Re: On the future of WO (here we go again)
Re: On the future of WO (here we go again)
- Subject: Re: On the future of WO (here we go again)
- From: Maik Musall <email@hidden>
- Date: Mon, 18 Mar 2019 15:11:06 +0100
Hi Faizel,
the EOF model importer wasn't perfect. Quite a few steps on the way to Cayenne
involve manually dealing with edge cases around the lesser-used features of
EOF, and I suppose this is one of them. I didn't make important use of the
userInfo dictionaries in EOF, so I can't really offer something from experience
on that specific topic. (Maybe someone else can?)
AFAIK, the Cayenne model doesn't support attaching arbitrary key-value data to
entities, and Cayenne experts will probably point out that the model should be
as purely that as it can be, and that other info should be stored elsewhere,
like e.g. in Annotations on the subclasses or something.
Maik
> Am 18.03.2019 um 05:50 schrieb Faizel Dakri <email@hidden>:
>
> Hi Maik,
>
> In your conversion from EOF to Cayenne, did you have any issues with metadata
> embedded inside your EOModel userInfo dictionaries and if so, how you did you
> go about solving them? I’ve been looking at Cayenne as a replacement for EOF
> and one roadblock I am running into at the moment is support for a userInfo
> type structure in the Cayenne datamap. Admittedly, it’s been about 6 months
> or so since I’ve looked, but from what I could tell, this appeared to be an
> open issue with Cayenne. It looked like there might be some idea on how to
> implement it at one point, but I did not see any resolution or implementation
> of it.
>
> I make extensive use of the EOModel userInfo dictionary (primarily on
> entities, but also on properties) to identify things that are configured at
> runtime. For example, I have a key named “taggable” that drives the template
> generation of my _EOEntity.java files. I also use userInfo entries to
> specify information for my auditing framework (e.g. being able to specify
> whether or an entity is audited, which keys to audit, etc.) or my custom
> filtering framework (e.g. to be able to turn on custom searches for an entity
> and to identify the keys that can be used in custom searches). These are
> just a few examples. My current thinking is to separate this metadata from
> the model and move it into its own configuration file, but that opens up the
> risk of the metadata getting out of sync with the model. It also seems a
> little dirty to me.
>
> Curious if anyone else is in a similar position, or has already solved such
> an issue?
>
> Regards,
>
> F
>
> --
> Faizel Dakri
> email@hidden <mailto:email@hidden>
>
>
>
>> On Mar 15, 2019, at 10:18 AM, Maik Musall <email@hidden
>> <mailto:email@hidden>> wrote:
>>
>> Hi Mark,
>>
>> In 2017, Hugi and I converted a large project (>800.000 lines) from EOF to
>> Cayenne, within a few months. Had parallel branches for a while and then
>> switched in production, never looked back. Cayenne is similar enough that
>> most of the work is either boilerplate conversion or actually making use of
>> the newly-gained benefits. Very few hard problems encountered, and all
>> solved.
>>
>> Let's have a talk in Frankfurt about what your EOF specifics actually are.
>>
>> Maik
>>
>>
>>> Am 15.03.2019 um 15:34 schrieb Morris, Mark <email@hidden
>>> <mailto:email@hidden>>:
>>>
>>> Just to throw our 2¢ in, we have an extremely large codebase that is very
>>> heavily invested in EOF, using it in several ways that dive deep into its
>>> bowels. ;-) Of course, we also use the WOF part of WO, and all of Wonder.
>>>
>>> Regards,
>>> Mark
>>>
>>>> On Mar 15, 2019, at 5:51 AM, Hugi Thordarson <email@hidden
>>>> <mailto:email@hidden>> wrote:
>>>>
>>>> Hi all.
>>>> In preparation for the coming WODay in Frankfurt, I'd love it if you'd be
>>>> open to having a discussion on the status and future of WO, so we can
>>>> enter the coming work prepared.
