Yes, that's why it's a long task. I have to consider all cases carefully.
That's also why I added this last comment (quoted below) in OFBIZ-8341
"I'll sometimes create subtasks or new Jira issues to differentiate cases that
need to be discussed.
It would be good for instance for a type of exception and a type of file (service, event, helper/handler/test/etc.) to use and adopt a same type of
exception handling."
Having patterns would help everybody, when creating, reviewing, refactoring,
etc.
Jacques
Le 11/07/2017 à 09:47, Taher Alkhateeb a écrit :
This thread is a good example of refactoring. So mass fixing of
swallowed exceptions is not ideal IMHO because sometimes we want to
log things, sometimes we want to re-throw, sometimes we want a
different path. Hence each item should be refactored slowly and in
isolation because if you just throw a log in there then people would
think this code is probably okay and doesn't need review.
Anyway, again I appreciate all the help in your reviews, the feature
is more or less implemented in OFBIZ-9441.
On Tue, Jul 11, 2017 at 12:35 AM, Jacques Le Roux
<jacques.le.r...@les7arts.com> wrote:
I'll refrain to speak about swallowed exceptions ;)
I still want to continue on OFBIZ-8341 but accepted to get sidetracked in
multi ways :)
Jacques
Le 10/07/2017 à 21:36, Taher Alkhateeb a écrit :
Fixed it in the JIRA, the EntitySaxReader (should be the next class to
refactor) is logging an error but suppressing an exception. The logic
for "continue-on-error" had to go 3 classes deep to work correctly. I
always underestimate how much spaghetti code we have :)
On Mon, Jul 10, 2017 at 8:14 PM, Taher Alkhateeb
<slidingfilame...@gmail.com> wrote:
Quick update, to my surprise an exception is swallowed
deeeeeeeeeeeeeeeeeeeeeeeeeeeep in the call stack. So getting this
feature might require some intrusive changes. I'm still working on it
and will keep you posted, but as of right now, no exception is
bubbling up to be caught with "continue-on-failure"
On Mon, Jul 10, 2017 at 2:14 PM, Taher Alkhateeb
<slidingfilame...@gmail.com> wrote:
Thank you everyone for your feedback, I will let this discussion
continue for a few days before committing anything (testing is going
to take some time anyway).
Now, I need help, I have a big patch in [1] that does what we
discussed in this thread and a whole lot more. If you have the time, I
really need your help! Most useful help is testing, there are so many
properties and combinations to use, so this requires thorough testing.
Also for those familiar with the core framework API, a second look at
the code would help.
[1] https://issues.apache.org/jira/browse/OFBIZ-9441
On Mon, Jul 10, 2017 at 1:45 PM, Devanshu Vyas
<vyas.devansh...@gmail.com> wrote:
I agree with option #3 and the 'continue-on-failure' flag with default
value=false. :)
Thanks & Regards,
Devanshu Vyas.
On Mon, Jul 10, 2017 at 3:07 PM, Taher Alkhateeb
<slidingfilame...@gmail.com
wrote:
Hi Rishi,
So my suggestion is that if anything does not load, then immediately
fail.
Why am I suggesting this?
- You have to intentionally ignore data failure after being aware of
it (it does not slip between the cracks)
- The data will automatically get cleaned by committers because no
failing data will be committed to the code base.
I suspect we will actually catch some data loading failures that exist
in the code base and we are maybe unaware of.
On Mon, Jul 10, 2017 at 12:04 PM, Rishi Solanki
<rishisolan...@gmail.com>
wrote:
I'm good to go with option #3 and continue-on-failure.
Just wanted to mention one thing here; for which type of data we will
be
failing build. That means, we have several options seed, ext, demo.
Do we
need to discuss these points or we are fine for all type of data.
Like
demo
data fails only affect a process for that data set only, and for that
failing build is okay or not (as on data load we get logs if any file
didn't load).
Btw, I'm good with the proposal, just sharing a thought in case we
should
discuss or may be we can simply ignore if we are good with that.
Thaks!
Rishi Solanki
Sr Manager, Enterprise Software Development
HotWax Systems Pvt. Ltd.
Direct: +91-9893287847
http://www.hotwaxsystems.com
On Mon, Jul 10, 2017 at 2:15 PM, Deepak Dixit <
deepak.di...@hotwaxsystems.com> wrote:
Historically the data loader boolean props are false if ommitted
and
the
code expects that, but you have a point about the double negative.
We
can
instead call it "continue-on-failure" for example.
+1 continue-on-failure with default value false
Thanks & Regards
--
Deepak Dixit
www.hotwaxsystems.com
www.hotwax.co
On Jul 10, 2017 3:48 AM, "Paul Foxworthy" <p...@cohsoft.com.au>
wrote:
Hi all,
I agree with option 3. I recall in my own work I once needed to add
a
throw
where there was none to track down a problem.
However ignore-failure leads to a double negative. How about
"stop-on-failure", default value true?
Cheers
Paul Foxworthy
On 10 July 2017 at 05:27, Taher Alkhateeb
<slidingfilame...@gmail.com
wrote:
Correction: on item (2) in my post: fail immediately, not after
loading all files, otherwise there's no point.
On Sun, Jul 9, 2017 at 10:18 PM, Taher Alkhateeb
<slidingfilame...@gmail.com> wrote:
Hello Everyone,
For a long time I was annoyed by something in OFBiz: the build
system
does not fail if data loading fails for some files. I spend hours
hunting bugs only to discover that the data simply did not load.
Given that I'm working on refactoring the data loading container,
I
believe this issue should resolved. However, I'm not sure if the
community is interested in making such a change.
So I list below 3 options to select from:
1- Leave it as is, do not fail the build if some files do not
load
2- Continue loading until all files are done and then fail the
build
3- Provide a flag e.g. ignore-failure that tells the system
whether
to
fail or not with a default value of "false".
My personal preference is for (3)
WDYT?
--
Coherent Software Australia Pty Ltd
PO Box 2773
Cheltenham Vic 3192
Australia
Phone: +61 3 9585 6788
Web: http://www.coherentsoftware.com.au/
Email: i...@coherentsoftware.com.au
