IObject classes must give unique filenames for files produced by the objects

[mbakholdina]

As an option, a unique ID or timestamp can be inserted into filename. A code example can be found here.

Update unit tests as well.

[mbakholdina]

As a temporary solution, I've added an opportunity to pass file prefix into object config while creating tasks in SingleExperimentRunner constructor. This allows adding task key prefix as well as any other additional information, e.g., 1-snd1, 1-snd2, etc. will be helpful in future when one task will be responsible for starting several senders simultaneously. 1 here means task 1, snd1 - sender number 1 out of N. As a result, we now have the following filenames:

1-tshark-tracefile.pcapng
2-srt-xtransmit-stats-snd.csv
2-snd1-srt-xtransmit-stats-snd.csv
2-snd2-srt-xtransmit-stats-snd.csv

All these output files will be stored stored in the directory specified in task config (dirpath), e.g., _results. Next time when running the script with the same config and the same value of dirpath, all the files will be rewritten correspondingly.

To guarantee these files not to be overwritten, we should insert unique file id into the filename. There are two options to address this:

1. Either generate unique id in SingleExperimentRunner constructor, e.g., UUID based on the server configuration and timestamp, and insert it as postfix into the filename during the IObject object initiation. This will not require checking the files present on the remote machine, but we will not be able to guarantee 100% unique id in this case,

2. Or store the filename pattern in IObject object and implement the logic of checking whether the filename is unique and changing the filename in IObjectRunner classes. This should be implemented in start() method of IObjectRunner classes write before calling IObject make_args() and make_str() functions. Minus: this implementation will require changing the IObject object attributes after the object has been created and may lead to possible errors with filenames. Every time before calling make_args() and make_str() functions, you should remember about checking the filename and ideally this should be done only once.

def unique_filepath(dirpath, pattern):
    counter = 0
    while True:
        counter += 1
        filepath = dirpath / pattern.format(counter)
        if not filepath.exists():
            return filepath

[mbakholdina]

Once PR #39 will be merged:

Filenames are no longer created by the library and should be specified by the user within an experiment config. They must be unique, otherwise several applications might write in the same file and results would be messed up. Ideally there is a need for config validation before running an experiment. It would be nice to validate the config for such an error and throw an error message at the very beginning.