| Xtext formatter [message #62232] |
Wed, 22 July 2009 04:24  |
Eclipse User |
|
|
|
Hello,
I really like the formatting support in the new Xtext version. However,
creating a formatting configuration for a large grammer is often tedious.
As far as I understood the documentation, I have to configure the
formatter for each rule specific token, e.g.:
c.setNoSpace().around(f.getParamAccess().getColonKeyword_1() );
In our grammar we have 10 different rules in which we use a colon keyword
and therefore I have to configure each of them separately. In order to
avoid this, my idea was to define a terminal for the colon keyword and
then configure the formatter based on the terminal. This would enable the
reuse of formatting rules across different grammar rules:
c.setNoSpace().around(f.getDOTRule());
However, this did not work. Is there another way to configure the
formatter in a more general way?
This also applies to curly braces where would like to write something like:
c.setIndentation(
f.getLeftCurlyBracketRule(),
f.getRightCurlyBracketRule());
c.setLinewrap().after(
f.getLeftCurlyBracketRule());
c.setLinewrap().after(
f.getRightCurlyBracketRule());
Regards,
Sebastian
|
|
|
| Re: Xtext formatter [message #62355 is a reply to message #62232] |
Wed, 22 July 2009 05:22  |
| Eclipse User |
|
|
|
Hi Sebastian,
> I really like the formatting support in the new Xtext version.
good to hear!
> However,
> creating a formatting configuration for a large grammer is often
> tedious. As far as I understood the documentation, I have to configure
> the formatter for each rule specific token, e.g.:
>
> c.setNoSpace().around(f.getParamAccess().getColonKeyword_1() );
that is right.
> In our grammar we have 10 different rules in which we use a colon
> keyword and therefore I have to configure each of them separately. In
> order to avoid this, my idea was to define a terminal for the colon
> keyword and then configure the formatter based on the terminal. This
> would enable the reuse of formatting rules across different grammar rules:
>
> c.setNoSpace().around(f.getDOTRule());
Assigning formatting to rules currently only works with rules for hidden
tokens..
> However, this did not work. Is there another way to configure the
> formatter in a more general way?
you can build your own:
<pseudocode>
for(AbstractElement e: grammarAccess.getGrammar().eAllElements())
if(e instanceof Keyword) {
Keyword k = (Keyword) e;
if(".".equals(k.getValue()))
cfg.setNoSpace().around(k);
}
</psudocode>
> This also applies to curly braces where would like to write something like:
>
> c.setIndentation(
> f.getLeftCurlyBracketRule(),
> f.getRightCurlyBracketRule());
> c.setLinewrap().after(
> f.getLeftCurlyBracketRule());
> c.setLinewrap().after(
> f.getRightCurlyBracketRule());
I guess you can think of a similar loop for finding all pairs of curly
brackets...
Generally speaking, I can say that we a re aware of this situation and
thinking about how to improve it. Another issue is, that methods-names
of the GrammarAccess tend to change when you modify your grammar.
One idea is to extend the GrammarAccess with methods that return all
keywords of a certain type, e.g.
ga.getAllColonKeywords()
Another is to introduce find-methods:
ga.findKeywords(".")
hth,
Moritz
--
Need professional support for Eclipse Modeling?
Go visit: http://xtext.itemis.com
|
|
|
| Re: Xtext formatter [message #62731 is a reply to message #62355] |
Wed, 22 July 2009 10:40 |
| Eclipse User |
|
|
|
Hi Moritz,
thanks for your help! It works as you said:
protected void configureFormatting(FormattingConfig c) {
MyGrammarAccess f = (MyGrammarAccess) getGrammarAccess();
// colons
Iterable<Keyword> keywords = GrammarUtil.containedKeywords(f.getGrammar());
for (Keyword currentKeyword : keywords) {
if (".".equals(currentKeyword.getValue())) {
c.setNoSpace().around(keyword);
}
}
// braces
Iterable<Keyword> keywords = GrammarUtil.containedKeywords(f.getGrammar());
Stack<Keyword> openBraceStack = new Stack<Keyword>();
for (Keyword currentKeyword : keywords) {
if ("{".equals(currentKeyword.getValue())) {
openBraceStack.add(currentKeyword);
c.setLinewrap().after(currentKeyword);
}else if ("}".equals(currentKeyword.getValue())) {
c.setLinewrap().after(currentKeyword);
c.setLinewrap().before(currentKeyword);
if (!openBraceStack.isEmpty()) {
c.setIndentation(openBraceStack.pop(), currentKeyword);
}
}
}
}
I think it would be really helpful if one could configure the formatting
at rule level, for example:
Statement:
Assignment | Return
;
...
=>
c.setLinewrap().after(f.getStatementRule());
Regards,
Sebastian
Moritz Eysholdt wrote:
> Hi Sebastian,
>> I really like the formatting support in the new Xtext version.
> good to hear!
>> However,
>> creating a formatting configuration for a large grammer is often
>> tedious. As far as I understood the documentation, I have to configure
>> the formatter for each rule specific token, e.g.:
>>
>> c.setNoSpace().around(f.getParamAccess().getColonKeyword_1() );
> that is right.
>> In our grammar we have 10 different rules in which we use a colon
>> keyword and therefore I have to configure each of them separately. In
>> order to avoid this, my idea was to define a terminal for the colon
>> keyword and then configure the formatter based on the terminal. This
>> would enable the reuse of formatting rules across different grammar rules:
>>
>> c.setNoSpace().around(f.getDOTRule());
> Assigning formatting to rules currently only works with rules for hidden
> tokens..
>> However, this did not work. Is there another way to configure the
>> formatter in a more general way?
> you can build your own:
> <pseudocode>
> for(AbstractElement e: grammarAccess.getGrammar().eAllElements())
> if(e instanceof Keyword) {
> Keyword k = (Keyword) e;
> if(".".equals(k.getValue()))
> cfg.setNoSpace().around(k);
> }
> </psudocode>
>> This also applies to curly braces where would like to write something like:
>>
>> c.setIndentation(
>> f.getLeftCurlyBracketRule(),
>> f.getRightCurlyBracketRule());
>> c.setLinewrap().after(
>> f.getLeftCurlyBracketRule());
>> c.setLinewrap().after(
>> f.getRightCurlyBracketRule());
> I guess you can think of a similar loop for finding all pairs of curly
> brackets...
> Generally speaking, I can say that we a re aware of this situation and
> thinking about how to improve it. Another issue is, that methods-names
> of the GrammarAccess tend to change when you modify your grammar.
> One idea is to extend the GrammarAccess with methods that return all
> keywords of a certain type, e.g.
> ga.getAllColonKeywords()
> Another is to introduce find-methods:
> ga.findKeywords(".")
> hth,
> Moritz
|
|
|
Powered by
FUDForum. Page generated in 0.03690 seconds
Home