[Grinder-use] thread-save reading of testdata from csv file

[Thomas F. <tho...@lp...>]

Hi all,

this is not particulary grinder-specific, it's rather a jython-question but I guess some of you might have come across this before. 

In my script I use a simple filereader to read values for the tests:


from threading import Condition
from threading import Lock

#at the top of the script:
f = codecs.open(csvFile, 'r',encoding='iso-8859-1')

 def __call__(self):
 ...

       lock = Lock()
       lock.acquire()
       csvValues = f.readline().rstrip("\r\n").split(',')
       lock.release()
 ...

So from my understanding access to the file-object should be synchronized between the worker-threads. However, this does not work reliably. Sometimes the csvValues are just wrong or do not contain all bytes. Before I dig any deeper into this and write a testcase and test different mechanisms of the threading module, I would like to know if any of you know cause and the solution to this issue. 

How have you implemented test-data reading? Maybe it would be better to use the csv module, but I guess that would not solve the problem. 

Thanks
Thomas