Jython

On Thu, Mar 01, 2007 at 09:03:42PM +0000, James Abley wrote:
> Apparently, they aren't interested in fixing the issue in the older
> releases, since 2.5 is the current good release. Obviously, I don't

I have commit privileges to the CPython SVN.  Older release or not,
I'm happy to commit any changes to the 22-maint branch that will
enable the tests to be used with Jython.

> c) I'd prefer to use unittest anyway, rather than the custom framework
> that test_unicodedata.py currently uses.

Note that converting tests to unittest is an ongoing task for CPython,
so, if this hasn't already been done in the current trunk, please
contribute the patch back to CPython.

--amk

On 3/1/07, James Abley <james.abley@...> wrote:
> Hi all,

> 1. JUnit / TestNG / ANOther isn't used - just CPython tests. How well
> do people find that works with regard to refactoring?
Personally, I prefer unittest to Java-based unit testing.  I think
that making the CPython unit tests better for cross-implementation
testing is a place where we (the Jython development community) can
really make a contribution back to CPython.

> 2. At least for unicodedata, the existing 2.3 CPython tests aren't
> quite what I'd like.  CPython 2.3 and 2.4 unicodedata contains at
> least one bug, which has been confirmed by via private email
> discussion with one of the CPython developers (Martin von L=F6wis).
> Apparently, they aren't interested in fixing the issue in the older
> releases, since 2.5 is the current good release. Obviously, I don't
> intend to create the same bug in any implementation I provide, but
> that means that the 2.3 CPython tests won't pass. So, I would like to
> change those tests for three reasons.

> a) They aren't sufficiently granular.
> b) They are wrong.
> c) I'd prefer to use unittest anyway, rather than the custom framework
> that test_unicodedata.py currently uses.
I have no problem with pulling the 2.5 test in early (although not for
the 2.2 release) -- especially since we are going to up our target for
the next release to either 2.4 or 2.5.  As AMK points out any
improvements we make to the CPython unittests can and should be
contributed back.

-Frank

On 3/1/07, James Abley <james.abley@...> wrote:
> 1. JUnit / TestNG / ANOther isn't used - just CPython tests.
There is a set of Jython-specific tests that use a Jython-custom
framework -- see the "bugtests" directory which is a sibling directory
of "src".  Someday I'd like to get these tests to be somehow based on
unittest -- but I think this will be a large task.

-Frank

Hi all,

on behalf of JUnits:

there are some Java JUnit tests in the installer part.
And in fact I usually write a JUnit test in the core part, too.

But I didn't want to introduce them for 2.2.

There are some strong pro's and con's, especially regarding the
existing test base (as Frank pointed out).

If possible, I'd like to discuss this face to face with those
developers I'm going to meet at JavaOne.

best wishes,
Oti.

On 3/2/07, Frank Wierzbicki <fwierzbicki@...> wrote:
> On 3/1/07, James Abley <james.abley@...> wrote:
> > 1. JUnit / TestNG / ANOther isn't used - just CPython tests.
> There is a set of Jython-specific tests that use a Jython-custom
> framework -- see the "bugtests" directory which is a sibling directory
> of "src".  Someday I'd like to get these tests to be somehow based on
> unittest -- but I think this will be a large task.
>
> -Frank
>
> -------------------------------------------------------------------------
> Take Surveys. Earn Cash. Influence the Future of IT
> Join SourceForge.net's Techsay panel and you'll get the chance to share your
> opinions on IT & business topics through brief surveys-and earn cash
> http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
> _______________________________________________
> Jython-dev mailing list
> Jython-dev@...
> https://lists.sourceforge.net/lists/listinfo/jython-dev
>

On 02/03/07, A.M. Kuchling <amk@...> wrote:
> On Thu, Mar 01, 2007 at 09:03:42PM +0000, James Abley wrote:
> > Apparently, they aren't interested in fixing the issue in the older
> > releases, since 2.5 is the current good release. Obviously, I don't
>
> I have commit privileges to the CPython SVN.  Older release or not,
> I'm happy to commit any changes to the 22-maint branch that will
> enable the tests to be used with Jython.
>
> > c) I'd prefer to use unittest anyway, rather than the custom framework
> > that test_unicodedata.py currently uses.
>
> Note that converting tests to unittest is an ongoing task for CPython,
> so, if this hasn't already been done in the current trunk, please
> contribute the patch back to CPython.

