jmeter-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From sebb <seb...@gmail.com>
Subject Re: Groovy
Date Wed, 03 Jul 2013 23:06:51 GMT
On 3 July 2013 09:29, Danny Lade <D.Lade@bigpoint.net> wrote:
> Just an idea:
> What about some kind of additional bundle one can install if he want to?
> So you wouldn't have to decide what you bundle with JMeter but can deliver an easy way
for customers to add plugins if they like.
>
> For the "groovy thing":
> I would really wish to have scripts that are cached. But currently (using 2.9) this functionality
can't be used for Java or beanshell.
> As I found out it is deactivated in the code to "handle some bug". We would use scripts
much more often , but they are currently
> not useful for massive load tests (but for setup). For example we often manipulate JSON
input for REST requests.

So why not convert the Groovy script to a Java test element?

> An investigation showed:
> Each script is loading classes from classpath during runtime which is synchronized over
all threads - a huge lack of performance.
Only the first time a class is loaded; not sure that can be avoided,
though of course you can "warm up" your test by running a few threads
first before adding the main load.

>
> So IMO there are two ways:
> * either solve the problems caching scripts at all

If possible, that would the the way to go.

> * or bundle groovy which seems to allow caching

Including Groovy in the JMeter download won't change the behaviour
from downloading it separately.
In both cases the jar is present in the classpath; how can that affect caching?

> JM2C Danny
>
>> -----Urspr√ľngliche Nachricht-----
>> Von: Philippe Mouawad [mailto:philippe.mouawad@gmail.com]
>> Gesendet: Mittwoch, 3. Juli 2013 08:23
>> An: dev@jmeter.apache.org
>> Betreff: Re: Groovy
>>
>> In my experience Scripting elements are more than intended for prototyping .
>> In my 10 last recent scripting experience I have used them at least once. I have
>> met plenty of cases where you absolutely need them ( or you need to setup an
>> eclipse project with jmeter dependencies to finally generate a jar, I am sure
>> plenty of testers wouldn't do it).
>> I also feel it is one of JMeter best feature and gives it a real flexibility.
>>
>> With compilable they have nearly as good performance as if a jar had been
>> created with java classes.
>>
>> For the argument about having more jars in classpath, the perf loss is really very
>> tiny and most of it at startup and on class load.
>> And I feel it goes against reusing in that case.
>>
>> As for beanshell, ok but I think it will take some time before having the level of
>> features that groovy has.
>>
>> Regards
>> Philippe
>>
>> On Wednesday, July 3, 2013, Shmuel Krakower wrote:
>>
>> > Maybe scripting elements are intended for prototyping but from my
>> > experience I endup using few with most of the test plans I create, as
>> > I always miss some specific functionality, which can be answered by
>> > those script elements.
>> >
>> > Btw - what was the need when it was decided to add bean shell jars to
>> > jmeter?
>> >
>> > I am not for adding groovy or against.
>> > On Jul 3, 2013 1:49 AM, "sebb" <sebbaz@gmail.com <javascript:;>>
wrote:
>> >
>> > > On 2 July 2013 23:26, Philippe Mouawad
>> > > <philippe.mouawad@gmail.com<javascript:;>>
>> > wrote:
>> > > > On Wednesday, July 3, 2013, sebb wrote:
>> > > >
>> > > >> On 2 July 2013 22:43, Philippe Mouawad
>> > > >> <philippe.mouawad@gmail.com<javascript:;>
>> > > <javascript:;>>
>> > > >> wrote:
>> > > >> > hello,
>> > > >> > so we include it ?
>> > > >> >
>> > > >> > It's +1 for me:
>> > > >> > - faster than beanshell
>> > > >> > - maintaned and jdk6 syntax can be used
>> > > >> > - big community
>> > > >>
>> > > >> None of the above require it to be bundled with JMeter.
>> > > >>
>> > > >> > - embedding it will encourage its usage
>> > > >>
>> > > >> Possibly; I'm not convinced it's necessary to include it.
>> > > >>
>> > > >> I think we need to be very careful about adding stuff to JMeter
>> > > >> just because it may be useful to some users. It's all to easy
to
>> > > >> end up with a bloated product that performs poorly.
>> > > >>
>> > > >> on the contrary, adding it will encourage groovy usage instead
of
>> > > > beanshell , and this will give better performances.
>> > >
>> > > Only for test plans that use scripting, but having more jars in the
>> > > classpath can have an adverse effect on all tests.
>> > >
>> > > Besides, the scripting test elements are primarily intended for
>> > > prototyping.
>> > >
>> > > Furthermore, BeanShell is hopefully joining Apache Commons soon.
>> > > It may well end up as fast or faster than Groovy.
>> > >
>> > > >> > Regards
>> > > >> > Philippe
>> > > >> >
>> > > >> > On Sunday, June 30, 2013, Philippe Mouawad wrote:
>> > > >> >
>> > > >> >> Hello Sebb,
>> > > >> >> It's Apache 2 license
>> > > >> >> groovy-all.jar is around 6m
>> > > >> >>
>> > > >> >> Regards
>> > > >> >> Philippe
>> > > >> >> On Sun, Jun 30, 2013 at 3:41 PM, sebb
>> > > >> >> <sebbaz@gmail.com<javascript:;>
>> > <javascript:;><javascript:_e({},
>> > > >> 'cvml', 'sebbaz@gmail.com <javascript:;> <javascript:;>');>
>> > > >> >> > wrote:
>> > > >> >>
>> > > >> >>> On 29 June 2013 22:57, Philippe Mouawad <
>> > philippe.mouawad@gmail.com <javascript:;>
>> > > <javascript:;><javascript:_e({},
>> > > >> 'cvml', 'philippe.mouawad@gmail.com
>> > > >> <javascript:;><javascript:;>');>>
>> > > >> >>> wrote:
>> > > >> >>> > Hello,
>> > > >> >>> > what about embedding groovy in next release
so that it is
>> > > >> automatically
>> > > >> >>> > available to JSR223 sampler ?
>> > > >> >>>
>> > > >> >>> How big is it?
>> > > >> >>> What is the license?
>> > > >> >>>
>> > > >> >>> >
>> > > >> >>> >
>> > > >> >>> > Regards
>> > > >> >>> > Philippe
>> > > >> >>> >
>> > > >> >>> > --
>> > > >> >>> > Cordialement.
>> > > >> >>> > Philippe Mouawad.
>> > > >> >>>
>> > > >> >>
>> > > >> >>
>> > > >> >>
>> > > >> >> --
>> > > >> >> Cordialement.
>> > > >> >> Philippe Mouawad.
>> > > >> >>
>> > > >> >>
>> > > >> >>
>> > > >> >
>> > > >> > --
>> > > >> > Cordialement.
>> > > >> > Philippe Mouawad.
>> > > >>
>> > > >
>> > > >
>> > > > --
>> > > > Cordialement.
>> > > > Philippe Mouawad.
>> > >
>> >
>>
>>
>> --
>> Cordialement.
>> Philippe Mouawad.

Mime
View raw message