Re: [rules-dev] @metadata: status; compatibility issues
On 29 April 2011 19:18, Edson Tirelli <ed.tirelli(a)gmail.com> wrote:
Wolfgang,
It seems to me that what we need to to improve the solution to address
the problems you listed, but we are better now then we were before, as we
support everything that was supported in 5.1 + the new use cases.
Correct!
Did I miss something?
Only if you intend to do, in a later release, s.th. with those
expressions, such as
evaluate them if they are all literal, or maybe with
imported variables, if that can be done at all. If yes, the thing I'd
propose is to either announce (clearly!) the future change ("expressions
will not remain expressions returned as Strings") OR reduce the accepted
expressions to single literals of the basic types, which could be converted
to Integer/String/Boolean objects according to their natural syntax.
Perhaps the concatenation of duplicated inner Map values should be removed -
I'm afraid I did that, in an effort to avoid silent overwriting, which
definitely isn't better; both is bad.
Wolfgang
Edson
2011/4/29 Wolfgang Laun <wolfgang.laun(a)gmail.com>
> The current parser handles metadata annotations
>
> '@' *Identifier* '(' *Text* ')'
>
> so that the annotations for one entity (e.g., a rule) are available as a
> Map<String,Object>, where the key is given by the *Identifier *and the
> value of the entry is available according to
>
> *if* *Text* can be *completely* parsed as *Identifier* '=' *Expression
> * (where *Expression* is whatever the DRL parser accepts as an
> expression, i.e., an extended subset of Java expressions)
> *then*
> the value is a Map<String,String>
> *else
> the *value is a String, with trimmed leading and trailing white
> space, but preserving all embedded white space
>
> Changes compared to 5.1.1:
> - no Map was ever returned in 5.1.1,
> - @m( "foo" ) in 5.1.1. returned "foo", but 5.2.0 returns
"\"foo\"".
> - @m( "foo", "bar" ) in 5.1.1. returned "foo\",
\"bar", but 5.2.0
> returns "\"foo\", \"bar\""
> - comments between "@...(" and ")" are now handled
consistently but
> were not in 5.1.1
>
> Possibly considered a problem for 5.2.0 or later:
> - No check is made for duplicate "outer" map keys; entries are
> overwritten.
> - Repeated "inner" map keys produce concatenated entries, e.g. @m( k = 1,
> k = "a" ) returns "1a" as the value for key "k".
> - Returning (arbitrary) expressions, unchanged, as String now may cause
> incompatibilitie with more sophisticated processing (if this ever should be
> considered).
>
> -W
>
>
>
>
>
>
>
>
>
>
> _______________________________________________
> rules-dev mailing list
> rules-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-dev
>
>
--
Edson Tirelli
JBoss Drools Core Development
JBoss by Red Hat @ www.jboss.com
_______________________________________________
rules-dev mailing list
rules-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-dev
Attachments:
- attachment.html (text/html — 4.7 KB)