I can write the tests, but I have no intention of writing C to fix CPython!

Thanks.

>
> --amk
>
> -------------------------------------------------------------------------
> Take Surveys. Earn Cash. Influence the Future of IT
> Join SourceForge.net's Techsay panel and you'll get the chance to share your
> opinions on IT & business topics through brief surveys-and earn cash
> http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
> _______________________________________________
> Jython-dev mailing list
> Jython-dev@...
> https://lists.sourceforge.net/lists/listinfo/jython-dev
>

The main reason (that I've encountered so far, but I haven't written
any real code) is what feels like the slow rhythm of test, code, test.
Having to run a full build and then run the specific test case each
time is a lot slower than I'm used to and in an ideal situation, the
feedback loop is a lot shorter. But maybe that's just me.

I appreciate and applaud the rationale for using Python tests, but I
was curious as how other people feel when developing Jython.

Cheers,

James

On 02/03/07, Oti <ohumbel@...> wrote:
> Hi all,
>
> on behalf of JUnits:
>
> there are some Java JUnit tests in the installer part.
> And in fact I usually write a JUnit test in the core part, too.
>
> But I didn't want to introduce them for 2.2.
>
> There are some strong pro's and con's, especially regarding the
> existing test base (as Frank pointed out).
>
> If possible, I'd like to discuss this face to face with those
> developers I'm going to meet at JavaOne.
>
> best wishes,
> Oti.
>
> On 3/2/07, Frank Wierzbicki <fwierzbicki@...> wrote:
> > On 3/1/07, James Abley <james.abley@...> wrote:
> > > 1. JUnit / TestNG / ANOther isn't used - just CPython tests.
> > There is a set of Jython-specific tests that use a Jython-custom
> > framework -- see the "bugtests" directory which is a sibling directory
> > of "src".  Someday I'd like to get these tests to be somehow based on
> > unittest -- but I think this will be a large task.
> >
> > -Frank
> >
> > -------------------------------------------------------------------------
> > Take Surveys. Earn Cash. Influence the Future of IT
> > Join SourceForge.net's Techsay panel and you'll get the chance to share your
> > opinions on IT & business topics through brief surveys-and earn cash
> > http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
> > _______________________________________________
> > Jython-dev mailing list
> > Jython-dev@...
> > https://lists.sourceforge.net/lists/listinfo/jython-dev
> >
>

On Fri, Mar 02, 2007 at 06:13:29PM +0000, James Abley wrote:
> The main reason (that I've encountered so far, but I haven't written
> any real code) is what feels like the slow rhythm of test, code, test.
> Having to run a full build and then run the specific test case each
> time is a lot slower than I'm used to and in an ideal situation, the
> feedback loop is a lot shorter.

Isn't the slow step there the full build step, not the running of the
actual test?

--amk

On 3/2/07, James Abley <james.abley@...> wrote:
> The main reason (that I've encountered so far, but I haven't written
> any real code) is what feels like the slow rhythm of test, code, test.
> Having to run a full build and then run the specific test case each
> time is a lot slower than I'm used to and in an ideal situation, the
> feedback loop is a lot shorter. But maybe that's just me.
You shouldn't have to run a full build each time -- you could do the
default developer build, which should only operate on the files you
have changed.

