jmeter-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From sebb <seb...@gmail.com>
Subject Re: Random extraction in csv dataset
Date Thu, 05 Sep 2013 08:22:23 GMT
On 5 September 2013 05:29, Philippe Mouawad <philippe.mouawad@gmail.com> wrote:
> can't we open the file in random access and go to a line ?
> Kind of RandomAccessFile + search newline character ?
> Not sure it works

Only if the file has fixed length records, which is a bit of a restriction.
Alternatively, the user or JMeter would have to create an index file
of fixed length records that point to the start of each record before
starting the test run.

As I wrote already, it's much better to solve this before starting the
JMeter test.

> On Wednesday, September 4, 2013, sebb wrote:
>
>> On 4 September 2013 09:27, Philippe Mouawad <philippe.mouawad@gmail.com<javascript:;>>
>> wrote:
>> > One useful case is distributed testing where you don't want to bother
>> > splitting files for each injector.
>> > Without this all injector will process lines in same order.
>> > With it, no need for splitting.
>>
>> Separately randomised files may still result in two servers processing
>> the same entry.
>> Probably not at the same time, though that is not impossible either.
>>
>> The larger the input file, the less likely it is that the same entry
>> will be processed more than once, but the larger the file, the larger
>> the processing overheads.
>>
>> > of course I am not talking about uniqueness which is another matter.
>> >
>> > As always, it is a productivity feature, of course you can do it by hand.
>>
>> The problem is that randomising the lines at run-time is bound to
>> increase the CPU and/or memory requirements.
>>
>> I think it would be a mistake to include this feature in standard JMeter.
>>
>> > Regards
>> > Philippe
>> >
>> > On Tuesday, September 3, 2013, sebb wrote:
>> >
>> >> On 3 September 2013 07:45, Philippe Mouawad <philippe.mouawad@gmail.com<javascript:;>
>> <javascript:;>>
>> >> wrote:
>> >> > Hello,
>> >> > I think it would be useful to implement a mechanism to extract a line
>> >> > randomly from csv,
>> >> > I often meet this case
>> >>
>> >> What is the use case?
>> >>
>> >> > and it's annoying to handle, it would be nice to
>> >> > have it oob.
>> >> >
>> >> > One problem is how to ensure uniquness if it's another requirement.
>> >>
>> >> That's easy. Just randomise the CSV file before starting the test run.
>> >>
>> >> That's how String From File came about.
>> >> We had a long-running test that used millions of PINs that could only
>> >> be used once, so we randomised them and split them into separate files
>> >> which could be used once.
>> >>
>> >> > Work with a modulo per thread or something like it ?
>> >> > compute at start chunks of values ?
>> >> >
>> >> >
>> >> >
>> >> >
>> >> > --
>> >> > Cordialement.
>> >> > Philippe Mouawad.
>> >>
>> >
>> >
>> > --
>> > Cordialement.
>> > Philippe Mouawad.
>>
>
>
> --
> Cordialement.
> Philippe Mouawad.

Mime
View raw message