I don't think we will be able to eliminate mini-language entirely. As
David has mentioned, and I agree - it is very convenient to add a few
lines of XML to a screen widget (or service definition in his example).
The purpose of the overhaul is to fix the numerous issues in
mini-language that have been there for years. I believe those issues are
partially responsible for the waning interest in mini-language. My hope
is that the overhaul will inspire a renewed interest in the language. If
that doesn't happen, then at least we will have a decent reference and
implementation that can be spun off to a separate project.
If a migration from mini-language to groovy were to take place, then I
would recommend creating a utility to perform the conversion. As I have
mentioned previously, there is a mountain of mini-language code in
OFBiz, and converting it all manually will be impossible. A conversion
utility will depend upon reliable mini-language source code (the XML
files, not the Java source). My initial tests with mini-language <set>
element validation have shown that there are a lot of errors and
nonsensical code in the existing mini-language XML code base. A
conversion utility will have a hard time making sense of it. Therefore,
from my perspective, regardless of whether mini-language stays or goes,
some work will need to be done to clean up the existing mini-language
XML code base.
-Adrian
On 4/5/2012 10:54 AM, Jacques Le Roux wrote:
From a pragmatic point of view I'd not be against. The less syntaxes
the better, and I know a lot of users are not using minilang
for this reason. But it seems Adrian has already begin to work on the
Mini-language Overhaul...
Adrian should we not concentrate our efforts on reducing syntaxes in
OFBiz? On the other hand I know replacing all minilang code in OFBiz
would be a long and "risky" task...
Jacques
From: "Jacopo Cappellato" <jacopo.cappell...@hotwaxmedia.com>
I completely agree that having *one* set on Minilang operations for
all the actions (forms/screens/methods) is a good step forward.
My only doubt is about the effort required for the migration: I see
an opportunity to migrate all our applications code to Groovy
instead; at that point all OFBiz code will be either Java or Groovy
and Freemarker.
Jacopo
On Apr 5, 2012, at 1:45 AM, Adrian Crum wrote:
If we can agree on the mini-language syntax and get a good
implementation, then the next step would be to integrate it with the
screen widgets. Personally, I have always wanted the screen widget
<actions> element to behave the same way as the mini-language
<simple-method> element. That concept can be extended further to
have <simple-method> elements in the screen widget files
(https://issues.apache.org/jira/browse/OFBIZ-4090).
-Adrian
On 4/4/2012 6:41 PM, Jacques Le Roux wrote:
Hi Adrian,
This looks like a great step forward. I really wonder though if we
will still have many possible syntaxes ("std minilang"
(should not vary with the place it's used), bsh, grooy, uel, etc.)
or if it will be more constrained
I guess you will follow David's suggestion on this?
What really bothers me with the simple-method stuff in OFBiz
whenever I
use it is the inconsistency between the simple-method, screen
actions,
form actions, etc, and the inconsistent handling of expressions
and such
for field, from-field, value, etc attributes. Having everything
use the
same XSD and having everything actually be a groovy expression
(since it
is just transformed into a groovy script) made it much easier and
less
frustrating to use.
I plenty agree with him. Moreover, I think we don't need many
syntaxes, as long as the one provided supersedes all the others.
On a long term, I can help to refactor current minilang code (which
means also screens and form actions of course).
Thanks
Jacques
From: "Adrian Crum" <adrian.c...@sandglass-software.com>
I think I have come up with a solution to fix the mini-language
code: mini-language auto-correction. If enabled, the element
models correct common mistakes and save the corrections to the
original file. This will still leave some warnings that will
need to be fixed manually.
-Adrian
On 4/3/2012 2:07 PM, Adrian Crum wrote:
I added parsing validation to the <set> element to see how it
will work. The validation can be enabled/disabled through a
property setting.
Just running the load-demo ant task generates hundreds of
warnings - most of them are caused by <set> attributes being used
incorrectly. On the positive side, a lot of nonsensical code is
being revealed; on the negative side, the log is filled with
warnings.
I'm not sure what I will do with this. If I commit the validation
code, then we will have a lot of work to do to clean up the
mini-language code.
-Adrian
On 3/8/2012 6:55 PM, Adrian Crum wrote:
Some more food for thought...
Looking through the Java code, I can see that there is no
runtime validation being performed. Granted, a decent XML editor
will warn you about required attributes and elements and such,
but not everyone uses that type of XML editor. Worse yet,
there is no way to know you've done something wrong - because
mini-lang hides the errors in verbose log statements. So, I
would like to add runtime validation. If the script is coded
improperly, then it should throw an exception.
Another change I would like to make is to remove default
attribute values in the schema. From my perspective, defaults
should
be in the mini-language code. The wiki page has demonstrated to
me how difficult it is to understand what's going on when you
have to look through Java code, and then also look through the
schema to see what values it is supplying. Plus, it makes me
wonder how mini-language will behave when the server doesn't
have access to the schema.
Which brings up another point: Once the grammar has been cleaned
up, we will need a new schema. I think we need to start
giving our schemas version numbers so that XML editors and
runtime XML validation will work properly.
-Adrian
On 3/8/2012 6:19 PM, Jacopo Cappellato wrote:
On Mar 8, 2012, at 7:03 PM, Adrian Crum wrote:
B) instead of:
<fail-property resource="ErrorMessages" property="FooError" />
we could have
<fail-property property="ErrorMessages.FooError" />
Keep in mind that UEL would interpret FooError as an element
of a Map called ErrorMessages.
What you suggested can be done, but it will require more
modifications to the UEL integration - something I try to avoid
because it causes more problems than it solves. I recommend we
keep the resource attribute.
Np then, I was not even sure it was a good idea and if requires
customizations to uel then I agree it is not worth the
effort.
Thank you,
Jacopo
