[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
|
Re: [ease-dev] GSoC code competion commits
|
Hi Christian,
I can understand your concern about the IModuleWrapper interface, I
actually just implemented the code there because you suggested it in the
initial refactor mail. I guess the best solution would be to add a new
interface that provides code completion for each scripting language. The
actual implementation of the interface can be done very fast, because I
only need to move the corresponding methods.
One problem I have at the moment is where to integrate the querying for
the instance. The IScriptService interface has a method to get the
ModuleWrapper and it would probably be best to add a method to get the
CompletionAnalyzer there as well. If I change the IScriptService
interface though, we are back to where we started.
One solution would be to create a new extension for this code analyzer
and add a static method in the concrete ScriptService implementation to
query for it. Do you have any other/better solutions?
Regarding the hardcoded completion providers, I was already working on
this. I needed to change the extension point to take care of the things
you mentioned, that's why I haven't committed it yet. The providers were
hardcoded because it was the fastest way to show they are working.
I guess it will be easiest if I discard the commits after the one
breaking the interface, wait for the new interface to be accepted and
then commit the following changes again. Most of them do not depend on
the interface directly (they only need a refined context) and I could
reintegrate everything very fast (mostly copy + paste).
I actually gave every commit a different message but always started out
with the bug ID line and since Hudson only displays the first line of a
commit message this may have let to some confusion. I will open some
more bugs to make distinguishing a bit easier...
Hope this sheds some light on the situation and that we can fix this
fast and move on to more fun stuff :-)
Best regards,
Martin
Am 2015-07-31 um 23:24 schrieb Christian Pontesegger:
Hi Martin,
your latest commits brought us in a delicate situation.
You commited 14 topics, all depending on each other. Some of them will
need some adaptions/discussions. But if 1 patch set is declined, the
next elements in the chain cannot be merged anymore. So you have to
step back to the first declined topic and start from there to rework
the topic.
Also having 14 commits with the exact same commit message is not
exactly what I was expecting. As discussed offline: smaller commits do
not only help to avoid the 1000 LOC rule. They also help to cluster
topics and make reviews easier.
One major concern of mine are those changes to the IModuleWrapper
interface. While you adapted the one for Rhino, your changes will
break compatibility for our external interpreters for Python, Groovy
and Ruby. IModuleWrappers are there to generate code chunks, so I
think they might be useful when you want to convert a java call to a
scripted command, but currently I see no point in changing them for
code completion proposals. If you think these changes are needed, we
should discuss it on the list.
A second topic is that the script shell should not need knowledge of
several code completion implementations. Your adapter should take care
of chosing the correct completion providers depending whether we are
working on a live interpreter or within an editor. So one option would
be to set a script interpreter instance and/or a script type in the
adapter. Then the adapter can chose matching completion providers by
itself.
Completion providers generally should not be hard coded into the
system. We should use extension points to register them along with
some metadata (eg a language filter, whether they are applicable on a
script interpreter or also offline, ...). Then the adapter should load
them dynamically depending on the environment.
HTH
Christian
_______________________________________________
ease-dev mailing list
ease-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or
unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/ease-dev
