for- all-open-source.html
I fully understand your frustrations. If you state them in a
constructive way, others will commiserate and things will improve based
on your constructive feedback and your contributions. Eventually
pointy-haired bosses will become irrelevant, or more likely, they'll
sing a different tune to avoid irrelevance.
>
> Regards,
>
> Harald
>
>
--------------060504090908060900050606
Content-Type: text/html; charset=ISO-8859-15
Content-Transfer-Encoding: 8bit
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-15"
http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Harald,<br>
<br>
Comments below.<br>
<br>
<br>
Harald Wellmann wrote:
<blockquote
cite="mid:ef069983dbbf9a6980900f521601728f$1@www.eclipse.org"
type="cite">Ed Merks wrote:
<br>
<br>
<blockquote type="cite">Sometimes it seem to me there are two groups
of developers: those who use rocks and pointy sticks and those who use
high level tools.
|
|
| |
| Re: Please lower the barrier [message #41597 is a reply to message #41537] |
Fri, 24 April 2009 06:11 |
| Eclipse User |
|
|
|
Hi,
> My feeling is that EMF (and associated technologies like OAW) bring an
> amazing solution to the table. I'm fully convinced of the value they
> bring to very real projects with real deadlines. And this is coming from
> a person who thinks traditional MDSD and/or UML is a complete waste of
> time.
I'm also one of these persons ;-)
> That said, the threshold for entry into this world is very significant.
> You have to be willing to go through a **lot** of pain. And I see that
> I'm not alone in my opinion on this subject [1].
I agree it's of course important to lower the barrier.
And Harald is right that it's currently hard to find a way through this
framework. And he also has some reasonable suggestions how to improve on
that.
But I won't tell him as long as he sounds like that.
It's just that we're talking about a milestone build of something which
will have it's first! release in June. And that release will have the
version 0.7.0. IMHO this should be sufficient to tell people that it's
not yet polished, isn't it? Whenever somebody asks whether Xtext is
mature enough to use in production we officially say "no" (at least not
without professional support from us).
> And yes, maybe Harald could have been a bit more "friendly". But you
> know what? I have tried the "friendly" approach myself. And I'm afraid I
> don't always see "extensive and verbose" responses from the experts, if
> I see any response at all.
At least in the case of TMF Xtext and EMF I doubt that you didn't get help.
> In this case, I have the feeling that
> although Harald brings up some good points, the biggest reason you are
> replying is because he has challenged your world.
Yes, he challenged it in a way that I loose my motivation to help him.
All that said, I understand your and Harald's point. And be sure that we
will improve on that, no matter people ask for it in a friendly or
unfriendly way.
Cheers,
Sven
|
|
|
Goto Forum:
Current Time: Tue Apr 29 08:52:07 EDT 2025
Powered by FUDForum. Page generated in 0.27958 seconds |
Home