-Frank

On 02/03/07, Frank Wierzbicki <fwierzbicki@...> wrote:
> On 3/2/07, James Abley <james.abley@...> wrote:
> > The main reason (that I've encountered so far, but I haven't written
> > any real code) is what feels like the slow rhythm of test, code, test.
> > Having to run a full build and then run the specific test case each
> > time is a lot slower than I'm used to and in an ideal situation, the
> > feedback loop is a lot shorter. But maybe that's just me.
> You shouldn't have to run a full build each time -- you could do the
> default developer build, which should only operate on the files you
> have changed.
>
> -Frank

That is what I'm doing. In my day-to-day development, I normally get
less than a 30 second turnaround between making a code change and
seeing whether the test still fails. I'm not getting that currently,
which feels a little strange. The build is the slow step - rather than
background compilation in the IDE, it's using ANT. I think I'll have
to play around with my setup and see what scope there is for
optimizing parts of it.

Cheers,

James

2007/3/4, James Abley <james.abley@...>:
> The build is the slow step - rather than
> background compilation in the IDE, it's using ANT. I think I'll have
> to play around with my setup and see what scope there is for
> optimizing parts of it.

There's something related to this at the end of
http://wiki.python.org/jython/JythonDeveloperGuide/EclipseNotes. If
you figure out a nice setup you should perhaps write something more
there.

Cheers,
    .peke

(originally sent this only to Frank and not to the list... sorry for  
the delay)

James, jython's test-code-test rhythm feels pretty slow to me too.   
Frank, I was using the default developer build at the sprint at PyCon  
-- still felt too slow to me.  I asked Charlie about writing JUnit  
tests while we were at the sprint.  He mostly persuaded me that the  
python tests were better, for all the reasons already presented in  
this thread.

Charlie's rhythm sounds faster, but I haven't got my environment  
configured to match his yet.  It's been three years since I worked in  
Java and some of my own slowness is just getting re-acquainted with  
an IDE and such.

Anyway he's got a custom shell script for running jython development  
builds.  They run with ./build in the classpath.  Eclipse compiles  
changes directly into ./build, so he doesn't run the developer build  
at all.  He edits java code in Eclipse and then runs jython unit  
tests through his custom shell script.  Pretty fast cycle.  Charlie,  
will you post your script please?  And maybe some notes about how you  
configured Eclipse?

Take care.
-Eric

The only customization I make to Eclipse is to tell it to compile into
./build instead of its default of ./bin.  As long as it's set to
compile on save(which I think is the default), then your changes are
compiled without running ant.  You should be able to do the same thing
with any other IDE that will compile on save like NetBeans or IDEA.
Then you need to update your jython startup script to use its output:
add the build directory to the classpath, remove the -jar jython.jar
reference and give the org.python.util.jython class as the startup
class explicitly.  I've attached the script I use.

Charlie

On 3/5/07, Eric Dobbs <eric@...> wrote:
> (originally sent this only to Frank and not to the list... sorry for
> the delay)
>
> James, jython's test-code-test rhythm feels pretty slow to me too.
> Frank, I was using the default developer build at the sprint at PyCon
> -- still felt too slow to me.  I asked Charlie about writing JUnit
> tests while we were at the sprint.  He mostly persuaded me that the
> python tests were better, for all the reasons already presented in
> this thread.
>
> Charlie's rhythm sounds faster, but I haven't got my environment
> configured to match his yet.  It's been three years since I worked in
> Java and some of my own slowness is just getting re-acquainted with
> an IDE and such.
>
> Anyway he's got a custom shell script for running jython development
> builds.  They run with ./build in the classpath.  Eclipse compiles
> changes directly into ./build, so he doesn't run the developer build
> at all.  He edits java code in Eclipse and then runs jython unit
> tests through his custom shell script.  Pretty fast cycle.  Charlie,
> will you post your script please?  And maybe some notes about how you
> configured Eclipse?
>
> Take care.
> -Eric
>
> -------------------------------------------------------------------------
> Take Surveys. Earn Cash. Influence the Future of IT
> Join SourceForge.net's Techsay panel and you'll get the chance to share your
> opinions on IT & business topics through brief surveys-and earn cash
> http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
> _______________________________________________
> Jython-dev mailing list
> Jython-dev@...
> https://lists.sourceforge.net/lists/listinfo/jython-dev
>