>>>>
>>>> I'd like to begin by sharing my own thoughts on the matter, based on my
>>>> current stack and experience. It's a rehash of something I posted to our
>>>> Slack yesterday, may sound revolutionary and will no doubt be
>>>> controversial, but I think some outside-the-box thinking is required at
>>>> this time. This is lengthy, sorry about that…
>>>>
>>>> --
>>>>
>>>> In the past few years I've been working towards minimising the use and
>>>> effect of WO/Wonder on my stack, so when and if The Time comes, I and my
>>>> customers have a migration path forward. Among the things I've done is
>>>> move from EOF to Cayenne and from Ant to Maven (to make using 3rd party
>>>> jars, including Cayenne easier), both of which have turned out to have
>>>> been very happy decisions which I wholeheartedly recommend, regardless of
>>>> anything else you do.
>>>>
>>>> I love working with my WO/Cayenne stack, which is currently only
>>>> "polluted" by the following frameworks:
>>>>
>>>> -- WO:
>>>> * JavaFoundation (indirectly through WO, I never use foundation classes in
>>>> my code unless absolutely required by WO)
>>>> * JavaWebObjects
>>>>
>>>> -- Wonder (I consider Wonder "polluted" since it depends on WO/EOF)
>>>> * ERExtensions (only the WO stuff, not the EOF stuff)
>>>> • Ajax
>>>> • WOOgnl (indirectly for parsing Wonder-style inline templates)
>>>> …and of course then there's the deployment stuff (JavaMonitor,wotaskd,
>>>> adaptors).
>>>>
>>>> Given this, here's my proposal for a way forward:
>>>> * We abandon EOF (and, in fact, any ORM—this is not meant to be a full
>>>> stack effort, initially at least)
>>>> * We re-implement JavaWebObjects as required (and the absolutely necessary
>>>> parts of JavaFoundation, such as KVC and NSBundle) as a single framework
>>>> * We separate the necessary WO stuff from the EOF/D2W stuff in Wonder (as
>>>> well as other totally unrelated things like mail sending frameworks, other
>>>> utility frameworks and "useful applications") and include it in our
>>>> re-implementation
>>>> * We create a fork of WOLips that knows how to live within the New Universe
>>>> * We rule the world
>>>>
>>>> Ideally, what we end with is Just a Web Framework™ with IDE integration
>>>> (and nothing else) that can serve as a basis for future development. While
>>>> re-implementing WO may sound like a huge undertaking, I actually think
>>>> it's smaller than rewriting all of my solutions that depend on it. This
>>>> probably applies to more of you.
>>>>
>>>> Now, looking at my own stack I know this proposal might sound a bit
>>>> self-serving, but I'd like to hear other opinions. I believe it's a
>>>> realistic way forward with (comparatively) minimal development effort.
>>>> Turns out that WOF itself is the only part of the WO/Wonder stack that I
>>>> really just don't want to live without.
>>>>
>>>> This is something I'd like to do, and if anyone likes the idea and is
>>>> willing to participate, I'm confident we can make this work! Doing stuff
>>>> alone sucks.
>>>>
>>>> Cheers,
>>>> - hugi
>>>> _______________________________________________
>>>> Do not post admin requests to the list. They will be ignored.
>>>> Webobjects-dev mailing list (email@hidden
>>>> <mailto:email@hidden>)
>>>> Help/Unsubscribe/Update your Subscription:
>>>>
>>>>
>>>> This email sent to email@hidden
>>>> <mailto:email@hidden>
>>>
>>> _______________________________________________
>>> Do not post admin requests to the list. They will be ignored.
>>> Webobjects-dev mailing list (email@hidden
>>> <mailto:email@hidden>)
>>> Help/Unsubscribe/Update your Subscription:
>>>
>>>
>>> This email sent to email@hidden <mailto:email@hidden>
>> _______________________________________________
>> Do not post admin requests to the list. They will be ignored.
>> Webobjects-dev mailing list (email@hidden
>> <mailto:email@hidden>)
>> Help/Unsubscribe/Update your Subscription:
>>
>> This email sent to email@hidden <mailto:email@hidden>
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden